SECURITIZATION, THE BANK LENDING CHANNEL AND ASYMMETRIC MONETARY TRANSMISSION A Thesis Presented to the faculty of the Department of Economics California State University, Sacramento Submitted in partial satisfaction of the requirements for the degree of MASTER OF ARTS in Economics by Eric Mayes SPRING 2012 SECURITIZATION, THE BANK LENDING CHANNEL AND ASYMMETRIC MONETARY TRANSMISSION A Thesis by Eric Mayes Approved by: __________________________________, Committee Chair Kristin Van Gaasbeck __________________________________, Second Reader Ta-Chen Wang ____________________________ Date ii Student: Eric Mayes I certify that this student has met the requirements for format contained in the University format manual, and that this thesis is suitable for shelving in the Library and credit is to be awarded for the thesis. __________________________, Graduate Coordinator Kristin Kiesel Department of Economics iii ___________________ Date Abstract of SECURITIZATION, THE BANK LENDING CHANNEL AND ASYMMETRIC MONETARY TRANSMISSION by Eric Mayes Using a bank lending channel framework, this study investigates how securitization affects the efficacy of monetary policy. The effectiveness of expansionary and contractionary policies is evaluated separately on the loan growth of banks divided by asset size, capital adequacy, and securitized asset holdings. There is evidence that securitization weakens the bank-lending channel but this result is not robust across all specifications. The implication of the analysis is that the effect of securitization on the efficacy of monetary policy depends greatly on the structure of the financial system. _______________________, Committee Chair Kristin Van Gaasbeck _______________________ Date iv TABLE OF CONTENTS Page List of Tables .............................................................................................................. vi List of Figures ............................................................................................................ vii Chapter 1. INTRODUCTION .................................................................................................. 1 2. SECURITIZATION IN THE U.S. FINANCIAL SYSTEM .................................. 5 3. LITERATURE REVIEW ....................................................................................... 8 Monetary Transmission .................................................................................... 8 Securitization .................................................................................................. 11 Asymmetric Policy Effects…...........................................................................13 4. DATA AND EMPIRICAL METHODOLOGY ................................................... 16 Data Sources and Descriptions ....................................................................... 16 Empirical Methodology .................................................................................. 21 5. RESULTS .............................................................................................................. 24 Regression Results .......................................................................................... 24 Robustness Checks.......................................................................................... 31 6. CONCLUSION ...................................................................................................... 34 Appendix A F-Statistics for Differences in Coefficient Sums.................................... 37 Appendix B Model 1 Using Fixed Capital-to-Asset Ratios........................................ 39 References ................................................................................................................... 40 v LIST OF TABLES Tables Page 1. Cumulative Assets of Securitizing and Non-Securitizing Banks……………....6 2. Mean Balance Sheet Ratios (1980Q1-2010Q4)………… ............................... .17 3. Cumulative Bank Assets (1980-2010)…………………………………..…….18 4. Mean Balance Sheet Ratios By Capital………… .. …………………………. 20 5. Mean Balance Sheet Ratios (2001Q1-2010Q4)…………………………….....21 6. Model 1 Using 1996 Break (1980Q1-2010Q4)……………………………….24 7. Model 1 Using 2001 Break (1980Q1-2010Q4)…………………………….…27 8. Model 2 (2001Q1-2010Q4)…………………………………………………...30 9. Degree of Securitization Effects……………………………………………....32 vi LIST OF FIGURES Figures Page 1. Cumulative Assets of ABS and Non-ABS Banks……………………………...7 2. Cumulative Bank Assets (1980-2010)……………………………….… . ……19 vii 1 1. Introduction One of the most important questions in Keynesian macroeconomics is how monetary policy affects the real economy. In the IS-LM model, monetary policy affects market equilibrium in the market for real money balances, which leads to a shift in the LM curve. Shifts in the LM curve change the equilibrium level of output in the IS-LM model, changing aggregate demand at any given price level. This results in a shift in the aggregate demand curve. To better understand this process, economists began studying monetary transmission mechanisms. Monetary transmission mechanisms consist of different ways that monetary policy affects aggregate demand and reaches the real economy (Mishkin, 2007). The traditional view of monetary policy transmission is that the Federal Reserve (Fed) influences the economy through an interest rate channel. In the interest rate channel, contractionary policy leads to an increase in real interest rates because of nominal rigidities, which increases the cost of capital causing a decrease in investment spending. Since investment spending is a component of Gross Domestic Product (GDP) this reduces GDP growth and slows down the economy. An alternative view on monetary transmission is that the Federal Reserve affects the economy through credit channels that arise from asymmetric information in the financial system. When asymmetric information exists in financial markets, banks are necessary financial intermediaries that connect borrowers to lenders that would not otherwise have access to financial markets. Two forms of credit channels are the bank lending channel (BLC) and the balance sheet channel (BSC). In the BSC, monetary 2 policy reaches the real economy by affecting firms’ balance sheets. Contractionary policy can cause a decrease in stock prices which decreases firms’ net worth. This decrease in net worth increases asymmetric information issues which reduce lending and investment spending. In the BLC banks serve as intermediaries connecting borrowers to credit markets. Expansionary monetary policy operates by increasing bank reserves and deposits which results in banks making more loans. This increases investment and consumer spending as borrowers use bank loans to finance their expenditures. In the BLC, small firms are more sensitive to monetary policy because they depend more on bank finance than large firms which can more easily acquire funds through stock and bond markets. Even in welldeveloped financial systems, like those in the U.S., Europe, and Japan, banks and other financial intermediaries remain the most important source of borrowing for households and firms. The transmission of monetary policy through the BLC potentially magnifies the effects of monetary policy and redistributes wealth, as households, small banks, and small firms are disproportionately affected by the contraction in credit that accompanies monetary policy tightening. This study focuses on the BLC and how the strength of the BLC varies between banks based on asset size and capital strength as well as how the rise of securitization has impacted the bank lending channel. Existing literature has uncovered bank capital as an important constraint in determining a BLC. Contractionary policy has been shown to be effective in reducing the loan growth of capital-constrained banks. This is because these banks must fund lending primarily with deposits and bank reserves, a reduction in 3 reserves due to contractionary monetary policy greatly reducing their ability to generate new loans. In contrast, contractionary policy should be much less effective in reducing the loan growth of high-capital banks. Banks holding large amounts of capital can more easily fund loans when reserves or deposits decrease. The large growth of securitization taking place since the 1970s has important implications for the BLC. The ability of banks to combine loans into asset-backed securities has changed the business model of some banks from “originate and hold” to “originate, repackage and sell” (Altumbas, Fazylov & Molyneux, 2009). In traditional banking, banks had to finance illiquid loans with liquid deposits and bank reserves. This created a business model of originating loans and then holding on to loans. The ability to package loans into securities has allowed banks to originate loans and then sell loans for cash, making them much less dependent on traditional sources of funding. It follows that with the rise of securitization the BLC may become weaker. By securitizing and selling loans banks acquire and additional sources of funding to make more loans independent of traditional sources. The effect of widespread asset-backed securities has important implications for monetary policy. As more and more banks securitize loans the financial system as a whole can become less responsive to policy actions taken by the Federal Reserve. This will likely make managing the business cycle more difficult for the Fed as it will have to take more aggressive open market operations to influence an economy full of banks using asset-backed securities to fund their lending. 4 This study contributes to the BLC literature in two ways. First, the use of more recent data will allow for investigation into how banks respond to monetary policy through the BLC has changed in recent years. Existing literature has shown banks behave differently in the BLC based on their asset size and capital-to-assets ratio, this study will follow suit by dividing banks into high and low capital categories as well as three asset size categories. Second, the study will examine the impact of increased securitization on the BLC by dividing the sample into two eras, the period from 1980 to when collateralized debt obligations were developed in 1996, and from 1996 to 2010. In addition, the sample will be divided into banks that securitized their assets and those that did not in order to contrast the strength of the BLC between these two categories in addition to capital strength and asset size. The rest of this paper is organized as follows; Section 2 presents background information on the growth of asset-backed securities in U.S. financial system, Section 3 presents review of literature on the topics studied, Section 4 describes the data and empirical methodologies used, Section 5 reports results and section 6 concludes 5 2. Securitization in the U.S. Financial System The reasons a bank creates asset-backed securities could be to increase liquidity in its balance sheet or reduce regulatory capital requirements. The process begins with a bank originating a financial asset such as a home mortgage, auto loan or credit card receivable. The asset is then combined with other similar assets into a security that produces a return equal to the aggregate income of the pooled assets. This asset-backed security can then be sold to an investor and the cash used to fund more lending activities. Securitization began in home mortgage markets during the 1970s and has continued to grow with the trend intensifying in the 1980s and after 1996 (Estrella, 2002). Estrella (2002) identified the government’s sponsorship of the Federal Home Loan Mortgage Corporation, the Federal National Mortgage Association and the Government National Mortgage Association. These organizations issued securities consisting of pooled home mortgages that were originated by banks. In 1996 collateralized debt obligations were developed consisting of collateralized loan obligations and collateralized bond obligations. This intensified the growth of asset-backed securities which now are composed of not just mortgages but auto loans, student loans, credit card receivables, commercial and industrial loans as well as others (Estrella, 2002). Table 1 presents the cumulative total assets of banks that chose to securitize their assets (ABS Banks) and non-securitizing banks (Non-ABS Banks) from 2001-2007. The values in the table are the sum of all banks’ total assets in each category with the actual number of banks in parenthesis. Comparing cumulative assets and the quantities of banks in both categories it is immediately evident that banks that choose to securitize their 6 assets tend to be much larger than banks that do not. This fact is consistent with Adrian and Shin (2010), which showed that larger institutions were more likely to have the resources to create asset-backed securities. Table 1 also shows the asset share of securitizing banks surpassing non-securitizing banks in 2002 and continuing to grow at a faster rate through 2007. These trends are also visible in figure 1 which plots cumulative assets of the two bank categories over time. Table 1 Cumulative Assets of Securitizing and Non-Securitizing Banks Year ABS Banks Non-ABS Banks 2001 4075873 (181) 4238556 (8817) 2002 4515536 (170) 4324773 (8581) 2003 4891480 (152) 4583138 (8457) 2004 5719328 (134) 4849556 (8297) 2005 6148997 (135) 5427798 (8166) 2006 7120634 (132) 5949914 (8106) 2007 8127334 (130) 6621802 (7967) Thousands of Real Dollars 7 Figure 1 Cumulative Assets of ABS and Non-ABS Banks Cummulative Assets of Securitizing and Non-Securitizing Banks (Thousands of Real Dollars) 9000000 8000000 7000000 6000000 ABS Banks 5000000 Non-ABS Banks 4000000 3000000 2001 2002 2003 2004 2005 2006 2007 Widespread securitization gave birth to the “shadow banking system” which played a major role in the financial crises beginning in 2007 (Adrian & Shin, 2010). Adrian and Shin (2010) also determined that securitization increased the fragility of the financial system by allowing banks to leverage themselves by buying each other’s assetbacked securities. The growing share of total assets belonging to securitizing banks highlights the importance of studying how they respond differently to monetary policy. If securitizing banks are less responsive to monetary policy then these trends suggests the Fed should become more aggressive in open market operations, adjusting its strategy as the structure of the banking system changes. 8 3. Literature Review Monetary Transmission According to traditional Keynesian and neoclassical macroeconomic theory, monetary policy affects real economic variables in the short run through the interest rate, or money channel. Specially, the central bank controls the supply of money and because of nominal rigidities; changes in money supply affect nominal and real interest rates. In addition to the interest rate channel of monetary transmission, economists have completed substantial research identifying alternative transmission mechanisms. These alternative mechanisms include credit channels such as the bank lending channel (BLC). With the significant changes in the financial system in the late 1990s and 2000s, and the 2007-2009 financial crisis and recession, these channels of monetary transmission continue to receive attention in the literature. Bernanke and Blinder (1992) set early foundations for subsequent research on monetary transmission. The authors identify the federal funds rate as informative of the stance of monetary policy and link its changes to macroeconomic conditions. The federal funds rate is now used extensively as a measurement of the Federal Reserve’s policy stance. The authors also identified two channels of monetary transmission through bank loans and bank assets. In their analysis, contractionary policy causes banks to sell their security holdings and become reluctant to make new loans which have a negative effect on the real economy. This thesis uses the federal funds rate to indicate policy stance, but differs from Bernanke and Blinder (1992) in that we consider the possibility of asymmetry in policy shocks. 9 Kashyap and Stein (2000) use the 1976-1993 Report of Condition and Income Data (call reports) in order to identify a BLC in that period. The authors find the BLC is stronger among banks with low ratios of cash and securities to assets and these banks are mainly smaller banks in the sample. The authors use of a panel data set provided strong evidence in favor of the existence of a BLC, this was important because previous studies using aggregate data were unable to rule out the possibility of an interest rate channel. Results from Kashyap and Stein (2000) demonstrated the distributional effects of monetary policy through the BLC. For this reason, we estimate the strength of the BLC across banks based on size (measured by total assets). Kishan and Opiela (2000) also used the call reports to study the BLC. The authors divide banks by asset size and capital strength and find significant differences in BLC strength across their sample. The banks were divided into six asset size categories and three capital strength categories based on capital-to-asset ratios, undercapitalized (ratio<8%), adequately capitalized (8%<ratio<10%), and well capitalized (ratio>10%). The authors reasoned that the amount of equity capital bank held would impact its ability to finance loans and change its sensitivity to monetary policy. In contrast to Kashyap and Stein (2000) which excluded banks involved in mergers from the sample, Kishan and Opiela (2000) performed their analysis using both merger-adjusted and non-mergeradjusted versions of the sample. The merger-adjusted data was accomplished using a method employed by Peek and Rosengren (1995), which treats merged banks as single banks during the whole sample as if they merged at the beginning. This study used this approach as a robustness check and found no significant difference in results. The 10 authors’ find that the BLC is strongest among small undercapitalized banks, a result similar to Kashyap and Stein (2000) which found the BLC was strongest among small illiquid banks. The authors’ result that undercapitalized banks were more responsive to policy has important implications for the banking system as a whole. As regulatory changes call for banks to hold more capital the banking system may become less responsive to monetary policy. These findings indicate that it is not only important to consider bank size, but also bank capitalization when measuring the BLC. We similarly find that bank capitalization matters for monetary transmission through the BLC, when identified by capital-asset ratios. Ashcraft (2006) finds evidence consistent with Kashyap and Stein (2000) as well as Kishan and Opiela (2000) in that a bank lending channel exists and its strength varies between banks of different sizes and capital strength. However Ashcraft also casts some doubt on the importance of the BLC to the macroeconomy. Using aggregate data, the author estimates the elasticity of loan growth to output growth and finds relatively inelastic estimates. The results of the study suggest that the BLC may not be the whole picture when it comes to monetary transmission. This is consistent with an earlier survey by Kashyap and Stein (1994) which finds that the aggregate effects of the BLC are small in magnitude. The research surveyed so far has investigated bank lending channels in the United States; economists have also studied monetary transmission internationally. Using bank level data from European countries, Altumbas et al. (2002) identified a bank lending channel and found evidence consistent with Kishan and Opiela (2000) where 11 undercapitalized banks were more responsive to monetary policy. Using bank level data from 1994-2003, Matousek and Sarantis (2009) find further evidence of a BLC and obtain results suggesting that liquidity and banks size were most important in determining loan growth. Data on Polish banks revealed that the strength of the BLC vary by levels of deposit guarantees (Opiela, 2008). Bhaumik, Dang, and Kutan (2011) use bank level data from India to study how different types of banks responded to monetary policy shocks. The authors found that bank ownership type played a significant role in credit channel strength. The authors also determined that the BLC was more pronounced during periods of contractionary policy then expansionary periods. Securitization Estrella (2002) uses aggregate data to investigate the growth of securitization in the U.S. financial system and examine how the spread of asset-backed securities has influenced the efficacy of monetary policy. The author charts the growth of securitization in the in the United State from 1965 to 2000. The growth of securitization intensified in the 1980s and in 1996 when collateralized debt obligations were developed. The author uses a simple IS curve estimation with aggregate data to assess the effect of securitization on monetary policy efficacy. Estrella (2002) found that securitization growth significantly reduced the elasticity of output with respect to the federal funds rate. In addition, the researcher found results showing mortgage interest rates more closely following the federal funds rate as the value of securitized assets grew in the economy. 12 Estrella reasons this finding means that the reduction in policy effectiveness is due to changes in liquidity and credit channel transmission mechanisms. His research suggests that securitization has significant implications for monetary policy through a variety of channels. Loutskina and Strahan (2009) exploited Fannie Mae and Freddie Mac’s inability to purchase jumbo mortgages to study how securitization has affected banks’ lending behavior. The authors contrast traditional banking where banks originate and hold loans with banking under securitization which allows banks to originate loans and then distribute them. Traditional banks had to fund illiquid loans with liquid deposits which resulted in deposits and balance sheet liquidity determining the supply of credit. The authors report that securitizing banks rely only on deposits to make loans and that a bank’s financial condition has no impact on their supply of credit. The result that securitizing banks rely mostly on deposits to make loans contrasts with the behavior of non-securitizing banks. This study will divide banks into securitizing and nonsecuritizing categories to investigate if this behavior difference affects the BLC. Altumbas, Gambacorta and Marques-Ibanez (2009) use data on European banks to study securitization’s effect on the BLC. The authors used a unique data set which allowed investigating how banks that securitize their assets respond differently from banks that did not. Their results revealed that securitizing banks were less responsive to monetary policy. Altumbas et al. (2009) also found banks that securitized made more loans and this result was more pronounced during healthy macroeconomic conditions. Loutskina and Strahan (2009) finding that securitizing banks rely mostly on deposits 13 could explain why these banks mad more loans during healthy macroeconomic periods as deposits increase during economic expansions. Aysun and Hepp (2011) use call report data to study how securitizing and nonsecuritizing banks respond differently to monetary policy using balance sheet channel framework. In contrast to Altumbas et al. (2009) which found securitization weakened the BLC, Aysun and Hepp (2011) find that securitization strengthens the BSC. In their analysis, Aysun and Hepp (2011) no significant difference between securitizing and nonsecuritizing banks but found that the degree of securitization was positively related to banks responsiveness to monetary policy. Their results suggest that securitization has had an ambiguous effect on monetary transmission. To account for the result of Aysun and Hepp (2011) this study will also restrict the sample to securitizing banks and examine how the degree of securitization has impacted the BLC. Asymmetric Policy Effects While economists have studied asymmetric monetary policy effects for some time, only recently have they investigated asymmetric effects to alterative transmission mechanisms. Cover (1992) used aggregate data to examine asymmetric effects of money supply shocks. Cover found that positive money shocks had no effect on the real economy while negative shocks had a negative effect on the real economy. This result is similar to Bhaumik et al. (2011) which found that the BLC was more pronounced during contractionary policy than expansionary. In contrast to Cover (1992), Weise (1999) found contractionary and expansionary monetary policy to have nearly symmetric effects in 14 magnitude. Weise (1999) used an autoregression approach with nonlinear vectors using aggregate data. Given the mixed evidence on the potential for asymmetric effects of monetary policy, this thesis allows for differences in the effects of contractionary and expansionary monetary policy. Kishan and Opiela (2006) studied asymmetric effects of monetary policy on the BLC. Using call report data the authors split the sample into six asset size categories and two capital strength categories. Using banks’ capital-to-assets ratios, those with ratios<8% were undercapitalized and those with ratios>8% were well capitalized. The researchers found that contractionary policy was most effective in reducing the loan growth of small, undercapitalized banks while expansionary policy was most effective in increasing the loan growth of high capital banks. In addition, the authors divided there sample into pre-Basel and post-Basel eras pertaining to the implantation of the Basel 1 capital regulations. The authors noted that their results only held in the post-Basel era and concluded that stringent capital constraint enforcement was the reason for discrepancy between the two eras. This study extends the literature by combining investigations into the bank lending channel, securitization and asymmetric policy effects. Expansionary and contractionary monetary policy stances are isolated and how securitization has impacted the efficacy of these policy stances will be investigated. The following four hypotheses are tested in the analysis: 15 (1) Among low-capital banks, contractionary policy should be more effective in reducing loan growth than expansionary policy is in increasing loan growth. (2) Among high-capital banks, expansionary policy should be more effective in increasing loan growth than contractionary policy is in reducing loan growth. (3) After collateralized debt obligations were developed in 1996, both expansionary and contractionary policy should be less effective in influencing the loan growth of all banks. (4) Banks that securitize their assets should be less responsive to both contractionary and expansionary policy. 16 4. Data and Empirical Methodology Data Sources and Descriptions To test the four hypotheses outlined in the previous section, the study uses a combination of aggregate and bank level data. All aggregate data was taken from the Federal Reserve Bank of St. Louis’s Federal Reserve Economics Database (FRED), while bank level data were obtained from the Report of Condition and Income (call reports) that all insured banks are required to submit to the Federal Reserve. The data set is recorded in quarterly intervals from the first quarter of 1980 to the fourth quarter of 2010. All balance sheet variables are adjusted for inflation using the Consumer Price Index. This study uses methods to distinguish a BLC employed by Kishan and Opiela (2006), dividing the sample by bank size and capital constraint. Banks are split into three asset size categories; large (total assets>$1 billion), medium ($1 billion>total assets>$100 million) and small (total assets<$100 million). In order to examine the BLC it is also necessary to divide banks by capital constraint (Kishan and Opiela, 2000; Altunbus et al., 2002; Kishan and Opiela, 2006; Matousek and Sarantis, 2009). Using their capital-tototal asset ratios, banks are divided into two categories; constrained banks with a ratio greater than 8% and unconstrained banks with a ratio less than 8%. 17 Table 2 Mean Balance Sheet Ratios (1980Q1-2010Q4) Large Banks Medium Banks Small Banks Total Loans/Assets .5869 .5870 .5661 Real Estate Loans/Assets .2847 .2847 .2860 Consumer Loans/Assets .1124 .1128 .0972 Large Time .1155 .1154 .1110 Total Loans/Assets .5551 .5316 .5662 Real Estate Loans/Assets .2867 .2866 .2861 Consumer Loans/Assets .0891 .0894 .0972 Large Time .1086 .1085 .1117 Capital Ratio<8% Deposits/Assets Capital Ratio>8% Deposits/Assets Table 2 presents mean balance sheet ratios for large, medium and small banks divided by capital to total assets ratio. Large and medium sized banks with high capital to total asset ratios tended to have smaller total loans to assets ratios, as well as and less consumer loans relative to their total assets. However this trend is not visible with smaller banks which varied very little with response to capital to asset ratio. Cumulative total assets for banks in the three asset size categories from 19802010 is displayed in table 3. In 1980 banks in the small category controlled the majority of assets in the banking followed by medium banks then large banks; however, the relative asset shares changed dramatically over time. The cumulative assets of small and medium banks both increased steadily until 2000 when they begin to taper off. Large 18 banks on the other hand, increased at a rate similar to medium and small banks until 1995, when the growth of their cumulative assets increased dramatically. It is possible that the large increase in the cumulative assets of large banks is a result of securitization. It was discussed in section 2 of this study that securitizing banks tend to be much larger than non-securitizing banks and the growth of asset-backed securities increased rapidly after collateralized debt obligations were developed in 1996. Large securitizing banks would be able to increase the size of their balance sheets quickly by financing loans with asset-backed security sales. Unfortunately data on securitized assets are only available in the call reports after 2000, preventing any in depth analysis of securitizations role in the growth of large banks’ asset share after 1995. Table 3 Cumulative Bank Assets (1980-2010) Year Small Medium Large 1980 1434512 690304 433199 1985 2018824 1121523 603290 1990 2726674 1694430 792223 1995 3139996 1975441 1338404 2000 3209028 1756916 3257754 2005 3268455 1915595 5680753 2007 3422783 1892225 7011202 2010 3111419 1761482 8943852 Thousands of Real Dollars 19 Figure 2 Cumulative Bank Assets (1980-2010) Cumulative Bank Assets (Thousands of Real Dollars) 10000000 9000000 8000000 7000000 6000000 Small 5000000 Medium 4000000 Large 3000000 2000000 1000000 0 1980 1985 1990 1995 2000 2005 2007 2010 In order to examine the effects of securitization the sample is further divided into pre-1996/CDO eras and post-1996/CDO eras to examine the effect of the large growth in asset backed securities after the development of collateralized debt obligations in 1996. In addition, the sample was also divided into pre-2001 and post-2001 eras to examine the effects of when asset-backed securities were required in the call reports. To investigate the effects of securitization more closely the sample was also divided into banks that held asset-backed securities and banks that did not, allowing for a more in depth analysis of behavior differences between securitizing and non-securitizing banks. 20 Table 4 Mean Balance Sheet Ratios by Capital Capital/Assets <8% Capital/Assets >8% Total Loans/Assets .5473 .5594 Real Estate Loans/Assets .2473 .2750 Consumer Loans/Assets .1094 .0994 Large Time Deposits/Assets .1032 .1090 Total Loans/Assets .5984 .5648 Real Estate Loans/Assets .3283 .3005 Consumer Loans/Assets .0919 .0873 Large Time Deposits/Assets .1221 .1113 Pre-1996/CDO Post-1996/CDO Table 4 contrasts mean balance sheet ratios between low and high capital banks during the pre-1996/CDO and Post-1996/CDO eras. There is little noticeable difference between low capital and high capital banks. Moving between the two eras both classifications slightly increased their total loans to asset ratios while real estate loans became a larger proportion of total assets relative to consumer loans. Thus, securitization lead to an expansion of mortgages that is statistically significant. Total loans/assets and large time deposits/assets ratios for ABS and Non-ABS banks are reported in table 5. The table does not show any real differences between the two categories and none of the means were statistically different from each other. 21 Table 5 Mean Balance Sheet Ratios (2001Q1-2010Q4) ABS Banks Non-ABS Banks Total Loans/Assets .6466 .6326 Large Time Deposits/Assets .1415 .1373 Total Loans/Assets .6270 .6315 Large Time Deposits/Assets .1359 .1373 Capital Ratio<8% Capital Ratio>8% Empirical Methodology The analysis uses two dynamic panel regression models to investigate the relationship between loan growth and monetary policy. The first model explores possible structural breaks in 1996 and 2001 and the second model examines how banks that held asset-backed securities respond differently to monetary policy compared to banks with no asset-backed securities. The dependent variable is the growth rate of total loans in both models. The contemporary change in the federal funds rate and four lagged values of the federal funds rate are included as policy regressors. To account for asymmetric responses to the federal funds rate reported by Cover (1992), Morgan (1993), Bliss and Kaufman (2003) as well as Kishan and Opiela (2006), the federal funds rate variable is divided into contractionary and expansionary variables. The contractionary variable consists of positive changes in the federal funds rate or zero otherwise while the expansionary variable contains negative changes in the federal funds rate or zero otherwise. 22 For non-policy regressors, seasonal dummy variables and a time trend dummy are included in both models. To control for demand shocks caused by macroeconomic conditions, contemporary and four lagged values of real GDP growth are included. Kishan and Opiela (2006) identified a structural break affecting the BLC in the third quarter of 1990 caused by the Basel 1 capital requirements. To control for this regulatory change a dummy variable equal to zero before 1990q3 and one afterwards is added to both regressions. Both models also include a dummy variable to control for changes in variable definitions in the first quarter of 1984 and four lagged values of the dependent variable. To mitigate potential endogeneity bias and funding effects, the analysis will employ a two-stage regression process outlined by Kishan and Opiela (2006) which includes unexpected changes in large time deposits and securities as independent variables. In the first stage, the growth in large time deposits and securities are regressed on all other independent variables. In the second stage, the residuals from these regressions are included in the loan growth equations. To examine how the growth of asset-backed securities has impacted the BLC, the first model will use a dummy variable equal to zero for all observations before 1996Q1 and one for all observations after. This date was chosen because of the large increase in asset-backed security growth after collateralized debt obligations were developed in 1996. This dummy variable will then be interacted with the federal funds rate variables to measure the effect on the BLC. In addition to the 1996/CDO break, the study will also check how requiring asset-backed securities to be reported in the call reports affected the 23 BLC. A dummy variable will be set equal to one for all observations before 2001Q1 and equal to one for all observations afterwards and then interacted with all federal funds rate variables. This specification will allow us to compare the strength of the BLC across periods of varying asset-backed security presence. In addition to how the growth of asset-backed securities has affected the BLC, the second model will investigate how the strength of the BLC varies between banks that securitized their assets and those that did not. This analysis restricts the time period to 2001Q1-2010Q4. The restriction is necessary since data on asset-backed securities appears in the call reports only after 2000Q4. To separate the banks a securitization variable was created. This variable is the sum value of securitized assets appearing on the balance sheet including home equity lines, credit card receivables, auto loans, commercial and industrial loans, other consumer loans and all other loans. If the securitization variable is any value greater than zero the bank is categorized as a bank holding securitized assets. The loan growth equation is then regressed again using the new divided sample. This specification will allow us to test whether the BLC’s strength varies between banks the hold asset-backed securities and those that do not. 24 5. Results Regression Results Table 6 presents the results of the first empirical model using the 1996/CDO break. This analysis tests how both contractionary and expansionary policy stances impact the loan growth of banks based on their size and capital strength. The model then tests how the development of collateralized debt obligations in 1996 and the following growth in asset-backed securities affected this relationship. The table presents the sums of coefficients corresponding to changes in the federal funds rate. Expansionary policy measurements consist of the coefficient sums of negative changes in the federal funds rate and contractionary policy measurements are the corresponding sums of positive changes in the federal funds rate. F-Statistics for differences in Coefficient sums can be found in Appendix A. Table 6 Model 1 Using 1996 Break (1980Q1-2010Q4) Policy Pre-1996 Low Stance/Bank Capital Banks Size Large Banks -2.96*** (Expansionary) Large Banks -3.54* (Contractionary) Medium Banks -.812*** (Expansionary) Medium Banks -3.47** (Contractionary) Small Banks -.776*** (Expansionary) Small Banks -3.21** (Contractionary) Significant at 10%*, 5%**, 1%*** Pre-1996 High Capital Banks -2.46*** Post-1996 Low Capital Banks 7.48** Post-1996 High Capital Banks 1.81 8.527* -.533*** -6.51* 3.46*** -.689** -6.61** 4.20*** -.739*** 13.04** 3.05*** 1.077 -1.25* 2.258** -2.53* -1.25 25 Table 6 presents evidence consistent with the first hypothesis, that contractionary policy should have a stronger impact with low-capital banks. In the pre-1996 era the model shows both expansionary and contractionary policy for all banks showing negative and statistically significant relationships. Contractionary coefficient sums were greater in magnitude for all three bank sizes which suggests increases in the federal funds rate were more effective in reducing loan growth than decreases in the federal funds rate were at increasing loan growth. This relationship is more pronounced for small and medium banks than for large banks. The second hypothesis posits that expansionary policy would have a stronger effect then contractionary among high capital banks. Table 6 presents results compatible with this hypothesis for large banks but not for medium or small banks. For large highcapital banks, the model shows expansionary policy having a positive effect on loan growth during the pre-1996 era while banks in this category were able to bypass contractionary policy and continue to increase loan growth. The model shows that both small and medium sized banks were able to bypass both contractionary and expansionary policy. The third hypothesis is that both expansionary and contractionary policy became less effective in controlling loan growth after collateralized debt obligations were developed in 1996. In the post-1996 columns for both low and high capital banks Table 6 presents the sums of federal funds rate coefficients and their interaction terms in the Post-1996 columns for both low and high capital banks. For low capital banks the results are consistent with Hypothesis 3 for all rows except small banks facing expansionary 26 policy. For large and medium sized banks coefficient sums decreased in magnitude after the break, suggesting monetary policy became less effective in controlling loan growth. The estimated model shows large banks being able to bypass expansionary policy entirely by reducing loan growth when the federal funds rate decreases. For high capital banks, the model yielded evidence contrary to hypothesis 3 for all bank sizes. Among large banks, contractionary policy was more effective in reducing loan growth after 1996 if their capital to assets ratio was greater than 8%. Expansionary policy became more effective in stimulating loan growth in both medium and small banks with high capital to asset ratios. For medium sized banks with high capital to asset ratios, the sums of coefficients were significant and positive when facing contractionary policy both before and after 1996. However the coefficient sum after the break is greater in magnitude which may suggest the banks were even better able to bypass monetary policy by increasing loan growth under contractionary policy stances. It also needs to be noted whether first two hypotheses held during the post break era. Table 6 does present some evidence consistent with hypothesis 1. Contractionary policy was more effective than expansionary for both large and medium sized banks although this effect was small for medium banks. For small banks contractionary policy was negative and significant but expansionary policy was not statistically significant. The model shows significant evidence fitting with hypothesis 2 for only medium sized banks. In the post 1996 era the model shows medium banks with high capital-to-asset ratios able to increase loan growth during contractionary policy while expansionary policy was effective in increasing loan growth. Coefficient sums were also significant 27 and negative for large, high capital banks facing contractionary policy and small, high capital banks facing expansionary policy. Although these sums were significant and of the expected sign, not having statistically significant results for large, high-capital banks facing expansionary policy and small, low-capital banks facing contractionary policy gives little guidance for hypothesis 2. Table 7 Model 1 Using 2001 Break (1980Q1-2010Q4) Policy Pre-2001 Low Stance/Bank Capital Banks Size Large Banks -.174*** (Expansionary) Large Banks -.427*** (Contractionary) Medium Banks -.841** (Expansionary) Medium Banks 1.69 (Contractionary) Small Banks .464*** (Expansionary) Small Banks 9.61* (Contractionary) Significant at 10%*, 5%**, 1%*** Pre-2001 High Capital Banks -1.59*** Post-2001 Low Post-2001 Capital Banks High Capital Banks -3.36** 1.87* 5.89*** -1.11* .923 .791** 9.09** .851 2.94** -5.55** -5.82* -1.18*** 3.81* 1.36 4.52** -1.03 0.822 Table 7 presents mixed evidence from the first model using the 2001 break when data on asset backed securities begins being recorded in the call reports. In the Pre-2001 low-capital bank column evidence in favor of hypothesis 1 can be seen only with large banks. For medium sized banks expansionary policy was effective in stimulating loan growth but the coefficient sum was not significant for contractionary policy. The model shows small banks bypassing both expansionary and contractionary policy. Both large 28 and small banks show evidence in favor of hypothesis 2 while medium banks appear able to bypass both expansionary and contractionary policy stances, although the sum of expansionary coefficients being a smaller magnitude then the contractionary sum may suggest hypothesis held for medium banks. The post-2001 era columns show mixed evidence as well. Among low-capital banks, contractionary policy was less effective than expansionary which contrasts with our first hypothesis. However, among medium sized banks with low capital contractionary policy was effective in reducing loan growth while expansionary policy was not effective in increasing loan growth which is consistent with the first hypothesis. For small banks the expansionary coefficient sum was positive and significant while the contractionary sum was not significant. For high-capital banks in the post-2001 era only large banks facing expansionary policy and medium banks facing contractionary policy were statistically significant. Large banks were able to bypass expansionary policy and reduce their loan growth while the model shows contractionary policy being effective in reducing the loan growth of medium sized banks. When contrasting the pre-2001 and post-2001 eras, evidence conflicts with the third hypothesis. The model shows both expansionary and contractionary policy becoming more effective with large low capital banks. However, large high-capital banks became less responsive to expansionary policy and medium sized low-capital banks became less responsive to expansionary policy, both results are inconsistent with hypothesis 3. All other rows in the post-2001 era were not statistically significant. 29 In addition to the first analysis the second analysis investigates how the strength of the BLC varies between banks that held asset-backed securities and those that did not. This model restricts the sample to 2001Q1 to 2010Q4 due to data limitations and restricts the banks to two size categories, large and small. The asset size categories were reduced due the very small presence of medium sized banks during the 2001-2010 periods. Any banks that would have fallen into the medium sized category are in the large category while banks that held asset-backed securities are designated ABS banks and those that did not are Non-ABS banks. The results of the analysis are displayed in table 8. First looking at the large bank rows, there are noticeable differences in coefficient sums for both high and low capital banks with respect to asset-backed securities. Both expansionary and contractionary policy was effective in influencing the loan growth of low capital banks that did not hold asset-backed securities. This contrasts with low capital banks that did hold asset-backed securities which were able to bypass both expansionary and contractionary policy. This result is consistent with hypothesis 4 which stated banks holding asset-backed securities would be less responsive to monetary policy. This result is also visible among large high-capital banks and although it appears less pronounced the coefficient sums are statistically different from each other. Both policy stances had an effect on high-capital banks and the magnitude of the coefficients decreased among those that held asset-backed securities, providing evidence in support of the fourth hypothesis. 30 Table 8 Model 2 (2001Q1-2010Q4) Policy Stance/ Bank Size Low-Capital Non-ABS Banks -2.44** Large Banks (Expansionary) Large Banks -4.55*** (Contractionary Small Banks 5.37*** (Expansionary) Small Banks -3.57** (Contractionary) Significant at 10%*, 5%**, 1%*** Low-Capital ABS Banks High-Capital ABS Banks 4.836** High-Capital Non-ABS Banks -6.03*** 2.97*** -3.40*** -3.029*** 5.134*** -4.357*** 5.023*** -3.146*** -2.99*** -3.102*** -4.834*** The rows containing coefficient sums of small banks show conflicting evidence regards to our fourth hypothesis. No statistically significant difference exists between small low-capital banks with respect to asset-backed securities. The model shows both Non-ABS and ABS banks able to reduce loan growth during expansionary policy and contractionary policy was effective in reducing loan growth. Small banks with high capital-to-asset ratios were much less responsive to expansionary policy if they held asset-backed securities. The results show expansionary policy effective in increasing the loan growth of Non-ABS banks while ABS banks were able to reduce loan growth during expansionary policy, fitting with hypothesis 4. However, like the low-capital small banks there is no statistically significant difference between Non-ABS and ABS high-capital small banks with regards to contractionary policy. Examining coefficient sums for low-capital Non-ABS banks, contractionary policy was more effective in reducing loan growth than expansionary policy was in increasing loan growth, a result consistent with hypothesis 1. Low-capital ABS banks 31 also showed results consistent with hypothesis 1. Small low-capital banks holding assetbacked securities were able to bypass expansionary policy but responded to contractionary. While large low-capital banks holding asset-backed securities were able to bypass both policy stances the coefficient sum for contractionary policy was a lesser magnitude (and statistically different) which may be evidence in favor of hypothesis 1. The high-capital bank columns in table 8 show evidence in favor of hypothesis 2 in all categories except small banks holding asset backed securities. Small banks holding asset backed securities were not affected by expansionary policy but did respond to contractionary policy which is at odds with hypothesis 2. For all other categories of high capital banks expansionary policy was more effective in increasing loan growth than contractionary policy was in reducing loan growth which is consistent with hypothesis 2. Robustness Checks Over time some banks in the sample switch between the high-capital and lowcapital categories. This is a possible source of bias as some variation in loan behavior between high and low-capital banks may change over time as a result of banks switching categories as opposed to in response to the federal funds rate. To address this issue a third model was estimated using data from 1980Q1-2010Q2. The model specification was identical to model 1 except for the high and low-capital categories. New categories were created by locking banks into a category based on their capital-to-assets ratio in 1995. The date of 1995 was chosen because it was 5 years after the Basel 1 capital accord which resulted in many banks changing increasing the levels of capital they held 32 on their balance sheets. The results were consistent with model 1 and the results of the regression are presented in the appendix. The results in table 8 are divided into securitizing and non-securitizing banks but the results show no evidence regarding how the degree of securitization affects banks’ responsiveness to monetary policy. To examine how the level of securitization on a bank’s balance sheet impacts in responsiveness to policy a fourth model was estimated. In this analysis a variable was created equal to banks’ securitized assets to total assets ratio. This variable was then interacted with all monetary policy variables. The sample is restricted to only securitized banks using data from 2001Q1-2010Q4. The results are presented in table 9. Table 9 Degree of Securitization Effects Bank/Policy Low-Capital Banks Large Banks -1.300** (Expansionary) Large Banks -2.009* (Contractionary) Small Banks -.4938 (Expansionary) Small Banks .9286 (Contractionary) Significant at 10%*, 5%**, 1%*** High-Capital Banks .9021 -.6025* -.6346 .9021 The values in table 7 are the sums of the interaction term coefficients. The only statistically significant findings are for large low-capital banks and large high-capital banks facing contractionary policy. All statistically significant coefficients are negative suggesting that the larger a bank’s securitized assets to total assets ratio the more 33 responsive the bank is to monetary policy. This result suggests that differences between securitizing and non-securitizing banks reported in table 6 could be due to unobserved traits that vary between the two bank categories. It is also possible that the model is showing variation resulting from the balance sheet channel. Aysun and Hepp (2011) found evidence that the degree of securitization has a positive effect on the balance sheet channel of monetary transmission. The responsiveness of loans reported in the coefficient sums could be a result of the balance sheet channel overpowering the bank lending channel. 34 6. Conclusion Recent research into the bank lending channel has emphasized capital constraints and bank size as factors in the efficacy of monetary policy. It has also been shown that monetary policy impacts the BLC asymmetrically not just by bank size and capital constraint but by the direction of the policy as well. Research has uncovered that contractionary policy is especially effective in reducing the loan growth of small lowcapital banks but expansionary policy is not effective in increasing the loan growth of these banks. It has also been shown that contractionary policy is ineffective in curtailing the loan growth of high-capital banks while expansionary policy is effective in increasing the loan growth of these banks. Economists have also uncovered evidence that securitization can shut down the bank lending channel, giving banks alternative methods to finance loans by pooling assets into securities and selling these securities to investors. This study empirically investigated these asymmetric trends in the bank lending channel and how the rise of securitization has impacted these relationships. Among lowcapital banks, contractionary policy was more effective than expansionary policy for all bank asset size categories. After collateralized debt obligations were developed in 1996, the analysis shows evidence that low-capital banks became less responsive to both contractionary and expansionary policy. The policy stance asymmetry among lowcapital banks held after the 1996 break for all but small banks which did not have a statistically significant result for expansionary policy. The empirical analysis produced results showing small and medium high-capital banks being able to bypass both policy stances before 1996. This result was inconsistent with expectations formed by surveying 35 existing literature; however, large high-capital banks did reveal expected behavior by being much more responsive to expansionary policy. High-Capital banks revealed evidence completely at odds with the third hypothesis. After 1996, large banks were more responsive to contractionary policy and both medium and small banks became more responsive to expansionary policy. In the second empirical model estimated, banks were divided into those that securitized their assets and those that did not. This analysis revealed some evidence in favor with hypothesis 4. Among large banks, those that securitized their assets were less responsive to monetary policy than those that did not. While large banks behaved as expected the behavior of smaller banks was much more conflicting. There was very little difference in the behavior between securitizing and non-securitizing small low-capital banks and small banks with high capital-to-assets ratios showed conflicting behavior. Securitizing banks in this category were less responsive to expansionary policy but more responsive to contractionary policy. The results of this study suggest the notion that securitization weakens the bank lending channel is more complex than previous research has suggested. The impact of securitization varies depending on banks’ capital strength and size which suggests that the structure of the banking system is an important factor in determining how the growth of asset-backed securities influences the bank lending channel. For example, this analysis suggests that a banking system heavily populated by large securitizing banks would be less responsive to monetary policy. The results also suggest that securitization 36 affects the difference in effectiveness of expansionary and contractionary monetary policy, a result with implications for business cycle management. Future research would do well to directly contrast how the large growth of assetbacked securities has impacted different channels of monetary transmission. While there is evidence suggesting the bank lending channel has become weaker, Aysun and Hepp (2011) found evidence that the balance sheet channel may have been made stronger by securitization. To truly understand how securitization has affected the efficacy of monetary policy the relative strength of different transmission mechanisms and how these mechanisms have been impacted needs to be compared and contrasted. 37 Appendix A: F-Statistics for Differences in Coefficient Sums Pre and Post 1996 (Table 4) Category Small High Capital Bank Small Low Capital Bank Medium High Capital Bank Medium Low Capital Bank Big High Capital Bank Big Low Capital Bank Contractionary Policy 4.21*** 0.71 3.43*** 2.71** 5.68*** 8.15*** Expansionary Policy 7.08*** 8.20** 9.31*** 5.3*** 0.21 50.14*** Contractionary and Expansionary Policy (Table 4) Category Small High Capital Bank Small Low Capital Bank Medium High Capital Bank Medium Low Capital Bank Big High Capital Bank Big Low Capital Bank Pre-1996 2.69** 12.31*** 1.01 15.27*** 8.44*** 5.35*** Post-1996 0.89 0.43 4.21*** 0.88 3.33*** 17.81*** Contractionary Policy 4.88** 0.47 1.00 0.73 5.73** 4.88** Expansionary Policy 1.29 5.87*** 0.56 23.73*** 8.00*** 53.23*** Pre and Post 2001 (Table 5) Category Small High Capital Bank Small Low Capital Bank Medium High Capital Bank Medium Low Capital Bank Big High Capital Bank Big Low Capital Bank 38 Contractionary and Expansionary Policy (Table 5) Category Small High Capital Bank Small Low Capital Bank Medium High Capital Bank Medium Low Capital Bank Big High Capital Bank Big Low Capital Bank Pre-2001 17.48*** 6.91*** 2.68*** 1.00 9.11*** 7.24** Post-2001 0.13 0.47 1.32 7.47** 0.87 2.99** ABS and Non-ABS Banks (Table 6) Category Contractionary Policy Small High Capital Bank Small Low Capital Bank Big High Capital Bank Big Low Capital Bank 0.56 0.91 0.72 15.26*** Expansionary Policy 10.03*** 0.09 1.35 21.21*** Contractionary and Expansionary Policy (Table 6) Category Contractionary Policy Small High Capital Bank Small Low Capital Bank Big High Capital Bank Big Low Capital Bank 5.02** 4.63** 4.01** 8.88** Expansionary Policy 16.52*** 13.22*** 5.12** 4.24*** 39 Appendix B: Model 1 Using Fixed Capital-to-Asset Ratios Policy Pre-1996 Low Stance/Bank Capital Banks Size Large Banks -2.87*** (Expansionary) Large Banks -3.55* (Contractionary) Medium Banks -.547*** (Expansionary) Medium Banks -4.12** (Contractionary) Small Banks -1.20*** (Expansionary) Small Banks -3.87** (Contractionary) Significant at 10%*, 5%**, 1%*** Pre-1996 High Capital Banks Post-1996 Low Capital Banks -3.11*** 7.25** Post-1996 High Capital Banks 4.22 8.522* -1.49*** -7.47** 3.00*** -.689** -3.13** 3.99*** -.742*** 3.20** 3.58*** 1.81 -.232* 2.101** -2.53* -.041 40 References Altumbas, Y., Fazylov, O., & Molyneux, P. (2002). Evidence on the bank lending channel in Europe. Journal of Banking and Finance, 26, 2093-2110. Altumbas, Y., Gambacorta, L., & Marques-Ibanez, D. (2009). Securitization and the bank lending channel. European Economic Review, 53, 996-1009. Ashcraft, A. (2006). New evidence on the lending channel. Journal of Money, Credit and Banking, 38, 751-776. Aysun, U., & Hepp, R. (2011). Securitization and the balance sheet channel of monetary transmission. Journal of Banking and Finance, 35, 2111-2122. Berger, A., Herring, R., & Szego, G. (1995). The role of capital in financial institutions. Journal of Banking and Finance, 19, 393-430. Bernanke, B., & Blinder, A. (1992). The federal funds rate and the channels of monetary transmission. American Economic Review, 82, 901-922. Bhaumik, S., Dang, V., & Kutan, A. (2011). Implications of bank ownership for the credit channel of monetary policy transmission. Economics and Finance Research, 29, 2010-2998. Bliss, R., & Kaufman, G. (2003). Bank procyclicality, credit crunches, and asymmetric effects of monetary policy. Journal of Applied Finance , 13, 23-31. Cover, J. (1992). Asymmetric effects of positive and negative money supply shocks. The Quarterly Journal of Economics, 107(4), 1261-1282. 41 Estrella, A. (2002). Securitization and the efficacy of monetary policy. Economic Policy Review, 8(1), 1-12. Judson, R., & Klee, E. (2010). Big Bank, Small Bank: Monetary Policy Implementation nd Banks' Reserve Management Strategies. Journal of Economics and Business, 63, 306328. Stein, J., & Kashyap, A. (1994). Monetary Policy and Bank Lending. Monetary Policy (pp. 221-261). Chicago: The University of Chicago Press. Kashyap, A., & Stein, J. (2000). What do a million banks have to say about the transmission of monetary policy. American Economic Review, 90, 407-428. Kishan, R., & Opiela, T. (2000). Bank size, bank capital, and the bank lending channel. Journal of Money, Credit and Banking, 32, 121-141. Kishan, R., & Opiela, T. (2006). Bank capital and loan asymmetry in the transmission of monetary policy. Journal of Banking and Finance, 30, 259-285. Loutskina, E., & Strahan, P. (2009). Securitization and the declining impact of bank finance on loan supply: evidence from mortgage organizations. Journal of Finance, 64, 861-889. Matousek, R., & Sarantis, N. (2009). The bank lending channel and monetary transmission in central and eastern european countries. Journal of Comparative Economics, 37, 321-334. Miskin, F. (2007). The economics of money, banking, and financial markets. Boston: Pearson Addison Wesley 42 Opiela, T. (2008). Differential deposit guarantees and the effect of monetary policy on bank lending. Economic Inquery, 46, 610-623. Senda, T. (1995). Asymmetric effects of money supply shocks and trend inflation. Journal of Money, Credit and Banking, 33, 65-89. den Heuvel, S. V. (2002). Does bank capital matter for monetary policy transmission?. Economic Policy Review, Federal Reserve Bank of New York, 259-265. Weise, C. (1999). The asymmetric effects of monetary policy: A nonlinear vector autoregression approach. Journal of Money, Credit and Banking, 31, 85-108. 43