The Transmission of Bank Liquidity Shocks: Evidence from House Prices

The Transmission of Bank Liquidity Shocks: Evidence from House Prices Abstract This article uses the 2007–09 financial crisis as a negative liquidity shock on banks in the USA and analyzes its transmission to the real economy. The ex ante heterogeneity in the amount of long-term debt that matured during the crisis is used to measure the variation in banks’ exposure to the liquidity shock. I find that banks transmitted the liquidity shock to the real economy by reducing their loan supply. The reduction was particularly strong for real estate loans. As a result, house prices declined in the MSAs where these banks have branches. Bank capital plays a significant role in the transmission: under-capitalized banks transmitted the liquidity shock, whereas well-capitalized banks’ lending did not show any decline. 1. Introduction Do banks transmit liquidity shocks across markets by a reduction in their loan supply? If so, do these shocks have an impact on the real economy? The goal of this article is to answer these questions by using the 2007–09 financial crisis as a negative liquidity shock to banks in the USA. The crisis led to a systemic shock that affected many banks at the same time. To study the transmission of this shock to the real economy, I need to identify a variation in the exposure of similar banks to the liquidity shock. I follow Almeida et al. (2011) and use the heterogeneity in the amount of long-term debt that matured right after the onset of the crisis to measure the variation in banks’ exposure to the shock.1 During the crisis, financing costs for long-term debt increased sharply. For example, Citigroup’s investment-grade spreads increased from 1% in September 2007 to 3% in early 2008 (Almeida et al., 2011). The collapse of Lehman Brothers aggravated this effect further to around 7% at the end of 2008. Similarly, high-yield spreads, which had been around 3% in early 2007, approached 8% in early 2008 and reached a level of around 17% shortly after September 2008. As a result, banks with a larger amount of long-term debt that matured during the crisis are expected to be affected more heavily since these banks had to refinance their maturing debt when financing costs were high. This makes the amount of debt that matured during the crisis a good proxy to measure the individual bank’s exposure to the negative liquidity shock. It is important to use the long-term debt issued before the end of 2006 rather than using short-term debt.2 When the crisis hit, the amount of long-term debt due around that time was an exogenous shock to banks since they did not anticipate the coming crisis when they issued this debt before the end of 2006. This does not hold for their short-term debt which was issued when the crisis was anticipated. This generates an endogeneity problem since the amount of short-term debt depends on the anticipation of the crisis. Using long-term debt solves this endogeneity problem. In this way, I can identify a bank liquidity shock that is exogenous to the state of the economy. As a result, I can study the causal effect of this liquidity shock on banks’ loan supply and examine the transmission of this shock to the real economy. One might argue that the amount of long-term debt, on average 10% of total assets, is insignificant compared with other liabilities in bank balance sheets and, as a result, the effect of the long-term debt due during the crisis is expected to be insignificant. To understand the importance of the long-term debt that matured during the last financial crisis, one can look at the largest bank in my sample with total assets of $1.88 trillion at the end of 2006. This bank had $20.87 billion of long-term debt maturing during the crisis. This is a substantial amount of debt that needed to be rolled over for one bank. So the amount of long-term debt that matured during the crisis was significantly important, although the fraction of long-term debt is quite small. To study the effect of the liquidity shock on banks’ loan supply, I need to separate supply from demand. To do so, I use a difference in differences (DID) methodology where I match each treated bank that is affected by the liquidity shock with a set of unaffected control banks that are located in the same metropolitan statistical areas (MSAs). This matched bank approach that is proposed by Carlson, Shan, and Warusawitharana (2013) enables me to control for local environment for banks because banks in the same location face the same economic environment. The remaining difference in the amount of loans should then be related to differences in banks’ exposure to the liquidity shock during the crisis. To provide evidence on the transmission of the negative liquidity shock to the real economy, I proceed in the following steps: first, I show that an increase in the amount of long-term debt that matured during the crisis leads to a significant decrease in banks’ outstanding long-term debt. According to my results, a bank with 1 percentage point higher long-term debt ratio that matured during the crisis decreased its long-term debt by almost 0.04% of its total assets. This result suggests that banks had a hard time rolling over their maturing long-term debt and had to cut their long-term debt holdings. This could be attributed to the sharp increase in financing costs during the crisis. Second, I document the transmission of this shock from the liability side to the asset side of a bank’s balance sheet through a reduction in its loan supply. According to my results, a 0.01 higher fraction of long-term debt maturing during the crisis led to a significant reduction in a bank’s loan supply by almost 0.09% of its total assets. This result holds particularly strong for banks with lower deposit ratios and banks with higher short-term debt holdings. This supports the earlier findings that banks with higher deposit ratios are more robust to liquidity shocks because deposits are a more stable source of financing (see, e.g., Gatev and Strahan, 2006; Ivashina and Scharfstein, 2010; Cornett et al., 2011; Dagher and Kazimov, 2015), and that banks with higher short-term debt ratios were affected more heavily by the last financial crisis since these banks faced higher roll-over risk (see, e.g., Cornett et al., 2011; Huang and Ratnovski, 2011; Dagher and Kazimov, 2015). To understand the motivation behind the reduction in the loan supply, I study the effect on nonperforming loans, loan loss reserves, and cash holdings of banks, so that I can answer the question whether banks decreased their loan supply to reduce their portfolio risk or to preserve liquidity. I find that 10 percentage point increase in the fraction of long-term debt that matured during the crisis led to a decrease in a bank’s nonperforming loans by 0.08% of its total assets and a decrease in its loan loss reserves by 0.03% of its total assets, whereas there was no significant change in the cash holdings. This suggests that banks that were affected by the liquidity shock decreased their lending to reduce their portfolio risk. I next analyze the effect of the liquidity shock on three types of loans: real estate loans, consumer loans, and commercial and industrial (C&I) loans. The results indicate that the observed decrease in total loans was a result of a significant decrease in real estate loans and consumer loans, whereas C&I loans were not affected. One percentage point higher fraction of long-term debt that matured during the crisis led to a decrease in a bank’s real estate loans by 0.06% of its total assets and a decrease in its consumer loans by almost 0.02% of its total assets. This suggests that when banks cut their outstanding long-term debt due to the high financing costs, they adjusted their assets by a reduction in the real estate loans and consumer loans. Thus, the negative liquidity shock triggered by the last financial crisis is transmitted from the liability side to the asset side of a bank’s balance sheet. Third, I explore the transmission of this shock to the real economy through the reduction in the real estate loans by studying the effect on house prices. For each MSA, a weighted average fraction of long-term debt that matured during the crisis is calculated. The fraction of deposits that each bank has in an MSA is used as weights. This measures each MSA’s exposure to the liquidity shock through banks that have branches in this MSA.3 The results show that house prices declined more in the MSAs where strongly affected banks have more deposits. 1 percentage point increase in the weighted average fraction of long-term debt that matured during the crisis resulted in a 0.13 percentage point decrease in the growth rate of house prices. The results suggest that banks with a larger amount of long-term debt that matured during the crisis could not find affordable financing sources to cover their shortfalls. The effect of high financing costs is expected to be stronger for under-capitalized banks since these banks are perceived as riskier compared to well-capitalized banks. Financial frictions prevent these banks from finding alternative financing sources to continue financing their loans during contraction times (Kashyap and Stein, 2000; Kishan and Opiela, 2000; Meh and Moran, 2010). To study whether the transmission of the negative liquidity shock depends on banks’ capital ratios, I split the sample into under-capitalized and well-capitalized banks. The results show that only under-capitalized banks are affected by the negative liquidity shock. An under-capitalized bank with 1 percentage point increase in its long-term debt ratio that matured during the crisis decreased its total loans by 0.18% of its total assets and its real estate loans by almost 0.08% of its total assets. On the contrary, well-capitalized banks do not show any significant reduction in their loan supply. The results are consistent at the MSA level: for MSAs with more well-capitalized banks, 1 percentage point increase in the weighted average fraction of long-term debt that matured during the crisis leads to a 0.12 percentage point decrease in the growth rate of house prices, whereas the decrease is 0.18 percentage point in the growth rate of house prices for MSAs with more under-capitalized banks. This suggests that house prices in the MSAs with more well-capitalized banks are affected less relative to MSAs with more under-capitalized banks although the difference is not significant. This article contributes to several strands of literature. First, it relates to a large literature studying the real consequences of bank liquidity shocks (see, e.g., Gan, 2007; Khwaja and Mian, 2008; Paravisini, 2008; Chava and Purnanandam, 2011; Schnabl, 2012). Gan (2007) studies how a negative shock to the financial health of banks, caused by a land market collapse in Japan, affects the real economy through firms’ investment and market valuation. Khwaja and Mian (2008) study the effect of bank liquidity shocks on borrowing firms induced by the unexpected nuclear tests of Pakistan in 1998. According to their results, small firms that borrow from affected banks are significantly more likely to be in financial distress a year after the nuclear tests. Paravisini (2008) uses an exogenous allocation of government funds across banks in Argentina to study the effect on banks’ loan supply. According to his results, financial shocks to constrained banks have an amplified effect on the aggregate supply of credit. Chava and Purnanandam (2011) use the 1998 Russian default as an exogenous shock to the US banking system and provide evidence that adverse capital shocks to banks affect their borrowers’ performance negatively. Similarly, Schnabl (2012) exploits the 1998 Russian default as an exogenous shock to international banks and analyzes the transmission of this shock to Peru. Overall, this strand of literature examines whether bank liquidity shocks are transmitted to the real economy through a change in banks’ loan supply. This article contributes to this literature by first studying the effect of a negative liquidity shock on banks’ loan supply and then linking this change in the loan supply to the real economy with a focus on house prices. This article is also related to research estimating the real costs of financial crises (see, e.g., Peek and Rosengren, 2000; Ongena, Smith, and Michalsen, 2003; Dell’Ariccia, Detragiache, and Rajan, 2008; Ivashina and Scharfstein, 2010; Iyer and Peydro, 2011; Puri, Rocholl, and Steffen, 2011; Iyer et al., 2014; Cingano, Manaresi, and Sette, 2016). Peek and Rosengren (2000) use the Japanese banking crisis as an exogenous loan supply shock in the USA and show that loan supply shocks originated in Japan had real effects on the economic activity in the USA. Ongena, Smith, and Michalsen (2003) use the Norwegian banking crisis to study the impact of bank distress announcements on the stock performance of borrowers. To explore the effect of banking crises on the real activity, Dell’Ariccia, Detragiache, and Rajan (2008) study differential effects of banking crises across different sectors. By exploiting the sudden failure of a large bank in India, Iyer and Peydro (2011) inspect the effect of financial contagion through interbank linkages. Iyer et al. (2014) study the effect of the unexpected shock to the interbank markets in August 2007 on the credit supply in Portugal. Similarly, Cingano, Manaresi, and Sette (2016) study the effect of the liquidity drought in interbank markets on the credit supply and real economic activity in Italy. My article particularly adds to the literature that investigates the costs of the last financial crisis. Almeida et al. (2011) analyze the effect of financial contracting on firms’ investment decisions during the last financial crisis. They find that firms whose long-term debt largely matured after the third quarter of 2007 cut their investment-to-capital ratio significantly more compared to otherwise similar firms. Puri, Rocholl, and Steffen (2011) and Ivashina and Scharfstein (2010) examine the effects of the US financial crisis on the retail bank lending in Germany and on new loans in the USA, respectively. Puri, Rocholl, and Steffen (2011) find that the US financial crisis induced a contraction in the supply of retail lending in Germany. Ivashina and Scharfstein (2010) show that new loans issued to large borrowers fell significantly during the last financial crisis. This article contributes to this literature by using the last financial crisis as a negative liquidity shock on the banking system and examining the transmission of this shock to the real economy through a reduction in banks’ loan supply. Finally, my findings on house prices are related to the literature that studies the effect of bank loan supply on house prices. Mian and Sufi (2009) show that the increase in securitization in the early 2000 s led to an expansion in subprime mortgage credit. This expansion had contributed significantly to the increasing house prices from 2002 to 2005. Adelino, Schoar, and Severino (2012) find that easier access to credit, made possible by changes in the conforming loan limit, significantly increased house prices. Di Maggio and Kermani (2017) use the federal preemption of national banks from local laws against predatory lending in 2004 to show that the supply of credit has significant consequences on house prices. Favara and Imbs (2015) show the significant effects of credit expansion on house prices by exploiting the US branching deregulations during 1994 and 2005 as an exogenous expansion in mortgage credit. This article adds to this literature by examining the effect of negative liquidity shocks on house prices through a reduction in the real estate loans during the last financial crisis. The rest of the article is organized as follows. Section 2 introduces the data. Section 3 presents the empirical analysis and main results. In Section 4, I introduce robustness tests and additional controls. Section 5 concludes. 2. Data In the first part of my empirical analysis, I use data on the bank balance sheet, income statement, bank capital structure, and branch deposits for US bank holding companies (BHC). The data are collected from SNL Financial. The balance sheet and income statement on BHCs are taken from the Consolidated Financial Statements of Bank Holding Companies (FR Y-9 C) for the time period from 2006 to 2009.4 The bank branch data come from the Summary of Deposits filings of the Federal Deposit Insurance Corporation (FDIC). In the second part of my empirical analysis, I use data on population, income per capita, house price index (HPI), median age in the population, owner occupied housing, population with different levels of education, and unemployment rate at the MSA level. The data on population and income per capita are collected by the Bureau of Economic Analysis. The HPI data come from the Federal Housing Finance Agency (FHFA). The data on median age, owner occupied housing, levels of education, and unemployment rate are taken from the U.S. Census Bureau and collected through SNL Financial. To eliminate outliers, I exclude bank-years with negative values of total equity ratios. I focus on deposit taking institutions that have positive deposit holdings and positive loans outstanding. The final unbalanced sample consists of 913 bank-year observations and 242 banks. To measure a bank’s financial condition, several bank characteristics that proxy for the components of the CAMELS rating are used. The CAMELS rating is a supervisory rating used by federal banking regulators to provide a convenient summary of bank conditions. The acronym CAMELS refers to the six components of a bank’s condition: capital adequacy, asset quality, management, earnings, liquidity, and sensitivity to market risk. Banking regulators use this rating to provide a comprehensive assessment of a bank’s overall condition (Lopez, 1999). In this article, I use total equity to proxy capital adequacy, loan loss reserves for asset quality, net interest income for management, return on assets (ROA) for earnings, cash for liquidity, and total deposits for the sensitivity to the market risk. These six bank characteristics are included as controls in each regression to control for differences. In addition, I control for the size of the banks and their long-term debt ratios. Table I reports summary statistics for the variables used in the analysis. The average bank in my sample has $27.3 billion in total assets, and the standard variation is $178.2 billion. The smallest bank is Community Capital Bancshares with $0.19 billion in total assets, whereas the largest bank in the sample is Citigroup with total assets of $2.19 trillion. In my sample, the largest asset class is total loans which constitutes on average 71.7% of total assets. On average, 53% of the total loans are real estate loans, followed by C&I loans that constitute almost 21%. The third largest type of loans is the consumer loans that represent 13% of the total loans. On the liability side, a bank holds total equity that is on average 8.9% of its total assets. Total deposits are the largest liability class which is 74.5% of total assets. The total debt other than deposits (non-depository debt) is 14.7% of the total assets. 68.1% of this debt is long-term debt with at least 1 year maturity. The remaining 31.9% is short-term debt with a maturity shorter than 1 year. Table I. Definitions and summary statistics for variables Variable Definition Mean Standard deviation Panel A: Bank characteristics Total equity Total equity divided by total assets to proxy Capital adequacy (C) 0.089 0.023 Loan loss reserves Loan loss reserves divided by total loans to proxy Asset quality (A) 0.011 0.005 Net interest income Net interest income divided by total assets to proxy Management quality (M) 0.032 0.007 ROA Net income divided by total assets to proxy Earnings (E) 0.003 0.015 Cash Cash and noninterest-bearing deposits divided by total assets to proxy Liquidity (L) 0.055 0.052 Deposits Total deposits divided by total assets to proxy Sensitivity to market risk (S) 0.745 0.092 Total assets (billions) Total assets 27.338 178.200 Total loans Total loans divided by total assets 0.717 0.132 Real-estate loans Real-estate loans divided by total loans 0.530 0.147 C&I loans Commercial and industrial loans divided by total loans 0.209 0.122 Consumer loans Consumer loans divided by total loans 0.130 0.096 Total debt Total debt other than deposits 0.147 0.080 Short-term debt Total debt with a maturity of 1 year or less divided by total debt 0.319 0.234 Long-term debt Total debt minus short-term debt divided by total debt 0.681 0.234 Debt Total long-term debt issued from the beginning of 2000 until the end of 2006 divided by total assets 0.036 0.021 Maturing debt The fraction of long-term debt issued from the beginning of 2000 until the end of 2006 that matured during 2008–09 0.348 0.336 Bank-year observations 913 Panel B: MSA characteristics HPI HPI at the MSA level 185.834 36.779 Population (million) Population at the MSA level 0.476 0.694 Income per capita (thousand) Income divided by total population 35.075 6.502 Median age Population median age at the MSA level 36.333 3.969 Owner occupied houses Owner occupied housing units divided by total housing units 62.187 6.197 Population with a college degree Population aged 25+ with at least a college degree divided by total population aged 25+ 25.173 8.120 Unemployment rate Unemployment rate at the MSA level 6.116 2.214 MSA-year observations 1480 Variable Definition Mean Standard deviation Panel A: Bank characteristics Total equity Total equity divided by total assets to proxy Capital adequacy (C) 0.089 0.023 Loan loss reserves Loan loss reserves divided by total loans to proxy Asset quality (A) 0.011 0.005 Net interest income Net interest income divided by total assets to proxy Management quality (M) 0.032 0.007 ROA Net income divided by total assets to proxy Earnings (E) 0.003 0.015 Cash Cash and noninterest-bearing deposits divided by total assets to proxy Liquidity (L) 0.055 0.052 Deposits Total deposits divided by total assets to proxy Sensitivity to market risk (S) 0.745 0.092 Total assets (billions) Total assets 27.338 178.200 Total loans Total loans divided by total assets 0.717 0.132 Real-estate loans Real-estate loans divided by total loans 0.530 0.147 C&I loans Commercial and industrial loans divided by total loans 0.209 0.122 Consumer loans Consumer loans divided by total loans 0.130 0.096 Total debt Total debt other than deposits 0.147 0.080 Short-term debt Total debt with a maturity of 1 year or less divided by total debt 0.319 0.234 Long-term debt Total debt minus short-term debt divided by total debt 0.681 0.234 Debt Total long-term debt issued from the beginning of 2000 until the end of 2006 divided by total assets 0.036 0.021 Maturing debt The fraction of long-term debt issued from the beginning of 2000 until the end of 2006 that matured during 2008–09 0.348 0.336 Bank-year observations 913 Panel B: MSA characteristics HPI HPI at the MSA level 185.834 36.779 Population (million) Population at the MSA level 0.476 0.694 Income per capita (thousand) Income divided by total population 35.075 6.502 Median age Population median age at the MSA level 36.333 3.969 Owner occupied houses Owner occupied housing units divided by total housing units 62.187 6.197 Population with a college degree Population aged 25+ with at least a college degree divided by total population aged 25+ 25.173 8.120 Unemployment rate Unemployment rate at the MSA level 6.116 2.214 MSA-year observations 1480 Table I. Definitions and summary statistics for variables Variable Definition Mean Standard deviation Panel A: Bank characteristics Total equity Total equity divided by total assets to proxy Capital adequacy (C) 0.089 0.023 Loan loss reserves Loan loss reserves divided by total loans to proxy Asset quality (A) 0.011 0.005 Net interest income Net interest income divided by total assets to proxy Management quality (M) 0.032 0.007 ROA Net income divided by total assets to proxy Earnings (E) 0.003 0.015 Cash Cash and noninterest-bearing deposits divided by total assets to proxy Liquidity (L) 0.055 0.052 Deposits Total deposits divided by total assets to proxy Sensitivity to market risk (S) 0.745 0.092 Total assets (billions) Total assets 27.338 178.200 Total loans Total loans divided by total assets 0.717 0.132 Real-estate loans Real-estate loans divided by total loans 0.530 0.147 C&I loans Commercial and industrial loans divided by total loans 0.209 0.122 Consumer loans Consumer loans divided by total loans 0.130 0.096 Total debt Total debt other than deposits 0.147 0.080 Short-term debt Total debt with a maturity of 1 year or less divided by total debt 0.319 0.234 Long-term debt Total debt minus short-term debt divided by total debt 0.681 0.234 Debt Total long-term debt issued from the beginning of 2000 until the end of 2006 divided by total assets 0.036 0.021 Maturing debt The fraction of long-term debt issued from the beginning of 2000 until the end of 2006 that matured during 2008–09 0.348 0.336 Bank-year observations 913 Panel B: MSA characteristics HPI HPI at the MSA level 185.834 36.779 Population (million) Population at the MSA level 0.476 0.694 Income per capita (thousand) Income divided by total population 35.075 6.502 Median age Population median age at the MSA level 36.333 3.969 Owner occupied houses Owner occupied housing units divided by total housing units 62.187 6.197 Population with a college degree Population aged 25+ with at least a college degree divided by total population aged 25+ 25.173 8.120 Unemployment rate Unemployment rate at the MSA level 6.116 2.214 MSA-year observations 1480 Variable Definition Mean Standard deviation Panel A: Bank characteristics Total equity Total equity divided by total assets to proxy Capital adequacy (C) 0.089 0.023 Loan loss reserves Loan loss reserves divided by total loans to proxy Asset quality (A) 0.011 0.005 Net interest income Net interest income divided by total assets to proxy Management quality (M) 0.032 0.007 ROA Net income divided by total assets to proxy Earnings (E) 0.003 0.015 Cash Cash and noninterest-bearing deposits divided by total assets to proxy Liquidity (L) 0.055 0.052 Deposits Total deposits divided by total assets to proxy Sensitivity to market risk (S) 0.745 0.092 Total assets (billions) Total assets 27.338 178.200 Total loans Total loans divided by total assets 0.717 0.132 Real-estate loans Real-estate loans divided by total loans 0.530 0.147 C&I loans Commercial and industrial loans divided by total loans 0.209 0.122 Consumer loans Consumer loans divided by total loans 0.130 0.096 Total debt Total debt other than deposits 0.147 0.080 Short-term debt Total debt with a maturity of 1 year or less divided by total debt 0.319 0.234 Long-term debt Total debt minus short-term debt divided by total debt 0.681 0.234 Debt Total long-term debt issued from the beginning of 2000 until the end of 2006 divided by total assets 0.036 0.021 Maturing debt The fraction of long-term debt issued from the beginning of 2000 until the end of 2006 that matured during 2008–09 0.348 0.336 Bank-year observations 913 Panel B: MSA characteristics HPI HPI at the MSA level 185.834 36.779 Population (million) Population at the MSA level 0.476 0.694 Income per capita (thousand) Income divided by total population 35.075 6.502 Median age Population median age at the MSA level 36.333 3.969 Owner occupied houses Owner occupied housing units divided by total housing units 62.187 6.197 Population with a college degree Population aged 25+ with at least a college degree divided by total population aged 25+ 25.173 8.120 Unemployment rate Unemployment rate at the MSA level 6.116 2.214 MSA-year observations 1480 To examine whether any specific pre-crisis bank characteristics are correlated with the amount of maturing long-term debt during the crisis, I examine the effect of an increase in the maturing long-term debt on pre-crisis bank characteristics that proxy CAMELS rating. According to the results presented in Table II, the banks that had more maturing long-term debt during the crisis held significantly higher net interest income. More precisely, a bank with a 1 percentage point higher fraction of long-term debt that matured during the crisis had significantly more net interest income by 0.13 of a percentage point. Table II. The relationship between maturing long-term debt and pre-crisis bank characteristics The regressions in this table investigate the effect of an increase in the long-term debt maturing during the crisis on pre-crisis bank characteristics. Dependent variables are the pre-crisis bank characteristics, measured in 2006 and 2007, that proxy CAMELS supervisory rating. The variable “Maturing debt” is the fraction of long-term debt that matured during the crisis. All other variables are defined in Table I. All columns include controls for year fixed effects and 1 year lagged bank characteristics. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Dependent variable Total equity Loan loss reserves Net interest income ROA Cash Deposits Maturing debt 0.070 –0.002 0.131** 0.132 1.997 1.682 (0.170) (0.036) (0.065) (0.139) (1.517) (1.241) Total equity 0.927*** 0.000 0.022** –0.047 0.105 0.089 (0.024) (0.004) (0.009) (0.042) (0.073) (0.092) Loan loss reserves 0.001 0.903*** 0.036 –0.114 –2.524** 0.213 (0.202) (0.037) (0.063) (0.081) (1.036) (0.725) Net interest income 0.055 0.031* 0.930*** 0.026 0.145 –0.346 (0.098) (0.017) (0.035) (0.066) (0.917) (0.409) ROA –0.102 –0.018 –0.081* 0.793*** –0.893 –0.581 (0.139) (0.026) (0.042) (0.118) (0.788) (0.577) Cash –0.008 –0.004*** 0.001 0.004 0.866*** 0.076 (0.009) (0.002) (0.003) (0.005) (0.076) (0.047) Deposits 0.003 –0.002 0.002 0.006 0.008 0.933*** (0.006) (0.001) (0.002) (0.004) (0.042) (0.039) Size –0.004 0.005 –0.003 0.017 0.361** –0.265 (0.034) (0.007) (0.011) (0.019) (0.145) (0.161) Debt –0.048* 0.010* –0.001 –0.020 –0.007 –0.309* (0.028) (0.006) (0.010) (0.017) (0.120) (0.184) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.873 0.779 0.862 0.398 0.672 0.869 Observations 394 394 394 394 394 394 Dependent variable Total equity Loan loss reserves Net interest income ROA Cash Deposits Maturing debt 0.070 –0.002 0.131** 0.132 1.997 1.682 (0.170) (0.036) (0.065) (0.139) (1.517) (1.241) Total equity 0.927*** 0.000 0.022** –0.047 0.105 0.089 (0.024) (0.004) (0.009) (0.042) (0.073) (0.092) Loan loss reserves 0.001 0.903*** 0.036 –0.114 –2.524** 0.213 (0.202) (0.037) (0.063) (0.081) (1.036) (0.725) Net interest income 0.055 0.031* 0.930*** 0.026 0.145 –0.346 (0.098) (0.017) (0.035) (0.066) (0.917) (0.409) ROA –0.102 –0.018 –0.081* 0.793*** –0.893 –0.581 (0.139) (0.026) (0.042) (0.118) (0.788) (0.577) Cash –0.008 –0.004*** 0.001 0.004 0.866*** 0.076 (0.009) (0.002) (0.003) (0.005) (0.076) (0.047) Deposits 0.003 –0.002 0.002 0.006 0.008 0.933*** (0.006) (0.001) (0.002) (0.004) (0.042) (0.039) Size –0.004 0.005 –0.003 0.017 0.361** –0.265 (0.034) (0.007) (0.011) (0.019) (0.145) (0.161) Debt –0.048* 0.010* –0.001 –0.020 –0.007 –0.309* (0.028) (0.006) (0.010) (0.017) (0.120) (0.184) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.873 0.779 0.862 0.398 0.672 0.869 Observations 394 394 394 394 394 394 Table II. The relationship between maturing long-term debt and pre-crisis bank characteristics The regressions in this table investigate the effect of an increase in the long-term debt maturing during the crisis on pre-crisis bank characteristics. Dependent variables are the pre-crisis bank characteristics, measured in 2006 and 2007, that proxy CAMELS supervisory rating. The variable “Maturing debt” is the fraction of long-term debt that matured during the crisis. All other variables are defined in Table I. All columns include controls for year fixed effects and 1 year lagged bank characteristics. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Dependent variable Total equity Loan loss reserves Net interest income ROA Cash Deposits Maturing debt 0.070 –0.002 0.131** 0.132 1.997 1.682 (0.170) (0.036) (0.065) (0.139) (1.517) (1.241) Total equity 0.927*** 0.000 0.022** –0.047 0.105 0.089 (0.024) (0.004) (0.009) (0.042) (0.073) (0.092) Loan loss reserves 0.001 0.903*** 0.036 –0.114 –2.524** 0.213 (0.202) (0.037) (0.063) (0.081) (1.036) (0.725) Net interest income 0.055 0.031* 0.930*** 0.026 0.145 –0.346 (0.098) (0.017) (0.035) (0.066) (0.917) (0.409) ROA –0.102 –0.018 –0.081* 0.793*** –0.893 –0.581 (0.139) (0.026) (0.042) (0.118) (0.788) (0.577) Cash –0.008 –0.004*** 0.001 0.004 0.866*** 0.076 (0.009) (0.002) (0.003) (0.005) (0.076) (0.047) Deposits 0.003 –0.002 0.002 0.006 0.008 0.933*** (0.006) (0.001) (0.002) (0.004) (0.042) (0.039) Size –0.004 0.005 –0.003 0.017 0.361** –0.265 (0.034) (0.007) (0.011) (0.019) (0.145) (0.161) Debt –0.048* 0.010* –0.001 –0.020 –0.007 –0.309* (0.028) (0.006) (0.010) (0.017) (0.120) (0.184) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.873 0.779 0.862 0.398 0.672 0.869 Observations 394 394 394 394 394 394 Dependent variable Total equity Loan loss reserves Net interest income ROA Cash Deposits Maturing debt 0.070 –0.002 0.131** 0.132 1.997 1.682 (0.170) (0.036) (0.065) (0.139) (1.517) (1.241) Total equity 0.927*** 0.000 0.022** –0.047 0.105 0.089 (0.024) (0.004) (0.009) (0.042) (0.073) (0.092) Loan loss reserves 0.001 0.903*** 0.036 –0.114 –2.524** 0.213 (0.202) (0.037) (0.063) (0.081) (1.036) (0.725) Net interest income 0.055 0.031* 0.930*** 0.026 0.145 –0.346 (0.098) (0.017) (0.035) (0.066) (0.917) (0.409) ROA –0.102 –0.018 –0.081* 0.793*** –0.893 –0.581 (0.139) (0.026) (0.042) (0.118) (0.788) (0.577) Cash –0.008 –0.004*** 0.001 0.004 0.866*** 0.076 (0.009) (0.002) (0.003) (0.005) (0.076) (0.047) Deposits 0.003 –0.002 0.002 0.006 0.008 0.933*** (0.006) (0.001) (0.002) (0.004) (0.042) (0.039) Size –0.004 0.005 –0.003 0.017 0.361** –0.265 (0.034) (0.007) (0.011) (0.019) (0.145) (0.161) Debt –0.048* 0.010* –0.001 –0.020 –0.007 –0.309* (0.028) (0.006) (0.010) (0.017) (0.120) (0.184) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.873 0.779 0.862 0.398 0.672 0.869 Observations 394 394 394 394 394 394 Summary statistics for the variables at the MSA level are reported in Panel B in Table I. The mean HPI is 185.834 with Naples–Immokalee–Marco Island (FL) having the maximum HPI at 363.33 and Merced (CA) having the minimum HPI at 117.68. In my sample, the average MSA has 0.476 million people. The highest population belongs to New York–Newark–Jersey City (NY–NJ–PA) with 19.60 million people. The lowest belongs to Carson City (NV) with 55.29 thousand people. The average income per capita is $35.08 thousand with a minimum value of $18.73 thousand in McAllen–Edinburg–Mission (TX) and a maximum value of $88.81 in Bridgeport–Stamford–Norwalk (CT). The average median age in my sample is 36.33. Provo–Orem (UT) has the lowest median age with 23.4, whereas Punta Gorda (FL) has the highest median age with 54.4. The mean ratio of owner occupied housing units is 62.19% where the minimum ratio, 34.37%, belongs to Ocean City (NJ) and the maximum ratio, 77.09%, belongs to Monroe (MI). On average, people with at least a college degree at an associate level5 constitute 25.17% of total population. Dalton (GA) has the lowest fraction of people with a college degree at 14.64% and Boulder (CO) has the highest fraction at 58.70%. The mean unemployment rate is 6.116% where the highest unemployment rate, 13.6%, is in Hanford–Corcoran (CA) and the lowest unemployment rate, 2.6%, is in Sheboygan (WI). 2.1 Long-term debt due during the financial crisis of 2007–09 The most important variable of my analysis is the amount of long-term debt that matured during the 2007–09 financial crisis. The years 2008 and 2009 are used as the crisis following the National Bureau of Economic Research (NBER). The variation in banks’ exposure to the liquidity shock is measured by the ex ante heterogeneity in the fraction of their long-term debt that matured during the crisis. As mentioned above, the main advantage of using the fraction of long-term debt due during the crisis to identify an exogenous variation in banks’ exposure to the negative liquidity shock is that it mitigates endogeneity concerns. However, it also requires assumptions that can lead to a potential measurement error in this exposure. I collect the information on the issuance and the due date of the long-term debt for each bank from SNL Financial capital structure. I define the amount of long-term debt that matured during the crisis as the debt issued before the end of 2006 and due between the end of 2007 and 2009. There is one limitation with these data: the database SNL Financial has the data on capital structure from the end of 1999 onwards. By using these data, I neglect the long-term debt issued before the end of 1999 and due during the crisis. To understand how large this part of the debt is, I calculate the amount of long-term debt that was outstanding at the end of 2006 which was issued from the end of 1999 until the end of 2006. When I compare this amount with the total amount of long-term debt outstanding at the end of 2006, I find that approximately 50% of the total long-term debt was issued from the end of 1999 onwards given that these banks have debt due during the crisis. This means that half of the long-term debt was issued before the end of 1999. This might lead to a potential measurement error in this identification since the fraction of long-term debt that matured during the crisis which was issued after the end of 1999 does not take into account a bank that had a large amount of long-term debt that matured during the crisis where most of the debt was issued before the end of 1999. This introduces noise in the measurement of a bank’s exposure to the negative liquidity shock, so the estimation of the coefficient for this measure is expected to be biased downward unless there is a positive correlation between the long-term debt that was issued before the end of 1999 and the long-term debt that was issued after the end of 1999. In that case, positive correlation might lead to an upward bias in the coefficient estimate. The second concern related to this identification strategy is the importance of long-term debt relative to other types of liabilities in a bank’s balance sheet. In my sample, a bank’s total equity is on average 9% of its total assets. So the largest fraction, 91%, of its assets is financed by other types of liabilities. Out of these liabilities, deposits constitute 75% of the total assets. Other non-depository debt is 15% of the total assets. On average, 68% of the non-depository debt has a maturity of at least 1 year, defined as long-term debt, and the rest is short-term debt. One might argue that the amount of long-term debt is insignificant compared with other liabilities. So the effect of the long-term debt due during the crisis would also be insignificant. To understand the importance of the long-term debt that matured during the last financial crisis, one can think of the average bank in my sample with total assets of $27.3 billion. If we assume that the average bank in my sample has an average fraction of long-term debt that matured during the crisis (approximately 1.25%), the amount of this debt would be $341.3 million. This is an important amount of debt that matured during the crisis for one bank. If we look at the largest bank in my sample, the amount of long-term debt maturing during the crisis was $20.87 billion and this bank had $1.88 trillion total assets at the end of 2006.6 Although the fraction of long-term debt is quite small, the amount of long-term debt that matured during the crisis was still significantly important. 2.2 Matching banks to control for loan demand One of the key issues in determining the effects of a negative liquidity shock on banks’ loan supply is controlling for changes in the loan demand. The reason behind this concern is that the same economic conditions that lead to negative liquidity shocks can have a direct effect on the demand for loans. Employing a DID methodology helps to control for the systemic shock on loan demand during the crisis that affects all banks in my sample and for the possible differences in loan demand between banks with different exposure to the liquidity shock independent of the year. However, the effect of a negative liquidity shock on loan demand might differ geographically. To control for the differences in the effect of the last financial crisis on local loan demand, I use a matched bank approach that is proposed by Carlson, Shan, and Warusawitharana (2013) where each treated bank affected by the liquidity shock is matched with a set of control banks that are located in the same MSAs. Treated banks are defined as affected banks that had a positive amount of long-term debt matured during the crisis. Control banks are defined as unaffected neighboring banks that are located in the same MSAs with treated banks for each year. The reason behind this approach is that banks located in the same neighborhood face the same economic environment, so any change in the loan demand for the affected bank is expected to be the same for the matched unaffected bank. One important assumption in the DID estimation method is that the average change in the amount of loans would have been the same for both the treated and control banks in the absence of the negative liquidity shock. Table III compares the changes in the amount of loans over the period 2003–07 prior to the financial crisis. The Wilcoxon rank-sum test results show no statistical difference in the changes in the amount of loans for treated banks relative to control banks. This supports the identifying assumption of the empirical strategy. Figure 1 further presents the plots of the amount of loans for treated versus control banks before and after the crisis. The graph shows parallel trends in the amount of loans between treated and control banks before the beginning of the financial crisis. One can see the apparent difference in the trends between these groups after the crisis which provides further motivation for the DID estimation method. Table III. Trend comparison for treated and control banks before the crisis This table compares changes in total loans for treated and control banks over the period 2004–07 prior to the crisis. Treated banks are the banks with a positive amount of long-term debt that matured during the crisis. Control banks are defined as unaffected neighboring banks that are located in the same MSAs with treated banks. Change in total loans is calculated as the difference in total loans for 2 years divided by total loans in the previous year. The p-values of the Wilcoxon rank-sum test are presented in the rightmost column. Time period Treated Control Wilcoxon rank-sum test 2006–07 0.124 0.140 0.304 2005–06 0.168 0.183 0.674 2004–05 0.154 0.217 0.150 2003–04 0.182 0.260 0.500 Time period Treated Control Wilcoxon rank-sum test 2006–07 0.124 0.140 0.304 2005–06 0.168 0.183 0.674 2004–05 0.154 0.217 0.150 2003–04 0.182 0.260 0.500 Table III. Trend comparison for treated and control banks before the crisis This table compares changes in total loans for treated and control banks over the period 2004–07 prior to the crisis. Treated banks are the banks with a positive amount of long-term debt that matured during the crisis. Control banks are defined as unaffected neighboring banks that are located in the same MSAs with treated banks. Change in total loans is calculated as the difference in total loans for 2 years divided by total loans in the previous year. The p-values of the Wilcoxon rank-sum test are presented in the rightmost column. Time period Treated Control Wilcoxon rank-sum test 2006–07 0.124 0.140 0.304 2005–06 0.168 0.183 0.674 2004–05 0.154 0.217 0.150 2003–04 0.182 0.260 0.500 Time period Treated Control Wilcoxon rank-sum test 2006–07 0.124 0.140 0.304 2005–06 0.168 0.183 0.674 2004–05 0.154 0.217 0.150 2003–04 0.182 0.260 0.500 Figure 1. View largeDownload slide Trends in total loans before and after the crisis. This figure shows the trends in total loans before and after the crisis for treated and control banks. Treated banks are the banks with a positive amount of long-term debt that matured during the crisis. Control banks are defined as unaffected neighboring banks that are located in the same MSAs with treated banks. Total loans is calculated as the amount of total loans divided by 2005 total assets. Figure 1. View largeDownload slide Trends in total loans before and after the crisis. This figure shows the trends in total loans before and after the crisis for treated and control banks. Treated banks are the banks with a positive amount of long-term debt that matured during the crisis. Control banks are defined as unaffected neighboring banks that are located in the same MSAs with treated banks. Total loans is calculated as the amount of total loans divided by 2005 total assets. 3. The Transmission of Negative Liquidity Shocks In this section, I explore the transmission of negative liquidity shocks to the real economy through a reduction in banks’ loan supply in three steps: first, I show the effect of an increase in the fraction of long-term debt that matured during the crisis on the banks’ outstanding long-term debt. I expect that banks with more long-term debt that matured during the crisis would experience a larger reduction in their outstanding long-term debt since the cost of issuing long-term debt increased significantly in this time period. Second, I study whether the negative liquidity shock affects banks’ loan supply. I expect that banks with a higher exposure to the liquidity shock would reduce their lending more. Third, I analyze the impact of the negative liquidity shock on the real economy by studying the effect on house prices at the MSA level. I expect that house prices would decline more strongly in the MSAs where banks that are affected more by the negative liquidity shock have branches. 3.1 The effect on the outstanding long-term debt This section studies the effect of the negative liquidity shock on banks’ outstanding long-term debt. To study the effect of the shock, I use a DID estimation method with a continuous treatment variable, Di, and estimate LTDit=α0+α1Di×Pt+α2Di+α3Pt+θXi+δt+uit, (1) where LTDit is the outstanding long-term debt for bank i in time period t divided by total assets calculated in 2006.7 The time period captures 4 years: 2006, 2007, 2008, and 2009. Pt is the posttreatment indicator that is equal to one in 2008 and 2009, and zero in 2006 and 2007. Including the level Pt controls for trends common to all banks independent of their exposure to the liquidity shock. For example, if the debt holdings of banks are decreasing during the crisis due to uncertainty in the markets, α3 will capture this variation. Di is the continuous treatment variable and used to measure each treated bank’s exposure to the negative liquidity shock. It is the fraction of long-term debt issued before the end of 2006 and matured during the crisis, and calculated as the long-term debt issued before the end of 2006 with a maturity from the beginning of 2008 until the end of 2009 divided by the long-term debt issued before the end of 2006 with a maturity from the beginning of 2008. The level Di controls for permanent differences between banks with different exposure to the liquidity shock independent of the year. For example, if banks with more exposure to the liquidity shock have a larger amount of long-term debt than banks with less exposure, then α2 should capture this variation. The coefficient of interest is α1, which corresponds to the interaction term between Di and Pt. This variable measures whether banks with higher exposure to the negative liquidity shock experienced a greater reduction in their outstanding long-term debt from the pre-crisis years 2006 and 2007 to the crisis years 2008 and 2009 (Roberts and Whited, 2013). Xi includes the 2006 values of the size of the banks, the long-term debt issued before the end of 2006, and a set of bank characteristics that proxy the CAMELS supervisory rating, for example, total equity, loan loss reserves, net interest income, ROA, cash, and deposits, calculated in 2006.8 δt is the year fixed effects. Table IV presents the results. Column (1) shows that a bank with 1 percentage point increase in its long-term debt ratio that matured during the crisis decreased its long-term debt by almost 0.04% of its total assets. To study the effect of bank capital ratios on the transmission of the liquidity shock to the real economy, I split the sample into under-capitalized and well-capitalized banks. A bank is defined as under-capitalized if its total equity ratio was below the median in 2006 and as well-capitalized otherwise. I expect that the negative liquidity shock affects an under-capitalized bank significantly more relative to a well-capitalized bank. Columns (3) and (5) present the results for well-capitalized and under-capitalized banks, respectively. Column (3) shows that a well-capitalized bank with a higher long-term debt ratio that matured during the crisis does not experience a significant change in its outstanding long-term debt. On the contrary, the decrease in the outstanding long-term debt is stronger for under-capitalized banks according to Column (5). An under-capitalized bank with 1 percentage point higher fraction of long-term debt that matured during the crisis decreased its long-term debt by 0.05% of its total assets. When I repeat the analysis using the natural logarithm of long-term debt rather than normalizing it by total assets, the results are very similar as shown in Table V. Table IV. The effect of the negative liquidity shock on banks’ outstanding long-term debt and their loan supply The regressions in this table examine the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. Columns (1) and (2) show results for all banks and Columns (3)–(6) show results for well-capitalized and under-capitalized banks. A bank is defined as well-capitalized if its total equity ratio was above the median in 2006, and as under-capitalized otherwise. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.037** –0.085** –0.011 –0.035 –0.053** –0.121** (0.016) (0.038) (0.027) (0.044) (0.023) (0.054) Maturing debt –0.022 –0.140*** –0.008 –0.079* –0.017 –0.180** (0.020) (0.034) (0.029) (0.045) (0.032) (0.068) Crisis 0.017*** 0.019 0.043*** 0.069** 0.043*** 0.016 (0.004) (0.019) (0.006) (0.031) (0.007) (0.025) Total equity –0.003** –0.009* –0.005** –0.016*** –0.017* –0.014 (0.001) (0.005) (0.002) (0.006) (0.009) (0.015) Loan loss reserves –0.021 0.015 –0.011 –0.037 –0.021 0.113 (0.021) (0.045) (0.021) (0.067) (0.029) (0.073) Net interest income –0.009 0.040* 0.012 0.069** –0.023 0.014 (0.010) (0.024) (0.010) (0.028) (0.015) (0.053) ROA 0.008 0.021 –0.016 –0.032 0.019 0.073 (0.013) (0.036) (0.017) (0.057) (0.022) (0.051) Cash –0.003** –0.005*** –0.007*** –0.011*** –0.003** –0.004 (0.001) (0.002) (0.002) (0.004) (0.001) (0.003) Deposits –0.004*** 0.004** –0.005*** –0.001 –0.003*** 0.006** (0.001) (0.001) (0.001) (0.002) (0.001) (0.003) Size –0.007* –0.001 –0.003 0.003 –0.010 –0.005 (0.004) (0.009) (0.004) (0.013) (0.007) (0.016) Debt 0.009*** 0.011 0.014*** 0.009 0.003 0.008 (0.003) (0.008) (0.004) (0.009) (0.004) (0.012) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R-squared 0.351 0.145 0.442 0.073 0.358 0.242 Observations 613 712 319 372 294 340 All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.037** –0.085** –0.011 –0.035 –0.053** –0.121** (0.016) (0.038) (0.027) (0.044) (0.023) (0.054) Maturing debt –0.022 –0.140*** –0.008 –0.079* –0.017 –0.180** (0.020) (0.034) (0.029) (0.045) (0.032) (0.068) Crisis 0.017*** 0.019 0.043*** 0.069** 0.043*** 0.016 (0.004) (0.019) (0.006) (0.031) (0.007) (0.025) Total equity –0.003** –0.009* –0.005** –0.016*** –0.017* –0.014 (0.001) (0.005) (0.002) (0.006) (0.009) (0.015) Loan loss reserves –0.021 0.015 –0.011 –0.037 –0.021 0.113 (0.021) (0.045) (0.021) (0.067) (0.029) (0.073) Net interest income –0.009 0.040* 0.012 0.069** –0.023 0.014 (0.010) (0.024) (0.010) (0.028) (0.015) (0.053) ROA 0.008 0.021 –0.016 –0.032 0.019 0.073 (0.013) (0.036) (0.017) (0.057) (0.022) (0.051) Cash –0.003** –0.005*** –0.007*** –0.011*** –0.003** –0.004 (0.001) (0.002) (0.002) (0.004) (0.001) (0.003) Deposits –0.004*** 0.004** –0.005*** –0.001 –0.003*** 0.006** (0.001) (0.001) (0.001) (0.002) (0.001) (0.003) Size –0.007* –0.001 –0.003 0.003 –0.010 –0.005 (0.004) (0.009) (0.004) (0.013) (0.007) (0.016) Debt 0.009*** 0.011 0.014*** 0.009 0.003 0.008 (0.003) (0.008) (0.004) (0.009) (0.004) (0.012) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R-squared 0.351 0.145 0.442 0.073 0.358 0.242 Observations 613 712 319 372 294 340 Table IV. The effect of the negative liquidity shock on banks’ outstanding long-term debt and their loan supply The regressions in this table examine the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. Columns (1) and (2) show results for all banks and Columns (3)–(6) show results for well-capitalized and under-capitalized banks. A bank is defined as well-capitalized if its total equity ratio was above the median in 2006, and as under-capitalized otherwise. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.037** –0.085** –0.011 –0.035 –0.053** –0.121** (0.016) (0.038) (0.027) (0.044) (0.023) (0.054) Maturing debt –0.022 –0.140*** –0.008 –0.079* –0.017 –0.180** (0.020) (0.034) (0.029) (0.045) (0.032) (0.068) Crisis 0.017*** 0.019 0.043*** 0.069** 0.043*** 0.016 (0.004) (0.019) (0.006) (0.031) (0.007) (0.025) Total equity –0.003** –0.009* –0.005** –0.016*** –0.017* –0.014 (0.001) (0.005) (0.002) (0.006) (0.009) (0.015) Loan loss reserves –0.021 0.015 –0.011 –0.037 –0.021 0.113 (0.021) (0.045) (0.021) (0.067) (0.029) (0.073) Net interest income –0.009 0.040* 0.012 0.069** –0.023 0.014 (0.010) (0.024) (0.010) (0.028) (0.015) (0.053) ROA 0.008 0.021 –0.016 –0.032 0.019 0.073 (0.013) (0.036) (0.017) (0.057) (0.022) (0.051) Cash –0.003** –0.005*** –0.007*** –0.011*** –0.003** –0.004 (0.001) (0.002) (0.002) (0.004) (0.001) (0.003) Deposits –0.004*** 0.004** –0.005*** –0.001 –0.003*** 0.006** (0.001) (0.001) (0.001) (0.002) (0.001) (0.003) Size –0.007* –0.001 –0.003 0.003 –0.010 –0.005 (0.004) (0.009) (0.004) (0.013) (0.007) (0.016) Debt 0.009*** 0.011 0.014*** 0.009 0.003 0.008 (0.003) (0.008) (0.004) (0.009) (0.004) (0.012) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R-squared 0.351 0.145 0.442 0.073 0.358 0.242 Observations 613 712 319 372 294 340 All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.037** –0.085** –0.011 –0.035 –0.053** –0.121** (0.016) (0.038) (0.027) (0.044) (0.023) (0.054) Maturing debt –0.022 –0.140*** –0.008 –0.079* –0.017 –0.180** (0.020) (0.034) (0.029) (0.045) (0.032) (0.068) Crisis 0.017*** 0.019 0.043*** 0.069** 0.043*** 0.016 (0.004) (0.019) (0.006) (0.031) (0.007) (0.025) Total equity –0.003** –0.009* –0.005** –0.016*** –0.017* –0.014 (0.001) (0.005) (0.002) (0.006) (0.009) (0.015) Loan loss reserves –0.021 0.015 –0.011 –0.037 –0.021 0.113 (0.021) (0.045) (0.021) (0.067) (0.029) (0.073) Net interest income –0.009 0.040* 0.012 0.069** –0.023 0.014 (0.010) (0.024) (0.010) (0.028) (0.015) (0.053) ROA 0.008 0.021 –0.016 –0.032 0.019 0.073 (0.013) (0.036) (0.017) (0.057) (0.022) (0.051) Cash –0.003** –0.005*** –0.007*** –0.011*** –0.003** –0.004 (0.001) (0.002) (0.002) (0.004) (0.001) (0.003) Deposits –0.004*** 0.004** –0.005*** –0.001 –0.003*** 0.006** (0.001) (0.001) (0.001) (0.002) (0.001) (0.003) Size –0.007* –0.001 –0.003 0.003 –0.010 –0.005 (0.004) (0.009) (0.004) (0.013) (0.007) (0.016) Debt 0.009*** 0.011 0.014*** 0.009 0.003 0.008 (0.003) (0.008) (0.004) (0.009) (0.004) (0.012) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R-squared 0.351 0.145 0.442 0.073 0.358 0.242 Observations 613 712 319 372 294 340 Table V. The effect of the negative liquidity shock on the natural logarithm of banks’ outstanding long-term debt and their loan supply The regressions in this table explore the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply. Dependent variables are the natural logarithm of outstanding long-term debt and total loans. Columns (1) and (2) show results for all banks and Columns (3)–(6) show results for well-capitalized and under-capitalized banks. A bank is defined as well-capitalized if its total equity ratio was above the median in 2006, and as under-capitalized otherwise. The variable “Crisis” is the post-treatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.609*** –0.108* –0.428 –0.018 –0.804*** –0.180** (0.155) (0.055) (0.300) (0.057) (0.163) (0.073) Maturing debt –0.497* –0.181*** –0.418 –0.120* –0.442 –0.218** (0.279) (0.046) (0.415) (0.064) (0.412) (0.085) Crisis 0.168*** –0.006 0.219*** 0.061** 0.383*** –0.012 (0.041) (0.019) (0.073) (0.030) (0.063) (0.024) Total equity –0.028* –0.011* –0.058** –0.020** –0.088 –0.014 (0.017) (0.006) (0.026) (0.008) (0.081) (0.019) Loan loss reserves –0.346 –0.009 –0.470* –0.075 –0.017 0.116 (0.256) (0.064) (0.278) (0.103) (0.296) (0.088) Net interest income –0.058 0.061** 0.212 0.097*** –0.254* 0.042 (0.123) (0.029) (0.142) (0.035) (0.153) (0.058) ROA –0.083 0.027 –0.388 –0.030 0.164 0.070 (0.180) (0.051) (0.241) (0.083) (0.233) (0.063) Cash –0.034** –0.009*** –0.099*** –0.015*** –0.024** –0.006** (0.016) (0.002) (0.030) (0.005) (0.011) (0.003) Deposits –0.032*** 0.005** –0.049*** –0.001 –0.026** 0.007** (0.010) (0.002) (0.015) (0.002) (0.011) (0.003) Size 0.936*** 1.001*** 0.965*** 1.011*** 0.915*** 0.986*** (0.048) (0.011) (0.058) (0.015) (0.079) (0.018) Debt 0.163*** 0.011 0.236*** 0.009 0.081** 0.003 (0.042) (0.010) (0.064) (0.011) (0.039) (0.014) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.889 0.983 0.899 0.985 0.895 0.981 Observations 612 712 318 372 294 340 All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.609*** –0.108* –0.428 –0.018 –0.804*** –0.180** (0.155) (0.055) (0.300) (0.057) (0.163) (0.073) Maturing debt –0.497* –0.181*** –0.418 –0.120* –0.442 –0.218** (0.279) (0.046) (0.415) (0.064) (0.412) (0.085) Crisis 0.168*** –0.006 0.219*** 0.061** 0.383*** –0.012 (0.041) (0.019) (0.073) (0.030) (0.063) (0.024) Total equity –0.028* –0.011* –0.058** –0.020** –0.088 –0.014 (0.017) (0.006) (0.026) (0.008) (0.081) (0.019) Loan loss reserves –0.346 –0.009 –0.470* –0.075 –0.017 0.116 (0.256) (0.064) (0.278) (0.103) (0.296) (0.088) Net interest income –0.058 0.061** 0.212 0.097*** –0.254* 0.042 (0.123) (0.029) (0.142) (0.035) (0.153) (0.058) ROA –0.083 0.027 –0.388 –0.030 0.164 0.070 (0.180) (0.051) (0.241) (0.083) (0.233) (0.063) Cash –0.034** –0.009*** –0.099*** –0.015*** –0.024** –0.006** (0.016) (0.002) (0.030) (0.005) (0.011) (0.003) Deposits –0.032*** 0.005** –0.049*** –0.001 –0.026** 0.007** (0.010) (0.002) (0.015) (0.002) (0.011) (0.003) Size 0.936*** 1.001*** 0.965*** 1.011*** 0.915*** 0.986*** (0.048) (0.011) (0.058) (0.015) (0.079) (0.018) Debt 0.163*** 0.011 0.236*** 0.009 0.081** 0.003 (0.042) (0.010) (0.064) (0.011) (0.039) (0.014) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.889 0.983 0.899 0.985 0.895 0.981 Observations 612 712 318 372 294 340 Table V. The effect of the negative liquidity shock on the natural logarithm of banks’ outstanding long-term debt and their loan supply The regressions in this table explore the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply. Dependent variables are the natural logarithm of outstanding long-term debt and total loans. Columns (1) and (2) show results for all banks and Columns (3)–(6) show results for well-capitalized and under-capitalized banks. A bank is defined as well-capitalized if its total equity ratio was above the median in 2006, and as under-capitalized otherwise. The variable “Crisis” is the post-treatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.609*** –0.108* –0.428 –0.018 –0.804*** –0.180** (0.155) (0.055) (0.300) (0.057) (0.163) (0.073) Maturing debt –0.497* –0.181*** –0.418 –0.120* –0.442 –0.218** (0.279) (0.046) (0.415) (0.064) (0.412) (0.085) Crisis 0.168*** –0.006 0.219*** 0.061** 0.383*** –0.012 (0.041) (0.019) (0.073) (0.030) (0.063) (0.024) Total equity –0.028* –0.011* –0.058** –0.020** –0.088 –0.014 (0.017) (0.006) (0.026) (0.008) (0.081) (0.019) Loan loss reserves –0.346 –0.009 –0.470* –0.075 –0.017 0.116 (0.256) (0.064) (0.278) (0.103) (0.296) (0.088) Net interest income –0.058 0.061** 0.212 0.097*** –0.254* 0.042 (0.123) (0.029) (0.142) (0.035) (0.153) (0.058) ROA –0.083 0.027 –0.388 –0.030 0.164 0.070 (0.180) (0.051) (0.241) (0.083) (0.233) (0.063) Cash –0.034** –0.009*** –0.099*** –0.015*** –0.024** –0.006** (0.016) (0.002) (0.030) (0.005) (0.011) (0.003) Deposits –0.032*** 0.005** –0.049*** –0.001 –0.026** 0.007** (0.010) (0.002) (0.015) (0.002) (0.011) (0.003) Size 0.936*** 1.001*** 0.965*** 1.011*** 0.915*** 0.986*** (0.048) (0.011) (0.058) (0.015) (0.079) (0.018) Debt 0.163*** 0.011 0.236*** 0.009 0.081** 0.003 (0.042) (0.010) (0.064) (0.011) (0.039) (0.014) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.889 0.983 0.899 0.985 0.895 0.981 Observations 612 712 318 372 294 340 All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.609*** –0.108* –0.428 –0.018 –0.804*** –0.180** (0.155) (0.055) (0.300) (0.057) (0.163) (0.073) Maturing debt –0.497* –0.181*** –0.418 –0.120* –0.442 –0.218** (0.279) (0.046) (0.415) (0.064) (0.412) (0.085) Crisis 0.168*** –0.006 0.219*** 0.061** 0.383*** –0.012 (0.041) (0.019) (0.073) (0.030) (0.063) (0.024) Total equity –0.028* –0.011* –0.058** –0.020** –0.088 –0.014 (0.017) (0.006) (0.026) (0.008) (0.081) (0.019) Loan loss reserves –0.346 –0.009 –0.470* –0.075 –0.017 0.116 (0.256) (0.064) (0.278) (0.103) (0.296) (0.088) Net interest income –0.058 0.061** 0.212 0.097*** –0.254* 0.042 (0.123) (0.029) (0.142) (0.035) (0.153) (0.058) ROA –0.083 0.027 –0.388 –0.030 0.164 0.070 (0.180) (0.051) (0.241) (0.083) (0.233) (0.063) Cash –0.034** –0.009*** –0.099*** –0.015*** –0.024** –0.006** (0.016) (0.002) (0.030) (0.005) (0.011) (0.003) Deposits –0.032*** 0.005** –0.049*** –0.001 –0.026** 0.007** (0.010) (0.002) (0.015) (0.002) (0.011) (0.003) Size 0.936*** 1.001*** 0.965*** 1.011*** 0.915*** 0.986*** (0.048) (0.011) (0.058) (0.015) (0.079) (0.018) Debt 0.163*** 0.011 0.236*** 0.009 0.081** 0.003 (0.042) (0.010) (0.064) (0.011) (0.039) (0.014) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.889 0.983 0.899 0.985 0.895 0.981 Observations 612 712 318 372 294 340 These results indicate that banks with a higher exposure to the liquidity shock decreased their outstanding long-term debt significantly more. The results on the difference between well-capitalized and under-capitalized banks further adds to the literature by showing that the effect of higher financing costs is stronger for under-capitalized banks during contraction times. 3.2 Within bank transmission This section studies whether the liquidity shock is transmitted from the liability side to the asset side of a bank’s balance sheet by a reduction in its loan supply. To study the effect of the shock on the amount of loans, I again use a DID estimation method with a continuous treatment variable, Di, and estimate a similar regression to the Equation (1) Lit=β0+β1Di×Pt+β2Di+β3Pt+θXi+δt+uit, (2) where the only difference is the dependent variable Lit which is the total loans for bank i in time period t divided by total assets calculated in 2006. All other independent variables and controls are the same with Equation (1). The coefficient of interest is β1 which captures whether banks with higher exposure to the negative liquidity shock experienced a greater reduction in their loan supply from the pre-crisis years 2006 and 2007 to the crisis years 2008 and 2009. Table IV presents the results of the regression. Column (2) estimates the impact of the liquidity shock on the amount of total loans for all banks. I find that 1 percentage point increase in the fraction of a bank’s long-term debt that matured during the crisis led to a decrease in its loan supply by almost 0.09% of its total assets, which is economically and statistically highly significant. According to Column (4), the liquidity shock did not have any significant effect on well-capitalized banks’ loan supply during the crisis. On the contrary, the effect is stronger for under-capitalized banks as shown in Column (6). An under-capitalized bank with 1 percentage point higher long-term debt ratio that matured during the crisis decreased its loan supply by 0.12% of its total assets.9 These results indicate that banks that are affected more by the liquidity shock decreased their loan supply significantly more. In addition, the negative effect is stronger for under-capitalized banks, whereas it disappears for well-capitalized banks. I follow Puri, Rocholl, and Steffen (2011) and ask the question whether banks affected by the liquidity shock reduced their lending to reduce portfolio risk or to preserve liquidity. To investigate this, I first analyze the effect of the liquidity shock on banks’ nonperforming loans, loan loss reserves until the end of 2011 to capture the long-term effects. Column (1) in Table VI show that 1 percentage point increase in the long-term debt ratio that matured during the crisis led to a significant decrease in a bank’s nonperforming loans by almost 0.01% of its total assets. This suggests that banks affected by the liquidity shock reduced the risk of their loan portfolio. This is supported by the reduction in loan loss reserves: a bank with a 0.10 higher long-term debt ratio that matured during the crisis decreased its loan loss reserves by 0.03% of its total assets. I next analyze the effect on banks’ cash holdings to answer the question whether these affected banks preserved liquidity by reducing their lending activities. As shown in Column (3), I find that an increase in the long-term debt ratio matured during the crisis did not have any significant effect on the cash holdings of a bank. This indicates that affected banks did not preserve liquidity. Taken together, these results suggest that the banks affected heavily by the negative liquidity shock decreased their loan supply significantly, and this enabled them to reduce their portfolio risk. Table VI. The effect of the negative liquidity shock on nonperforming loans, loan loss reserves and cash holdings The regressions in this table investigate the effect of the liquidity shock on banks’ nonperforming loans, loan loss reserves and cash holdings. Dependent variables are the nonperforming loans, loan loss reserves and cash holdings divided by 2006 total assets. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. For results on nonperforming loans and loan loss reserves, as shown in Columns (1) and (2), the crisis dummy is extended 2 more years and includes 2010 and 2011 to capture the long-term effect. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Dependent variable Nonperforming loans Loan loss reserves Cash holdings Maturing debt × Crisis –0.008* –0.003* –0.020 (0.005) (0.002) (0.029) Maturing debt –0.001 –0.002* 0.021 (0.003) (0.001) (0.015) Crisis 0.027*** 0.008*** 0.032*** (0.002) (0.001) (0.004) Total equity –0.001 –0.000 0.000 (0.000) (0.000) (0.001) Loan loss reserves 0.001 0.009*** –0.035* (0.004) (0.001) (0.020) Net interest income 0.006*** 0.001* 0.020 (0.002) (0.001) (0.015) ROA –0.006* 0.001 –0.018 (0.004) (0.001) (0.015) Cash –0.000 –0.000** 0.009*** (0.000) (0.000) (0.001) Deposits 0.000 0.000 –0.000 (0.000) (0.000) (0.001) Size 0.001 0.001** 0.003 (0.001) (0.000) (0.002) Debt 0.001 0.000** 0.001 (0.001) (0.000) (0.002) Time fixed effects Yes Yes Yes Adjusted R2 0.342 0.432 0.500 Observations 1045 1058 712 Dependent variable Nonperforming loans Loan loss reserves Cash holdings Maturing debt × Crisis –0.008* –0.003* –0.020 (0.005) (0.002) (0.029) Maturing debt –0.001 –0.002* 0.021 (0.003) (0.001) (0.015) Crisis 0.027*** 0.008*** 0.032*** (0.002) (0.001) (0.004) Total equity –0.001 –0.000 0.000 (0.000) (0.000) (0.001) Loan loss reserves 0.001 0.009*** –0.035* (0.004) (0.001) (0.020) Net interest income 0.006*** 0.001* 0.020 (0.002) (0.001) (0.015) ROA –0.006* 0.001 –0.018 (0.004) (0.001) (0.015) Cash –0.000 –0.000** 0.009*** (0.000) (0.000) (0.001) Deposits 0.000 0.000 –0.000 (0.000) (0.000) (0.001) Size 0.001 0.001** 0.003 (0.001) (0.000) (0.002) Debt 0.001 0.000** 0.001 (0.001) (0.000) (0.002) Time fixed effects Yes Yes Yes Adjusted R2 0.342 0.432 0.500 Observations 1045 1058 712 Table VI. The effect of the negative liquidity shock on nonperforming loans, loan loss reserves and cash holdings The regressions in this table investigate the effect of the liquidity shock on banks’ nonperforming loans, loan loss reserves and cash holdings. Dependent variables are the nonperforming loans, loan loss reserves and cash holdings divided by 2006 total assets. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. For results on nonperforming loans and loan loss reserves, as shown in Columns (1) and (2), the crisis dummy is extended 2 more years and includes 2010 and 2011 to capture the long-term effect. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Dependent variable Nonperforming loans Loan loss reserves Cash holdings Maturing debt × Crisis –0.008* –0.003* –0.020 (0.005) (0.002) (0.029) Maturing debt –0.001 –0.002* 0.021 (0.003) (0.001) (0.015) Crisis 0.027*** 0.008*** 0.032*** (0.002) (0.001) (0.004) Total equity –0.001 –0.000 0.000 (0.000) (0.000) (0.001) Loan loss reserves 0.001 0.009*** –0.035* (0.004) (0.001) (0.020) Net interest income 0.006*** 0.001* 0.020 (0.002) (0.001) (0.015) ROA –0.006* 0.001 –0.018 (0.004) (0.001) (0.015) Cash –0.000 –0.000** 0.009*** (0.000) (0.000) (0.001) Deposits 0.000 0.000 –0.000 (0.000) (0.000) (0.001) Size 0.001 0.001** 0.003 (0.001) (0.000) (0.002) Debt 0.001 0.000** 0.001 (0.001) (0.000) (0.002) Time fixed effects Yes Yes Yes Adjusted R2 0.342 0.432 0.500 Observations 1045 1058 712 Dependent variable Nonperforming loans Loan loss reserves Cash holdings Maturing debt × Crisis –0.008* –0.003* –0.020 (0.005) (0.002) (0.029) Maturing debt –0.001 –0.002* 0.021 (0.003) (0.001) (0.015) Crisis 0.027*** 0.008*** 0.032*** (0.002) (0.001) (0.004) Total equity –0.001 –0.000 0.000 (0.000) (0.000) (0.001) Loan loss reserves 0.001 0.009*** –0.035* (0.004) (0.001) (0.020) Net interest income 0.006*** 0.001* 0.020 (0.002) (0.001) (0.015) ROA –0.006* 0.001 –0.018 (0.004) (0.001) (0.015) Cash –0.000 –0.000** 0.009*** (0.000) (0.000) (0.001) Deposits 0.000 0.000 –0.000 (0.000) (0.000) (0.001) Size 0.001 0.001** 0.003 (0.001) (0.000) (0.002) Debt 0.001 0.000** 0.001 (0.001) (0.000) (0.002) Time fixed effects Yes Yes Yes Adjusted R2 0.342 0.432 0.500 Observations 1045 1058 712 3.2.a. Cross-sectional results An important question is whether bank-specific characteristics affect the impact of the liquidity shock on banks’ loan supply. Following the literature, I am interested in two bank characteristics: deposits and short-term debt. First, I estimate the regressions separately for banks with higher deposits and banks with lower deposits. A bank is defined to have higher deposits if its total deposits ratio was above the median in 2006 and lower deposits otherwise. Recent studies show that banks with higher deposits cut their overall lending less severely during the last financial crisis (see, e.g., Ivashina and Scharfstein, 2010; Cornett et al., 2011; Dagher and Kazimov, 2015). They argue that banks with better access to deposit financing are more robust to liquidity shocks because deposits are a more stable source of funding (Gatev and Strahan, 2006). Thus, I expect a smaller effect of the liquidity shock on banks with higher deposits relative to banks with lower deposits. Table VII shows that the liquidity shock does not have any significant effect on the change in the amount of total loans for banks with higher deposits although they decreased their long-term debt significantly. On the contrary, banks with lower deposits experienced a significant decline in their loans when they were hit by the liquidity shock. A bank with 1 percentage point higher long-term debt ratio that matured during the crisis had a significant reduction in its loan supply by almost 0.10% of its total assets and in its long-term debt by almost 0.04% of its total assets.10 These results support the recent literature on the conclusion that banks with larger amount of deposits are more robust to negative liquidity shocks. Table VII. Cross-sectional results: deposits Regressions show the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply by deposits. A bank is defined to have higher deposits if its total deposits ratio was above the median in 2006 and lower deposits otherwise. Columns (1) and (2) show results for banks with higher deposits and Columns (3) and (4) show results for banks with lower deposits. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Higher deposits Lower deposits Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.036** –0.055 –0.035 –0.095** (0.017) (0.057) (0.022) (0.038) Maturing debt –0.003 –0.105* –0.030 –0.192*** (0.020) (0.053) (0.033) (0.048) Crisis 0.022*** 0.005 0.043*** 0.055*** (0.005) (0.018) (0.006) (0.018) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.375 0.043 0.211 0.210 Observations 285 353 328 359 Higher deposits Lower deposits Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.036** –0.055 –0.035 –0.095** (0.017) (0.057) (0.022) (0.038) Maturing debt –0.003 –0.105* –0.030 –0.192*** (0.020) (0.053) (0.033) (0.048) Crisis 0.022*** 0.005 0.043*** 0.055*** (0.005) (0.018) (0.006) (0.018) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.375 0.043 0.211 0.210 Observations 285 353 328 359 Table VII. Cross-sectional results: deposits Regressions show the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply by deposits. A bank is defined to have higher deposits if its total deposits ratio was above the median in 2006 and lower deposits otherwise. Columns (1) and (2) show results for banks with higher deposits and Columns (3) and (4) show results for banks with lower deposits. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Higher deposits Lower deposits Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.036** –0.055 –0.035 –0.095** (0.017) (0.057) (0.022) (0.038) Maturing debt –0.003 –0.105* –0.030 –0.192*** (0.020) (0.053) (0.033) (0.048) Crisis 0.022*** 0.005 0.043*** 0.055*** (0.005) (0.018) (0.006) (0.018) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.375 0.043 0.211 0.210 Observations 285 353 328 359 Higher deposits Lower deposits Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.036** –0.055 –0.035 –0.095** (0.017) (0.057) (0.022) (0.038) Maturing debt –0.003 –0.105* –0.030 –0.192*** (0.020) (0.053) (0.033) (0.048) Crisis 0.022*** 0.005 0.043*** 0.055*** (0.005) (0.018) (0.006) (0.018) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.375 0.043 0.211 0.210 Observations 285 353 328 359 Second, I estimate the main specification separately for banks with higher short-term debt holdings and banks with lower short-term debt holdings. I define a bank with higher short-term debt if its short-term debt ratio was above the median in 2006, and as a bank with lower short-term debt otherwise. A bank with more short-term debt holdings is expected to be affected more strongly by the negative liquidity shock since these banks faced higher roll-over risk during the 2007–09 financial crisis (see, e.g., Cornett et al., 2011; Huang and Ratnovski, 2011; Dagher and Kazimov, 2015). This is due to the fact that short-term financiers are uninsured creditors and more at risk of realizing losses. Banks that rely more on short-term debt had to find alternative financing sources to cover their shortfalls during the crisis. Thus, a bank with a high amount of short-term debt was affected more strongly from its maturing long-term debt than a bank with a low amount of short-term debt. Table VIII reports the results. I find that banks with higher short-term debt ratios decreased their long-term debt by 0.04% of their total assets and their loan supply by almost 0.12% of their total assets if there is 1 percentage point increase in the fraction of their long-term debt that matured during the crisis. The liquidity shock does not have any significant effect on the banks with lower short-term debt ratios. According to my results, banks with higher short-term debt holdings were more prone to the negative liquidity shock which is consistent with the literature. Table VIII. Cross-sectional results: short-term debt Regressions present the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply by short-term debt holdings. A bank is defined to have higher short-term debt if its short-term debt ratio was above the median in 2006, and lower short-term debt otherwise. Columns (1) and (2) show results for banks with higher short-term debt and Columns (3) and (4) show results for banks with lower short-term debt. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Higher short-term debt Lower short-term debt Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.040** –0.118** –0.039 0.052 (0.019) (0.046) (0.024) (0.079) Maturing debt –0.031* –0.146*** 0.059* –0.151*** (0.017) (0.042) (0.032) (0.057) Crisis 0.043*** 0.065** 0.025*** 0.072*** (0.006) (0.032) (0.005) (0.016) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R-squared 0.344 0.147 0.438 0.136 Observations 333 407 280 305 Higher short-term debt Lower short-term debt Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.040** –0.118** –0.039 0.052 (0.019) (0.046) (0.024) (0.079) Maturing debt –0.031* –0.146*** 0.059* –0.151*** (0.017) (0.042) (0.032) (0.057) Crisis 0.043*** 0.065** 0.025*** 0.072*** (0.006) (0.032) (0.005) (0.016) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R-squared 0.344 0.147 0.438 0.136 Observations 333 407 280 305 Table VIII. Cross-sectional results: short-term debt Regressions present the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply by short-term debt holdings. A bank is defined to have higher short-term debt if its short-term debt ratio was above the median in 2006, and lower short-term debt otherwise. Columns (1) and (2) show results for banks with higher short-term debt and Columns (3) and (4) show results for banks with lower short-term debt. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Higher short-term debt Lower short-term debt Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.040** –0.118** –0.039 0.052 (0.019) (0.046) (0.024) (0.079) Maturing debt –0.031* –0.146*** 0.059* –0.151*** (0.017) (0.042) (0.032) (0.057) Crisis 0.043*** 0.065** 0.025*** 0.072*** (0.006) (0.032) (0.005) (0.016) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R-squared 0.344 0.147 0.438 0.136 Observations 333 407 280 305 Higher short-term debt Lower short-term debt Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.040** –0.118** –0.039 0.052 (0.019) (0.046) (0.024) (0.079) Maturing debt –0.031* –0.146*** 0.059* –0.151*** (0.017) (0.042) (0.032) (0.057) Crisis 0.043*** 0.065** 0.025*** 0.072*** (0.006) (0.032) (0.005) (0.016) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R-squared 0.344 0.147 0.438 0.136 Observations 333 407 280 305 3.2.b. Results for different types of loans I next investigate the impact of the liquidity shock on different types of loans. I consider three types of loans which constitute 86.9% of total loans. The largest one is the real estate loans which are 53% of total loans. The second one is the C&I loans with 20.9% followed by consumer loans that represent 13% of total loans. Table IX presents the results of the regressions for three types of loans. Column (1) estimates the impact of the liquidity shock on real estate loans for all banks. I find that 1 percentage point increase in the fraction of a bank’s long-term debt that matured during the crisis generated a decrease in its real estate loans by 0.06% of its total assets, which is economically and statistically highly significant. Column (2) shows that a bank with 1 percentage point higher long-term debt ratio that matured during the crisis decreased its consumer loans by 0.02% of its total assets. According to Column (3), the effect of the liquidity shock on C&I loans is insignificant. I then repeat the analysis for well-capitalized and under-capitalized banks. The results for well-capitalized banks, shown in Columns (1), (2), and (3) in Table X, indicate that the liquidity shock did not have a significant effect on any of the three types of loans. However, the negative effect of the liquidity shock on the real estate loans and consumer loans is stronger for under-capitalized banks. As shown in Columns (4) and (5), an under-capitalized bank with 1 percentage point increase in its long-term debt ratio that matured during the crisis had a significant decrease in its real estate loans by 0.08% of its total assets and in its consumer loans by 0.02% of its total assets. Table IX. The effect of the negative liquidity shock on different types of loans This table shows the regression results for examining the effect of the liquidity shock on different types of loans: real-estate loans, consumer loans and C&I loans. Dependent variables are these three types of loans divided by 2006 total assets. The variable “Crisis” is the post-treatment indicator which is equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Dependent variable Real estate loans Consumer loans C&I loans Maturing debt × Crisis –0.063** –0.023*** 0.009 (0.028) (0.009) (0.016) Maturing debt –0.015 0.002 –0.090*** (0.034) (0.026) (0.025) Crisis 0.043*** 0.004 0.010 (0.009) (0.004) (0.007) Total equity –0.002 –0.005** –0.001 (0.005) (0.002) (0.003) Loan loss reserves –0.074 0.065** 0.021 (0.047) (0.027) (0.035) Net interest income 0.003 –0.009 0.017 (0.024) (0.011) (0.017) ROA 0.000 0.021 0.023 (0.034) (0.018) (0.025) Cash –0.003** –0.001** –0.001 (0.001) (0.001) (0.001) Deposits –0.000 0.001 0.002** (0.002) (0.001) (0.001) Size –0.033*** 0.015*** 0.009 (0.009) (0.005) (0.007) Debt –0.006 0.007 0.014** (0.009) (0.004) (0.006) Time fixed effects Yes Yes Yes Adjusted R-squared 0.169 0.157 0.070 Observations 611 646 667 All banks Dependent variable Real estate loans Consumer loans C&I loans Maturing debt × Crisis –0.063** –0.023*** 0.009 (0.028) (0.009) (0.016) Maturing debt –0.015 0.002 –0.090*** (0.034) (0.026) (0.025) Crisis 0.043*** 0.004 0.010 (0.009) (0.004) (0.007) Total equity –0.002 –0.005** –0.001 (0.005) (0.002) (0.003) Loan loss reserves –0.074 0.065** 0.021 (0.047) (0.027) (0.035) Net interest income 0.003 –0.009 0.017 (0.024) (0.011) (0.017) ROA 0.000 0.021 0.023 (0.034) (0.018) (0.025) Cash –0.003** –0.001** –0.001 (0.001) (0.001) (0.001) Deposits –0.000 0.001 0.002** (0.002) (0.001) (0.001) Size –0.033*** 0.015*** 0.009 (0.009) (0.005) (0.007) Debt –0.006 0.007 0.014** (0.009) (0.004) (0.006) Time fixed effects Yes Yes Yes Adjusted R-squared 0.169 0.157 0.070 Observations 611 646 667 Table IX. The effect of the negative liquidity shock on different types of loans This table shows the regression results for examining the effect of the liquidity shock on different types of loans: real-estate loans, consumer loans and C&I loans. Dependent variables are these three types of loans divided by 2006 total assets. The variable “Crisis” is the post-treatment indicator which is equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Dependent variable Real estate loans Consumer loans C&I loans Maturing debt × Crisis –0.063** –0.023*** 0.009 (0.028) (0.009) (0.016) Maturing debt –0.015 0.002 –0.090*** (0.034) (0.026) (0.025) Crisis 0.043*** 0.004 0.010 (0.009) (0.004) (0.007) Total equity –0.002 –0.005** –0.001 (0.005) (0.002) (0.003) Loan loss reserves –0.074 0.065** 0.021 (0.047) (0.027) (0.035) Net interest income 0.003 –0.009 0.017 (0.024) (0.011) (0.017) ROA 0.000 0.021 0.023 (0.034) (0.018) (0.025) Cash –0.003** –0.001** –0.001 (0.001) (0.001) (0.001) Deposits –0.000 0.001 0.002** (0.002) (0.001) (0.001) Size –0.033*** 0.015*** 0.009 (0.009) (0.005) (0.007) Debt –0.006 0.007 0.014** (0.009) (0.004) (0.006) Time fixed effects Yes Yes Yes Adjusted R-squared 0.169 0.157 0.070 Observations 611 646 667 All banks Dependent variable Real estate loans Consumer loans C&I loans Maturing debt × Crisis –0.063** –0.023*** 0.009 (0.028) (0.009) (0.016) Maturing debt –0.015 0.002 –0.090*** (0.034) (0.026) (0.025) Crisis 0.043*** 0.004 0.010 (0.009) (0.004) (0.007) Total equity –0.002 –0.005** –0.001 (0.005) (0.002) (0.003) Loan loss reserves –0.074 0.065** 0.021 (0.047) (0.027) (0.035) Net interest income 0.003 –0.009 0.017 (0.024) (0.011) (0.017) ROA 0.000 0.021 0.023 (0.034) (0.018) (0.025) Cash –0.003** –0.001** –0.001 (0.001) (0.001) (0.001) Deposits –0.000 0.001 0.002** (0.002) (0.001) (0.001) Size –0.033*** 0.015*** 0.009 (0.009) (0.005) (0.007) Debt –0.006 0.007 0.014** (0.009) (0.004) (0.006) Time fixed effects Yes Yes Yes Adjusted R-squared 0.169 0.157 0.070 Observations 611 646 667 Table X. The effect of the negative liquidity shock on different types of loans for well-capitalized and under-capitalized banks The regressions in this table examine the effect of the liquidity shock on different types of loans, namely real-estate loans, consumer loans and C&I loans, for well-capitalized and under-capitalized banks, separately. A bank is defined as well-capitalized if its total equity ratio was above the median in 2006, and as under-capitalized otherwise. Dependent variables are these three types of loans divided by 2006 total assets. The variable “Crisis” is the post-treatment indicator which is equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Well-capitalized banks Under-capitalized banks Dependent variable Real estate Consumer Commercial Real estate Consumer Commercial Maturing debt × Crisis –0.061 –0.006 0.012 –0.079** –0.024** 0.009 (0.040) (0.018) (0.016) (0.036) (0.011) (0.032) Maturing debt 0.041 0.054 –0.118*** –0.041 –0.010 –0.070 (0.056) (0.038) (0.038) (0.066) (0.030) (0.058) Crisis 0.041*** 0.014** –0.008 0.042*** 0.001 0.016 (0.011) (0.006) (0.006) (0.013) (0.003) (0.013) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.201 0.260 0.118 0.170 0.092 0.056 Observations 331 345 345 280 301 322 Well-capitalized banks Under-capitalized banks Dependent variable Real estate Consumer Commercial Real estate Consumer Commercial Maturing debt × Crisis –0.061 –0.006 0.012 –0.079** –0.024** 0.009 (0.040) (0.018) (0.016) (0.036) (0.011) (0.032) Maturing debt 0.041 0.054 –0.118*** –0.041 –0.010 –0.070 (0.056) (0.038) (0.038) (0.066) (0.030) (0.058) Crisis 0.041*** 0.014** –0.008 0.042*** 0.001 0.016 (0.011) (0.006) (0.006) (0.013) (0.003) (0.013) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.201 0.260 0.118 0.170 0.092 0.056 Observations 331 345 345 280 301 322 Table X. The effect of the negative liquidity shock on different types of loans for well-capitalized and under-capitalized banks The regressions in this table examine the effect of the liquidity shock on different types of loans, namely real-estate loans, consumer loans and C&I loans, for well-capitalized and under-capitalized banks, separately. A bank is defined as well-capitalized if its total equity ratio was above the median in 2006, and as under-capitalized otherwise. Dependent variables are these three types of loans divided by 2006 total assets. The variable “Crisis” is the post-treatment indicator which is equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Well-capitalized banks Under-capitalized banks Dependent variable Real estate Consumer Commercial Real estate Consumer Commercial Maturing debt × Crisis –0.061 –0.006 0.012 –0.079** –0.024** 0.009 (0.040) (0.018) (0.016) (0.036) (0.011) (0.032) Maturing debt 0.041 0.054 –0.118*** –0.041 –0.010 –0.070 (0.056) (0.038) (0.038) (0.066) (0.030) (0.058) Crisis 0.041*** 0.014** –0.008 0.042*** 0.001 0.016 (0.011) (0.006) (0.006) (0.013) (0.003) (0.013) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.201 0.260 0.118 0.170 0.092 0.056 Observations 331 345 345 280 301 322 Well-capitalized banks Under-capitalized banks Dependent variable Real estate Consumer Commercial Real estate Consumer Commercial Maturing debt × Crisis –0.061 –0.006 0.012 –0.079** –0.024** 0.009 (0.040) (0.018) (0.016) (0.036) (0.011) (0.032) Maturing debt 0.041 0.054 –0.118*** –0.041 –0.010 –0.070 (0.056) (0.038) (0.038) (0.066) (0.030) (0.058) Crisis 0.041*** 0.014** –0.008 0.042*** 0.001 0.016 (0.011) (0.006) (0.006) (0.013) (0.003) (0.013) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.201 0.260 0.118 0.170 0.092 0.056 Observations 331 345 345 280 301 322 According to these results, banks with a higher fraction of long-term debt that matured during the crisis decreased their real estate loans and consumer loans, whereas there are no significant reduction in their C&I loans. The reduction is, particularly, strong for real estate loans as compared with consumer loans. This supports the results of Puri, Rocholl, and Steffen (2011) on German savings banks. They find that banks affected by the US financial crisis rejected substantially more loan applications than non-affected banks, and this result is particularly strong for mortgage loans relative to consumer loans. One possible explanation for this result could be that banks are willing to continue their lending relationships with existing borrower firms so that they are more likely to cut on real estate and consumer loans relative to C&I loans in distress times. In relation to this, Iyer et al. (2014) show that higher strength of an existing lending relationship lowers the negative effect of liquidity shocks on banks’ loan supply. Their results suggest that banking relationships mitigate credit supply restrictions. Another explanation could be that borrower firms draw funds from existing credit lines with their banks when the supply of market liquidity falls. Ivashina and Scharfstein (2010) present that firms drew on their credit lines during the last financial crisis and this led to a spike in C&I loans reported on bank balance sheets. This might indicate that banks are more likely to reduce lending in real estate and consumer loans when liquidity becomes scarce because they might have to continue to provide liquidity for borrower firms through the existing credit lines. 3.3 Transmission to the real economy: evidence from house prices In normal times, when there are no frictions in lending, the effect of the negative liquidity shock on house prices through a reduction in real estate loans can be offset by switching across banks. However, during crisis times, frictions in the economy prevent switching across banks, and the liquidity shock is expected to affect the allocation of lending, and as a result house prices might be affected. As banks reduce their real estate loans, the demand for housing decreases and this pushes house prices down. I use the HPI data at the MSA level to study the transmission of the liquidity shock to the real economy by exploring the effect on house prices. To test for the effect, I again use a DID estimation method with a continuous treatment variable, Djt, and estimate △Hjt=γ0+γ1Djt×Pt+γ2Djt+γ3Pt+θYjt+μZjt+λj+δt+ujt, (3) where Hjt is the natural logarithm of the MSA HPI as given by FHFA for MSA j in time period t. △Hjt which is the change in the natural logarithm of HPIs, the growth rate of house prices, is used as the dependent variable because it is argued in the literature that house prices in the USA display heterogeneous trends (Favara and Imbs, 2015) and that HPI has no economic interpretation (Himmelberg, Mayer, and Sinai, 2005). The time period captures 4 years: 2006, 2007, 2008, and 2009. Pt is the posttreatment indicator that is equal to one in 2008 and 2009, and zero in 2006 and 2007. Djt is the weighted average fraction of long-term debt that was issued before the end of 2006 and that matured during the crisis at the MSA level. Following Berger and Bouwman (2009), Flannery and Lin (2015), and Dursun-de Neef and Schandlbauer (2017), the amount of deposits that each bank holds in an MSA is used as a weight that reflects how important an affected bank is for an MSA. Djt is used to measure the exposure of an MSA to the liquidity shock through the branches of the affected banks in an MSA.11 The intuition is that an MSA is affected by the liquidity shock if affected banks have branches in this MSA. Yjt includes the weighted average 2006 values of the size of the banks in each MSA and the long-term debt issued before the end of 2006. Zjt summarizes additional determinants of house prices documented in the literature. I follow Lamont and Stein (1999) and Favara and Imbs (2015), and include contemporaneous and lagged income per capita and population to control for local influences on the real estate market. Following Case and Shiller (1989), I additionally include a lagged value of the house prices and the change in the house prices. I also control for socioeconomic factors at the MSA level by including median age, ratio of owner occupied housing units, fraction of the population above 25 years old with at least a college degree and unemployment rate. λj is the MSA fixed effects and δt is the year fixed effects. The coefficient of interest is γ1, which measures whether MSAs with higher exposure to the negative liquidity shock, through the branches of affected banks that are located in these MSAs, experienced a greater reduction in the house prices from the pre-crisis years 2006 and 2007 to the crisis years 2008 and 2009. To study the effect of the negative liquidity shock on house prices, I use the fraction of deposits that each bank holds in its branches in an MSA as weights to measure the effect of a reduction in the amount of that bank’s real estate loans on house prices in that MSA. I cannot use the fraction of real estate loans due to the lack of data on the outstanding real estate loans at the branch level for each bank. This requires the assumption that the distribution of deposits is similar to the distribution of real estate loans among banks in an MSA. To provide evidence on this assumption, I collect data on the volume of mortgage loans originated by each bank in each MSA from SNL Financial. These data are taken from the Home Mortgage Disclosure Act (HMDA) database. Using the HMDA, data enable me to calculate the market share of each bank in the mortgage loans market for newly issued loans in each MSA and compare this with the deposits market share of that bank in that MSA.12Table XI presents that the correlation between the fraction of mortgage loans originated by a bank in an MSA and the fraction of that bank’s deposits in that MSA is 0.209, and it is highly significant at 1% level. This supports the argument that deposit distribution matches housing loan distribution among banks in an MSA. So the fraction of deposits that each bank holds in an MSA can be used to measure how important that bank is in the real estate loans market in that MSA. Table XI. Correlations: mortgage loans and deposits This table shows the correlation between the fraction of mortgage loans originated by each bank in an MSA with the fraction of deposits that the bank holds in that MSA for 4 years: 2006, 2007, 2008, and 2009. *p < 0.10, **p < 0.05, ***p < 0.01. Correlations Variable Mortgage loans Deposits Mortgage loans 1.000 Deposits 0.209*** 1.000 Correlations Variable Mortgage loans Deposits Mortgage loans 1.000 Deposits 0.209*** 1.000 Table XI. Correlations: mortgage loans and deposits This table shows the correlation between the fraction of mortgage loans originated by each bank in an MSA with the fraction of deposits that the bank holds in that MSA for 4 years: 2006, 2007, 2008, and 2009. *p < 0.10, **p < 0.05, ***p < 0.01. Correlations Variable Mortgage loans Deposits Mortgage loans 1.000 Deposits 0.209*** 1.000 Correlations Variable Mortgage loans Deposits Mortgage loans 1.000 Deposits 0.209*** 1.000 Table XII reports the results of the effect of the negative liquidity shock on the growth rate of house prices. The results show that 1 percentage point increase in the weighted average fraction of long-term debt that matured during the crisis resulted in a 0.13 percentage point decrease in the growth rate of house prices. I next repeat the analysis for MSAs with more well-capitalized banks and MSAs with more under-capitalized banks, separately. I define an MSA with more well-capitalized banks if the weighted average total equity ratio was above median total equity ratio in 2006, and as an MSA with more under-capitalized banks otherwise. I expect that the negative effect of the liquidity shock is stronger for MSAs with more under-capitalized banks since these are the banks that cut their real estate loans significantly. According to the results for MSAs with more well-capitalized banks, 1 percentage point increase in the weighted average fraction of long-term debt that matured during the crisis led to 0.12 percentage point decrease in the growth rate of house prices, whereas the decrease is 0.18 percentage point in the growth rate for MSAs with more under-capitalized banks. Table XII. The transmission of the negative liquidity shock to the real economy: evidence from MSA house prices The regressions in this table investigate the effect of the liquidity shock on MSA house prices. The dependent variable is the change in the natural logarithm of HPIs. The variable “Maturing debt” measures each MSA’s exposure to the negative liquidity shock and is calculated as the weighted average fraction of long-term debt that matured during the crisis, where the amount of deposits that each bank holds in an MSA is used as weights. All other variables are defined in Table I. All columns include controls for year fixed effects, MSA fixed effects and 1 year lagged MSA characteristics. Intercept, year fixed effects and MSA fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the MSA level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All MSAs Well-capitalized Under-capitalized Dependent variable △House price △House price △House price Maturing debt × Crisis –0.130*** –0.120* –0.178*** (0.045) (0.064) (0.061) Maturing debt –0.037 –0.174** 0.083 (0.072) (0.074) (0.101) Crisis –0.026** –0.015 –0.001 (0.012) (0.016) (0.012) Weighted size –0.014** –0.017** –0.011 (0.007) (0.007) (0.009) Weighted total debt 0.330*** 0.255 0.391** (0.120) (0.163) (0.196) Population 0.155 –0.002 0.825 (0.224) (0.202) (0.571) Lagged population 0.010 0.155 –0.644 (0.191) (0.135) (0.480) Income per capita 0.297*** 0.223*** 0.368*** (0.053) (0.079) (0.070) Lagged income per capita 0.147** 0.159* 0.131 (0.063) (0.092) (0.090) Lagged house price –0.574*** –0.535*** –0.624*** (0.031) (0.040) (0.050) Lagged house price growth 0.602*** 0.594*** 0.604*** (0.038) (0.052) (0.058) Median age 0.007 0.016*** 0.001 (0.009) (0.006) (0.013) Owner occupied houses –0.004 –0.001 –0.005* (0.003) (0.020) (0.003) Population with a college degree 0.002 0.007 –0.004 (0.005) (0.006) (0.010) Unemployment rate 0.026 0.048 –0.001 (0.024) (0.036) (0.045) Time fixed effects Yes Yes Yes MSA fixed effects Yes Yes Yes Adjusted R2 0.793 0.813 0.775 Observations 1450 718 732 All MSAs Well-capitalized Under-capitalized Dependent variable △House price △House price △House price Maturing debt × Crisis –0.130*** –0.120* –0.178*** (0.045) (0.064) (0.061) Maturing debt –0.037 –0.174** 0.083 (0.072) (0.074) (0.101) Crisis –0.026** –0.015 –0.001 (0.012) (0.016) (0.012) Weighted size –0.014** –0.017** –0.011 (0.007) (0.007) (0.009) Weighted total debt 0.330*** 0.255 0.391** (0.120) (0.163) (0.196) Population 0.155 –0.002 0.825 (0.224) (0.202) (0.571) Lagged population 0.010 0.155 –0.644 (0.191) (0.135) (0.480) Income per capita 0.297*** 0.223*** 0.368*** (0.053) (0.079) (0.070) Lagged income per capita 0.147** 0.159* 0.131 (0.063) (0.092) (0.090) Lagged house price –0.574*** –0.535*** –0.624*** (0.031) (0.040) (0.050) Lagged house price growth 0.602*** 0.594*** 0.604*** (0.038) (0.052) (0.058) Median age 0.007 0.016*** 0.001 (0.009) (0.006) (0.013) Owner occupied houses –0.004 –0.001 –0.005* (0.003) (0.020) (0.003) Population with a college degree 0.002 0.007 –0.004 (0.005) (0.006) (0.010) Unemployment rate 0.026 0.048 –0.001 (0.024) (0.036) (0.045) Time fixed effects Yes Yes Yes MSA fixed effects Yes Yes Yes Adjusted R2 0.793 0.813 0.775 Observations 1450 718 732 Table XII. The transmission of the negative liquidity shock to the real economy: evidence from MSA house prices The regressions in this table investigate the effect of the liquidity shock on MSA house prices. The dependent variable is the change in the natural logarithm of HPIs. The variable “Maturing debt” measures each MSA’s exposure to the negative liquidity shock and is calculated as the weighted average fraction of long-term debt that matured during the crisis, where the amount of deposits that each bank holds in an MSA is used as weights. All other variables are defined in Table I. All columns include controls for year fixed effects, MSA fixed effects and 1 year lagged MSA characteristics. Intercept, year fixed effects and MSA fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the MSA level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All MSAs Well-capitalized Under-capitalized Dependent variable △House price △House price △House price Maturing debt × Crisis –0.130*** –0.120* –0.178*** (0.045) (0.064) (0.061) Maturing debt –0.037 –0.174** 0.083 (0.072) (0.074) (0.101) Crisis –0.026** –0.015 –0.001 (0.012) (0.016) (0.012) Weighted size –0.014** –0.017** –0.011 (0.007) (0.007) (0.009) Weighted total debt 0.330*** 0.255 0.391** (0.120) (0.163) (0.196) Population 0.155 –0.002 0.825 (0.224) (0.202) (0.571) Lagged population 0.010 0.155 –0.644 (0.191) (0.135) (0.480) Income per capita 0.297*** 0.223*** 0.368*** (0.053) (0.079) (0.070) Lagged income per capita 0.147** 0.159* 0.131 (0.063) (0.092) (0.090) Lagged house price –0.574*** –0.535*** –0.624*** (0.031) (0.040) (0.050) Lagged house price growth 0.602*** 0.594*** 0.604*** (0.038) (0.052) (0.058) Median age 0.007 0.016*** 0.001 (0.009) (0.006) (0.013) Owner occupied houses –0.004 –0.001 –0.005* (0.003) (0.020) (0.003) Population with a college degree 0.002 0.007 –0.004 (0.005) (0.006) (0.010) Unemployment rate 0.026 0.048 –0.001 (0.024) (0.036) (0.045) Time fixed effects Yes Yes Yes MSA fixed effects Yes Yes Yes Adjusted R2 0.793 0.813 0.775 Observations 1450 718 732 All MSAs Well-capitalized Under-capitalized Dependent variable △House price △House price △House price Maturing debt × Crisis –0.130*** –0.120* –0.178*** (0.045) (0.064) (0.061) Maturing debt –0.037 –0.174** 0.083 (0.072) (0.074) (0.101) Crisis –0.026** –0.015 –0.001 (0.012) (0.016) (0.012) Weighted size –0.014** –0.017** –0.011 (0.007) (0.007) (0.009) Weighted total debt 0.330*** 0.255 0.391** (0.120) (0.163) (0.196) Population 0.155 –0.002 0.825 (0.224) (0.202) (0.571) Lagged population 0.010 0.155 –0.644 (0.191) (0.135) (0.480) Income per capita 0.297*** 0.223*** 0.368*** (0.053) (0.079) (0.070) Lagged income per capita 0.147** 0.159* 0.131 (0.063) (0.092) (0.090) Lagged house price –0.574*** –0.535*** –0.624*** (0.031) (0.040) (0.050) Lagged house price growth 0.602*** 0.594*** 0.604*** (0.038) (0.052) (0.058) Median age 0.007 0.016*** 0.001 (0.009) (0.006) (0.013) Owner occupied houses –0.004 –0.001 –0.005* (0.003) (0.020) (0.003) Population with a college degree 0.002 0.007 –0.004 (0.005) (0.006) (0.010) Unemployment rate 0.026 0.048 –0.001 (0.024) (0.036) (0.045) Time fixed effects Yes Yes Yes MSA fixed effects Yes Yes Yes Adjusted R2 0.793 0.813 0.775 Observations 1450 718 732 The results show that the negative liquidity shock is transmitted to the real economy by a significant reduction in house prices. A natural interpretation is that the negative liquidity shock decreased the supply of real estate loans, which in turn declined the demand for houses, and as a result house prices declined. Although the difference between MSAs with more under-capitalized banks and MSAs with more well-capitalized banks is not significant with a p-value of 0.44, house prices declined more in MSAs with more under-capitalized banks. This is consistent with my previous result that well-capitalized banks could avoid the transmission of the negative liquidity shock. 3.4 Economic magnitude It is important to evaluate whether the effect of maturing long-term debt on lending and house prices during the crisis was of a significant economic magnitude. To address this question, I estimate the hypothetical increase in total loans and house prices if all banks had a lower fraction of long-term debt that matured during the crisis by 10 percentage points. Based on the coefficient of the interaction term (–0.085) reported in Column (2) in Table IV, this would have led to an average relative increase (i.e., smaller decrease) of around $46.5 billion in total loans because total assets of all banks in my sample was around $5.47 trillion in 2006. This implies that total loans would have decreased less by around 31% in this hypothetical scenario because the average reduction in total lending during the crisis was approximately $150 billion in my sample. For house prices, the coefficient of the interaction term (–0.13) shown in Column (1) in Table XII indicates that this hypothetical scenario would have led to an average relative increase in the growth rate of house prices (i.e., smaller decrease) by around 1.3 percentage points. Given that the average drop in the growth rate of house prices during the crisis was 8.6 percentage points in my sample, this hypothetical decrease in long-term debt would have led to 15% less reduction in house prices during the crisis. These estimates indicate that 10 percentage points decrease in the fraction of long-term debt that matured during the crisis could have led to 31% less reduction in total loans and 15% less reduction in house prices, where both effects are economically significant. 4. Robustness 4.1 Falsification test The assumption of the DID methodology is that banks with a larger fraction of long-term debt that matured during the 2007–09 financial crisis were affected more severely by the negative liquidity shock because it was difficult for them to refinance their obligations through alternative financing sources in this contraction period of time. This assumption does not hold in times with easier credit opportunities. This implies that the effect of the maturing long-term debt must be insignificant if the same experiment is repeated for a period of time without a crisis. To verify whether this holds, the same experiment is replicated for the time period of 2003–06, where there was no crisis and it was easier to find alternative financing sources to refinance the maturing long-term debt. I falsely assume that 2005–06 is the crisis and 2003–04 is the pre-crisis period of time, and repeat the same empirical analysis. As reported in Table XIII, the results show that the effect of an increase in the fraction of long-term debt maturing during 2005–06 on the outstanding total long-term debt and the loan supply is indistinguishable from zero. This replication ensures that the observed change in the amount of loans is more likely due to the negative liquidity shock during the 2007–09 financial crisis. Table XIII. Robustness: falsification test assuming 2005–06 is the crisis The regressions in this table falsely assume that 2005–06 is a crisis period of time and show the effect of the fraction of long-term debt that matured during this period on banks’ outstanding long-term debt and their loan supply. Dependent variables are the outstanding long-term debt and total loans divided by 2003 total assets. Columns (1) and (2) show results for all banks and Columns (3)–(6) show results for well-capitalized and under-capitalized banks, where a bank is defined as well-capitalized if its total equity ratio was above the median in 2003, and as under-capitalized otherwise. The variable “Crisis” is the post-treatment indicator equal to one in 2005 and 2006, and zero in 2003 and 2004. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Well-capitalized banks Under-capitalized banks Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.000 –0.032 0.000 0.009 0.003 –0.055 (0.007) (0.028) (0.008) (0.036) (0.010) (0.039) Maturing debt –0.002 0.076* 0.003 0.092 –0.014 0.049 (0.008) (0.043) (0.011) (0.059) (0.009) (0.065) Crisis 0.001 0.132*** 0.006 0.262*** 0.006 0.239*** (0.006) (0.027) (0.007) (0.038) (0.012) (0.045) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.613 0.297 0.725 0.220 0.557 0.339 Observations 438 455 201 214 237 241 All banks Well-capitalized banks Under-capitalized banks Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.000 –0.032 0.000 0.009 0.003 –0.055 (0.007) (0.028) (0.008) (0.036) (0.010) (0.039) Maturing debt –0.002 0.076* 0.003 0.092 –0.014 0.049 (0.008) (0.043) (0.011) (0.059) (0.009) (0.065) Crisis 0.001 0.132*** 0.006 0.262*** 0.006 0.239*** (0.006) (0.027) (0.007) (0.038) (0.012) (0.045) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.613 0.297 0.725 0.220 0.557 0.339 Observations 438 455 201 214 237 241 Table XIII. Robustness: falsification test assuming 2005–06 is the crisis The regressions in this table falsely assume that 2005–06 is a crisis period of time and show the effect of the fraction of long-term debt that matured during this period on banks’ outstanding long-term debt and their loan supply. Dependent variables are the outstanding long-term debt and total loans divided by 2003 total assets. Columns (1) and (2) show results for all banks and Columns (3)–(6) show results for well-capitalized and under-capitalized banks, where a bank is defined as well-capitalized if its total equity ratio was above the median in 2003, and as under-capitalized otherwise. The variable “Crisis” is the post-treatment indicator equal to one in 2005 and 2006, and zero in 2003 and 2004. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Well-capitalized banks Under-capitalized banks Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.000 –0.032 0.000 0.009 0.003 –0.055 (0.007) (0.028) (0.008) (0.036) (0.010) (0.039) Maturing debt –0.002 0.076* 0.003 0.092 –0.014 0.049 (0.008) (0.043) (0.011) (0.059) (0.009) (0.065) Crisis 0.001 0.132*** 0.006 0.262*** 0.006 0.239*** (0.006) (0.027) (0.007) (0.038) (0.012) (0.045) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.613 0.297 0.725 0.220 0.557 0.339 Observations 438 455 201 214 237 241 All banks Well-capitalized banks Under-capitalized banks Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.000 –0.032 0.000 0.009 0.003 –0.055 (0.007) (0.028) (0.008) (0.036) (0.010) (0.039) Maturing debt –0.002 0.076* 0.003 0.092 –0.014 0.049 (0.008) (0.043) (0.011) (0.059) (0.009) (0.065) Crisis 0.001 0.132*** 0.006 0.262*** 0.006 0.239*** (0.006) (0.027) (0.007) (0.038) (0.012) (0.045) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.613 0.297 0.725 0.220 0.557 0.339 Observations 438 455 201 214 237 241 Table XIV. Robustness: TLGP The regressions in this table investigate the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply for TLGP banks versus non-TLGP banks using interaction terms. A bank is defined to be a TLGP bank if the bank issued TLGP debt under the DGP, and TLGP amount is the amount of debt that was issued by the TLGP bank divided by its 2006 total assets. Columns (1) and (2) show results for TLGP banks and Columns (3) and (4) show results for TLGP amounts. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.043*** –0.078** –0.042** –0.093** (0.015) (0.038) (0.016) (0.039) Maturing debt –0.016 –0.134*** –0.016 –0.131*** (0.020) (0.033) (0.022) (0.034) Crisis 0.017*** 0.024 0.017*** 0.020 (0.004) (0.020) (0.004) (0.019) Maturing debt × Crisis × TLGP 0.072 0.099 (0.093) (0.170) Maturing debt × TLGP –0.099 –0.277* (0.062) (0.167) Crisis × TLGP –0.012 –0.061 (0.024) (0.051) TLGP 0.031** 0.068** (0.012) (0.034) Maturing debt × Crisis × TLGP amount 0.020 0.060 (0.013) (0.051) Maturing debt × TLGP amount –0.027* –0.056 (0.015) (0.051) Crisis × TLGP amount –0.002 –0.013 (0.002) (0.008) TLGP amount 0.005*** 0.012* (0.002) (0.007) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.355 0.145 0.353 0.142 Observations 613 712 613 712 Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.043*** –0.078** –0.042** –0.093** (0.015) (0.038) (0.016) (0.039) Maturing debt –0.016 –0.134*** –0.016 –0.131*** (0.020) (0.033) (0.022) (0.034) Crisis 0.017*** 0.024 0.017*** 0.020 (0.004) (0.020) (0.004) (0.019) Maturing debt × Crisis × TLGP 0.072 0.099 (0.093) (0.170) Maturing debt × TLGP –0.099 –0.277* (0.062) (0.167) Crisis × TLGP –0.012 –0.061 (0.024) (0.051) TLGP 0.031** 0.068** (0.012) (0.034) Maturing debt × Crisis × TLGP amount 0.020 0.060 (0.013) (0.051) Maturing debt × TLGP amount –0.027* –0.056 (0.015) (0.051) Crisis × TLGP amount –0.002 –0.013 (0.002) (0.008) TLGP amount 0.005*** 0.012* (0.002) (0.007) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.355 0.145 0.353 0.142 Observations 613 712 613 712 Table XIV. Robustness: TLGP The regressions in this table investigate the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply for TLGP banks versus non-TLGP banks using interaction terms. A bank is defined to be a TLGP bank if the bank issued TLGP debt under the DGP, and TLGP amount is the amount of debt that was issued by the TLGP bank divided by its 2006 total assets. Columns (1) and (2) show results for TLGP banks and Columns (3) and (4) show results for TLGP amounts. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.043*** –0.078** –0.042** –0.093** (0.015) (0.038) (0.016) (0.039) Maturing debt –0.016 –0.134*** –0.016 –0.131*** (0.020) (0.033) (0.022) (0.034) Crisis 0.017*** 0.024 0.017*** 0.020 (0.004) (0.020) (0.004) (0.019) Maturing debt × Crisis × TLGP 0.072 0.099 (0.093) (0.170) Maturing debt × TLGP –0.099 –0.277* (0.062) (0.167) Crisis × TLGP –0.012 –0.061 (0.024) (0.051) TLGP 0.031** 0.068** (0.012) (0.034) Maturing debt × Crisis × TLGP amount 0.020 0.060 (0.013) (0.051) Maturing debt × TLGP amount –0.027* –0.056 (0.015) (0.051) Crisis × TLGP amount –0.002 –0.013 (0.002) (0.008) TLGP amount 0.005*** 0.012* (0.002) (0.007) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.355 0.145 0.353 0.142 Observations 613 712 613 712 Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.043*** –0.078** –0.042** –0.093** (0.015) (0.038) (0.016) (0.039) Maturing debt –0.016 –0.134*** –0.016 –0.131*** (0.020) (0.033) (0.022) (0.034) Crisis 0.017*** 0.024 0.017*** 0.020 (0.004) (0.020) (0.004) (0.019) Maturing debt × Crisis × TLGP 0.072 0.099 (0.093) (0.170) Maturing debt × TLGP –0.099 –0.277* (0.062) (0.167) Crisis × TLGP –0.012 –0.061 (0.024) (0.051) TLGP 0.031** 0.068** (0.012) (0.034) Maturing debt × Crisis × TLGP amount 0.020 0.060 (0.013) (0.051) Maturing debt × TLGP amount –0.027* –0.056 (0.015) (0.051) Crisis × TLGP amount –0.002 –0.013 (0.002) (0.008) TLGP amount 0.005*** 0.012* (0.002) (0.007) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.355 0.145 0.353 0.142 Observations 613 712 613 712 4.2 Temporary liquidity guarantee program On October 2008, the FDIC implemented the Temporary Liquidity Guarantee Program (TLGP). Under the Debt Guarantee Program (DGP), which was one of the components of the TLGP, the FDIC guaranteed in full the senior unsecured debt issued by a participating financial institution until the end of 2012. This enabled the banks and their holding companies to roll-over their short-term liabilities and issue longer-term debt during the financial crisis. In this section, I study whether the DGP has an impact on the transmission of the negative liquidity shock by enabling banks to roll over their maturing long-term debt at affordable costs. A bank is defined to be a TLGP bank if the bank issued debt under the guarantee of the DGP. Table XIV presents the results. Columns (1) and (2) show that being a TLGP bank did not have a significant effect on the reduction of banks’ outstanding long-term debt and their loan supply. I further examine whether the amount of debt that was issued under the DGP has an effect. According to the results reported in Columns (3) and (4), the reduction in the outstanding long-term debt and the loan supply is not affected by the amount of debt that each TLGP bank issued under the DGP. 5. Conclusion This article analyzes the transmission of negative liquidity shocks to the real economy through a reduction in banks’ loan supply by using the 2007–09 financial crisis as a negative liquidity shock on banks. The ex ante heterogeneity in the amount of long-term debt that matured during the crisis is used to measure bank-specific exposure to the negative liquidity shock. The banks that had high amounts of long-term debt that matured during the crisis had a hard time finding affordable financing sources to cover their shortfalls due to the increased costs and financial frictions. As a result, 1 percentage point increase in the fraction of long-term debt that matured during the crisis led to a significant decrease in banks’ outstanding long-term debt by almost 0.04% of their total assets. This is transmitted to the asset side of the banks’ balance sheets by a reduction in their loan supply by 0.09% of their total assets. This reduction is particularly strong for real estate loans as compared with consumer loans whereas there is no significant effect on C&I loans. I further analyze the transmission of this liquidity shock to the real economy through the reduction in the real estate loans by examining house prices in the MSAs where affected banks have branches. My findings show that 1 percentage point increase in the weighted average fraction of long-term debt that matured during the crisis resulted in a 0.13 percentage point decrease in the growth rate of house prices. According to my results, the liquidity shock did not have any significant effect on well-capitalized banks’ loan supply, whereas the effect is stronger for under-capitalized banks. An under-capitalized bank with 1 percentage point increase in its long-term debt ratio that matured during the crisis experienced a reduction in its loan supply by 0.12% of its total assets. This finding is consistent with the literature which shows that financial frictions prevent under-capitalized banks from finding alternative financing sources to continue financing their loans during contraction times since these banks are perceived as riskier than well-capitalized banks (see, e.g., Kashyap and Stein, 2000; Kishan and Opiela, 2000; Meh and Moran, 2010). Moreover, the effect of the liquidity shock is stronger on house prices in the MSAs with more under-capitalized banks relative to MSAs with more well-capitalized banks although the difference is not significant. Further cross-sectional results show that the effect of the negative liquidity shock is stronger for banks with lower deposits and banks with higher short-term debt holdings. The effect disappears for banks with higher deposits and banks with lower short-term debt holdings. The former result is consistent with the recent literature that argues that banks with better access to deposit financing are more robust to liquidity shocks (see, e.g., Gatev and Strahan, 2006; Ivashina and Scharfstein, 2010; Cornett et al., 2011; Dagher and Kazimov, 2015). The latter result is consistent with the recent studies that show that banks with higher short-term debt ratios faced higher roll-over risk during the last financial crisis (see, e.g., Cornett et al., 2011; Huang and Ratnovski, 2011; Dagher and Kazimov, 2015). Overall, these results suggest that bank lending establishes a transmission channel where negative liquidity shocks are transmitted from banks’ balance sheets to the real economy, and that holding higher bank capital ratios, higher deposit ratios, and lower short-term debt ratios mitigates the transmission through this channel. Footnotes 1 Almeida et al. (2011) use this identification to analyze the effect of financial contracting on firms’ investment decisions. They find that firms whose long-term debt largely matured after the third quarter of 2007 cut their investment-to-capital ratio by 2.5 percentage points more than otherwise similar firms. 2 Long-term debt is defined as the debt with at least 1 year maturity. 3 Following Berger and Bouwman (2009), Flannery and Lin (2015), and Dursun-de Neef and Schandlbauer (2017), the amount of deposits that each bank holds in an MSA is used as a weight that reflects how important that bank is for that MSA. Accordingly, to calculate an MSA’s exposure to the negative liquidity shock, a deposits-weighted average fraction of long-term debt that matured during the crisis is calculated for each MSA. 4 For the falsification test in the robustness section, I use data for the time period from 2003 to 2006. 5 This includes people above 25 years old with one of the following degrees: associate, bachelor, master, professional, and doctorate. 6 The GDP in 2006 was $13.86 trillion which means that the largest bank in my sample held total assets that were 13.6% of the US GDP. 7 By using a continuous treatment, I study the effect of a change in the exposure to the negative liquidity shock instead of defining a binary variable for treated and control groups. This is similar to Acemoglu, Autor, and Lyle (2004). 8 I use bank controls that are fixed in the year 2006 instead of 1 year lagged ones to control for possible endogeneity due to the possible effects of the crisis on these bank characteristics. 9 When I repeat the analysis using the natural logarithm of total loans rather than normalizing it by total assets, the results are very similar as shown in Table V. 10 The reduction in the long-term debt is not significant with a p-value of 12%. 11 It is calculated as the long-term debt issued before the end of 2006 with a maturity from the beginning of 2008 until the end of 2009 divided by the long-term debt issued before the end of 2006 with a maturity from the beginning of 2008 for each bank in an MSA. Afterwards, the amount of deposits that each bank holds in this MSA is used as weights to calculate the weighted average fraction of long-term debt that matured during the crisis. 12 HMDA reports information on mortgages originated by banks and independent mortgage companies, so I can collect data on the volume of mortgage loans that are originated in an MSA in a given year rather than outstanding mortgage loans that each bank holds in that MSA. References Acemoglu D. , Autor D. H. , Lyle D. ( 2004 ): Women, war, and wages: the effect of female labor supply on the wage structure at midcentury , Journal of Political Economy 112 , 497 – 551 . Google Scholar CrossRef Search ADS Adelino M. , Schoar A. , Severino F. ( 2012 ): Credit supply and house prices: evidence from mortgage market segmentation. Technical report, National Bureau of Economic Research. Almeida H. , Campello M. , Laranjeira B. , Weisbenner S. ( 2011 ): Corporate debt maturity and the real effects of the 2007 credit crisis , Critical Finance Review 1 , 3 – 58 . Google Scholar CrossRef Search ADS Berger A. N. , Bouwman C. H. ( 2009 ): Bank liquidity creation , Review of Financial Studies 22 , 3779 – 3837 . Google Scholar CrossRef Search ADS Carlson M. , Shan H. , Warusawitharana M. ( 2013 ): Capital ratios and bank lending: a matched bank approach , Journal of Financial Intermediation 22 , 663 – 687 . Google Scholar CrossRef Search ADS Case K. E. , Shiller R. J. ( 1989 ): The efficiency of the market for single-family homes , American Economic Review 79 , 125 – 137 . Chava S. , Purnanandam A. ( 2011 ): The effect of banking crisis on bank-dependent borrowers , Journal of Financial Economics 99 , 116 – 135 . Google Scholar CrossRef Search ADS Cingano F. , Manaresi F. , Sette E. ( 2016 ): Does credit crunch investment down? New evidence on the real effects of the bank-lending channel , Review of Financial Studies 29 , 2737 – 2773 . Google Scholar CrossRef Search ADS Cornett M. M. , McNutt J. J. , Strahan P. E. , Tehranian H. ( 2011 ): Liquidity risk management and credit supply in the financial crisis , Journal of Financial Economics 101 , 297 – 312 . Google Scholar CrossRef Search ADS Dagher J. , Kazimov K. ( 2015 ): Banks’ liability structure and mortgage lending during the financial crisis , Journal of Financial Economics 116 , 565 – 582 . Google Scholar CrossRef Search ADS Dell’Ariccia G. , Detragiache E. , Rajan R. ( 2008 ): The real effect of banking crises , Journal of Financial Intermediation 17 , 89 – 112 . Google Scholar CrossRef Search ADS Di Maggio M. , Kermani A. ( 2017 ): Credit-induced boom and bust , Review of Financial Studies 30 , 3711 – 3758 . Google Scholar CrossRef Search ADS Dursun-de Neef H. Ö. , Schandlbauer A. ( 2017 ): The causal relationship between bank loans and bank capital. Unpublished working paper. Favara G. , Imbs J. ( 2015 ): Credit supply and the price of housing , American Economic Review 105 , 958 – 992 . Google Scholar CrossRef Search ADS Flannery M. J. , Lin L. ( 2015 ): House prices, bank balance sheets, and bank credit supply. Unpublished working paper. Gan J. ( 2007 ): The real effects of asset market bubbles: loan-and firm-level evidence of a lending channel , Review of Financial Studies 20 , 1941 – 1973 . Google Scholar CrossRef Search ADS Gatev E. , Strahan P. E. ( 2006 ): Banks’ advantage in hedging liquidity risk: theory and evidence from the commercial paper market , Journal of Finance 61 , 867 – 892 . Google Scholar CrossRef Search ADS Himmelberg C. , Mayer C. , Sinai T. ( 2005 ): Assessing high house prices: bubbles, fundamentals and misperceptions , Journal of Economic Perspectives 19 , 67 – 92 . Google Scholar CrossRef Search ADS Huang R. , Ratnovski L. ( 2011 ): The dark side of bank wholesale funding , Journal of Financial Intermediation 20 , 248 – 263 . Google Scholar CrossRef Search ADS Ivashina V. , Scharfstein D. ( 2010 ): Bank lending during the financial crisis of 2008 , Journal of Financial Economics 97 , 319 – 338 . Google Scholar CrossRef Search ADS Iyer R. , Peydro J.-L. ( 2011 ): Interbank contagion at work: evidence from a natural experiment , Review of Financial Studies 24 , 1337 – 1377 . Google Scholar CrossRef Search ADS Iyer R. , Peydró J.-L. , da Rocha-Lopes S. , Schoar A. ( 2014 ): Interbank liquidity crunch and the firm credit crunch: evidence from the 2007–2009 crisis , Review of Financial Studies 27 , 347 – 372 . Google Scholar CrossRef Search ADS Kashyap A. K. , Stein J. C. ( 2000 ): What do a million observations on banks say about the transmission of monetary policy?, American Economic Review 90 , 407 – 428 . Google Scholar CrossRef Search ADS Khwaja A. I. , Mian A. ( 2008 ): Tracing the impact of bank liquidity shocks: evidence from an emerging market , American Economic Review 98 , 1413 – 1442 . Google Scholar CrossRef Search ADS Kishan R. P. , Opiela T. P. ( 2000 ): Bank size, bank capital, and the bank lending channel , Journal of Money, Credit and Banking 32 , 121 – 141 . Google Scholar CrossRef Search ADS Lamont O. , Stein J. C. ( 1999 ): Leverage and house-price dynamics in US cities , Journal of Economic Perspectives 30 , 498 – 514 . Lopez J. A. ( 1999 ): Using CAMELS ratings to monitor bank conditions, Federal Reserve Bank of San Francisco Economic Letter No. 19. Meh C. A. , Moran K. ( 2010 ): The role of bank capital in the propagation of shocks , Journal of Economic Dynamics and Control 34 , 555 – 576 . Google Scholar CrossRef Search ADS Mian A. , Sufi A. ( 2009 ): The consequences of mortgage credit expansion: evidence from the US mortgage default crisis , Quarterly Journal of Economics 124 , 1449 – 1496 . Google Scholar CrossRef Search ADS Ongena S. , Smith D. C. , Michalsen D. ( 2003 ): Firms and their distressed banks: lessons from the Norwegian banking crisis , Journal of Financial Economics 67 , 81 – 112 . Google Scholar CrossRef Search ADS Paravisini D. ( 2008 ): Local bank financial constraints and firm access to external finance , Journal of Finance 63 , 2161 – 2193 . Google Scholar CrossRef Search ADS Peek J. , Rosengren E. S. ( 2000 ): Collateral damage: effects of the Japanese bank crisis on real activity in the United States , American Economic Review 90 , 30 – 45 . Google Scholar CrossRef Search ADS Puri M. , Rocholl J. , Steffen S. ( 2011 ): Global retail lending in the aftermath of the US financial crisis: distinguishing between supply and demand effects , Journal of Financial Economics 100 , 556 – 578 . Google Scholar CrossRef Search ADS Roberts M. R. , Whited T. M. ( 2013 ): Endogeneity in empirical corporate finance , Handbook of the Economics of Finance 2 , 493 – 572 . Google Scholar CrossRef Search ADS Schnabl P. ( 2012 ): The international transmission of bank liquidity shocks: evidence from an emerging market , Journal of Finance 67 , 897 – 932 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the European Finance Association. All rights reserved. For Permissions, please email: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Review of Finance Oxford University Press

The Transmission of Bank Liquidity Shocks: Evidence from House Prices

Loading next page...
 
/lp/ou_press/the-transmission-of-bank-liquidity-shocks-evidence-from-house-prices-wCF6dPwCZ0
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the European Finance Association. All rights reserved. For Permissions, please email: journals.permissions@oup.com
ISSN
1572-3097
eISSN
1573-692X
D.O.I.
10.1093/rof/rfy001
Publisher site
See Article on Publisher Site

Abstract

Abstract This article uses the 2007–09 financial crisis as a negative liquidity shock on banks in the USA and analyzes its transmission to the real economy. The ex ante heterogeneity in the amount of long-term debt that matured during the crisis is used to measure the variation in banks’ exposure to the liquidity shock. I find that banks transmitted the liquidity shock to the real economy by reducing their loan supply. The reduction was particularly strong for real estate loans. As a result, house prices declined in the MSAs where these banks have branches. Bank capital plays a significant role in the transmission: under-capitalized banks transmitted the liquidity shock, whereas well-capitalized banks’ lending did not show any decline. 1. Introduction Do banks transmit liquidity shocks across markets by a reduction in their loan supply? If so, do these shocks have an impact on the real economy? The goal of this article is to answer these questions by using the 2007–09 financial crisis as a negative liquidity shock to banks in the USA. The crisis led to a systemic shock that affected many banks at the same time. To study the transmission of this shock to the real economy, I need to identify a variation in the exposure of similar banks to the liquidity shock. I follow Almeida et al. (2011) and use the heterogeneity in the amount of long-term debt that matured right after the onset of the crisis to measure the variation in banks’ exposure to the shock.1 During the crisis, financing costs for long-term debt increased sharply. For example, Citigroup’s investment-grade spreads increased from 1% in September 2007 to 3% in early 2008 (Almeida et al., 2011). The collapse of Lehman Brothers aggravated this effect further to around 7% at the end of 2008. Similarly, high-yield spreads, which had been around 3% in early 2007, approached 8% in early 2008 and reached a level of around 17% shortly after September 2008. As a result, banks with a larger amount of long-term debt that matured during the crisis are expected to be affected more heavily since these banks had to refinance their maturing debt when financing costs were high. This makes the amount of debt that matured during the crisis a good proxy to measure the individual bank’s exposure to the negative liquidity shock. It is important to use the long-term debt issued before the end of 2006 rather than using short-term debt.2 When the crisis hit, the amount of long-term debt due around that time was an exogenous shock to banks since they did not anticipate the coming crisis when they issued this debt before the end of 2006. This does not hold for their short-term debt which was issued when the crisis was anticipated. This generates an endogeneity problem since the amount of short-term debt depends on the anticipation of the crisis. Using long-term debt solves this endogeneity problem. In this way, I can identify a bank liquidity shock that is exogenous to the state of the economy. As a result, I can study the causal effect of this liquidity shock on banks’ loan supply and examine the transmission of this shock to the real economy. One might argue that the amount of long-term debt, on average 10% of total assets, is insignificant compared with other liabilities in bank balance sheets and, as a result, the effect of the long-term debt due during the crisis is expected to be insignificant. To understand the importance of the long-term debt that matured during the last financial crisis, one can look at the largest bank in my sample with total assets of $1.88 trillion at the end of 2006. This bank had $20.87 billion of long-term debt maturing during the crisis. This is a substantial amount of debt that needed to be rolled over for one bank. So the amount of long-term debt that matured during the crisis was significantly important, although the fraction of long-term debt is quite small. To study the effect of the liquidity shock on banks’ loan supply, I need to separate supply from demand. To do so, I use a difference in differences (DID) methodology where I match each treated bank that is affected by the liquidity shock with a set of unaffected control banks that are located in the same metropolitan statistical areas (MSAs). This matched bank approach that is proposed by Carlson, Shan, and Warusawitharana (2013) enables me to control for local environment for banks because banks in the same location face the same economic environment. The remaining difference in the amount of loans should then be related to differences in banks’ exposure to the liquidity shock during the crisis. To provide evidence on the transmission of the negative liquidity shock to the real economy, I proceed in the following steps: first, I show that an increase in the amount of long-term debt that matured during the crisis leads to a significant decrease in banks’ outstanding long-term debt. According to my results, a bank with 1 percentage point higher long-term debt ratio that matured during the crisis decreased its long-term debt by almost 0.04% of its total assets. This result suggests that banks had a hard time rolling over their maturing long-term debt and had to cut their long-term debt holdings. This could be attributed to the sharp increase in financing costs during the crisis. Second, I document the transmission of this shock from the liability side to the asset side of a bank’s balance sheet through a reduction in its loan supply. According to my results, a 0.01 higher fraction of long-term debt maturing during the crisis led to a significant reduction in a bank’s loan supply by almost 0.09% of its total assets. This result holds particularly strong for banks with lower deposit ratios and banks with higher short-term debt holdings. This supports the earlier findings that banks with higher deposit ratios are more robust to liquidity shocks because deposits are a more stable source of financing (see, e.g., Gatev and Strahan, 2006; Ivashina and Scharfstein, 2010; Cornett et al., 2011; Dagher and Kazimov, 2015), and that banks with higher short-term debt ratios were affected more heavily by the last financial crisis since these banks faced higher roll-over risk (see, e.g., Cornett et al., 2011; Huang and Ratnovski, 2011; Dagher and Kazimov, 2015). To understand the motivation behind the reduction in the loan supply, I study the effect on nonperforming loans, loan loss reserves, and cash holdings of banks, so that I can answer the question whether banks decreased their loan supply to reduce their portfolio risk or to preserve liquidity. I find that 10 percentage point increase in the fraction of long-term debt that matured during the crisis led to a decrease in a bank’s nonperforming loans by 0.08% of its total assets and a decrease in its loan loss reserves by 0.03% of its total assets, whereas there was no significant change in the cash holdings. This suggests that banks that were affected by the liquidity shock decreased their lending to reduce their portfolio risk. I next analyze the effect of the liquidity shock on three types of loans: real estate loans, consumer loans, and commercial and industrial (C&I) loans. The results indicate that the observed decrease in total loans was a result of a significant decrease in real estate loans and consumer loans, whereas C&I loans were not affected. One percentage point higher fraction of long-term debt that matured during the crisis led to a decrease in a bank’s real estate loans by 0.06% of its total assets and a decrease in its consumer loans by almost 0.02% of its total assets. This suggests that when banks cut their outstanding long-term debt due to the high financing costs, they adjusted their assets by a reduction in the real estate loans and consumer loans. Thus, the negative liquidity shock triggered by the last financial crisis is transmitted from the liability side to the asset side of a bank’s balance sheet. Third, I explore the transmission of this shock to the real economy through the reduction in the real estate loans by studying the effect on house prices. For each MSA, a weighted average fraction of long-term debt that matured during the crisis is calculated. The fraction of deposits that each bank has in an MSA is used as weights. This measures each MSA’s exposure to the liquidity shock through banks that have branches in this MSA.3 The results show that house prices declined more in the MSAs where strongly affected banks have more deposits. 1 percentage point increase in the weighted average fraction of long-term debt that matured during the crisis resulted in a 0.13 percentage point decrease in the growth rate of house prices. The results suggest that banks with a larger amount of long-term debt that matured during the crisis could not find affordable financing sources to cover their shortfalls. The effect of high financing costs is expected to be stronger for under-capitalized banks since these banks are perceived as riskier compared to well-capitalized banks. Financial frictions prevent these banks from finding alternative financing sources to continue financing their loans during contraction times (Kashyap and Stein, 2000; Kishan and Opiela, 2000; Meh and Moran, 2010). To study whether the transmission of the negative liquidity shock depends on banks’ capital ratios, I split the sample into under-capitalized and well-capitalized banks. The results show that only under-capitalized banks are affected by the negative liquidity shock. An under-capitalized bank with 1 percentage point increase in its long-term debt ratio that matured during the crisis decreased its total loans by 0.18% of its total assets and its real estate loans by almost 0.08% of its total assets. On the contrary, well-capitalized banks do not show any significant reduction in their loan supply. The results are consistent at the MSA level: for MSAs with more well-capitalized banks, 1 percentage point increase in the weighted average fraction of long-term debt that matured during the crisis leads to a 0.12 percentage point decrease in the growth rate of house prices, whereas the decrease is 0.18 percentage point in the growth rate of house prices for MSAs with more under-capitalized banks. This suggests that house prices in the MSAs with more well-capitalized banks are affected less relative to MSAs with more under-capitalized banks although the difference is not significant. This article contributes to several strands of literature. First, it relates to a large literature studying the real consequences of bank liquidity shocks (see, e.g., Gan, 2007; Khwaja and Mian, 2008; Paravisini, 2008; Chava and Purnanandam, 2011; Schnabl, 2012). Gan (2007) studies how a negative shock to the financial health of banks, caused by a land market collapse in Japan, affects the real economy through firms’ investment and market valuation. Khwaja and Mian (2008) study the effect of bank liquidity shocks on borrowing firms induced by the unexpected nuclear tests of Pakistan in 1998. According to their results, small firms that borrow from affected banks are significantly more likely to be in financial distress a year after the nuclear tests. Paravisini (2008) uses an exogenous allocation of government funds across banks in Argentina to study the effect on banks’ loan supply. According to his results, financial shocks to constrained banks have an amplified effect on the aggregate supply of credit. Chava and Purnanandam (2011) use the 1998 Russian default as an exogenous shock to the US banking system and provide evidence that adverse capital shocks to banks affect their borrowers’ performance negatively. Similarly, Schnabl (2012) exploits the 1998 Russian default as an exogenous shock to international banks and analyzes the transmission of this shock to Peru. Overall, this strand of literature examines whether bank liquidity shocks are transmitted to the real economy through a change in banks’ loan supply. This article contributes to this literature by first studying the effect of a negative liquidity shock on banks’ loan supply and then linking this change in the loan supply to the real economy with a focus on house prices. This article is also related to research estimating the real costs of financial crises (see, e.g., Peek and Rosengren, 2000; Ongena, Smith, and Michalsen, 2003; Dell’Ariccia, Detragiache, and Rajan, 2008; Ivashina and Scharfstein, 2010; Iyer and Peydro, 2011; Puri, Rocholl, and Steffen, 2011; Iyer et al., 2014; Cingano, Manaresi, and Sette, 2016). Peek and Rosengren (2000) use the Japanese banking crisis as an exogenous loan supply shock in the USA and show that loan supply shocks originated in Japan had real effects on the economic activity in the USA. Ongena, Smith, and Michalsen (2003) use the Norwegian banking crisis to study the impact of bank distress announcements on the stock performance of borrowers. To explore the effect of banking crises on the real activity, Dell’Ariccia, Detragiache, and Rajan (2008) study differential effects of banking crises across different sectors. By exploiting the sudden failure of a large bank in India, Iyer and Peydro (2011) inspect the effect of financial contagion through interbank linkages. Iyer et al. (2014) study the effect of the unexpected shock to the interbank markets in August 2007 on the credit supply in Portugal. Similarly, Cingano, Manaresi, and Sette (2016) study the effect of the liquidity drought in interbank markets on the credit supply and real economic activity in Italy. My article particularly adds to the literature that investigates the costs of the last financial crisis. Almeida et al. (2011) analyze the effect of financial contracting on firms’ investment decisions during the last financial crisis. They find that firms whose long-term debt largely matured after the third quarter of 2007 cut their investment-to-capital ratio significantly more compared to otherwise similar firms. Puri, Rocholl, and Steffen (2011) and Ivashina and Scharfstein (2010) examine the effects of the US financial crisis on the retail bank lending in Germany and on new loans in the USA, respectively. Puri, Rocholl, and Steffen (2011) find that the US financial crisis induced a contraction in the supply of retail lending in Germany. Ivashina and Scharfstein (2010) show that new loans issued to large borrowers fell significantly during the last financial crisis. This article contributes to this literature by using the last financial crisis as a negative liquidity shock on the banking system and examining the transmission of this shock to the real economy through a reduction in banks’ loan supply. Finally, my findings on house prices are related to the literature that studies the effect of bank loan supply on house prices. Mian and Sufi (2009) show that the increase in securitization in the early 2000 s led to an expansion in subprime mortgage credit. This expansion had contributed significantly to the increasing house prices from 2002 to 2005. Adelino, Schoar, and Severino (2012) find that easier access to credit, made possible by changes in the conforming loan limit, significantly increased house prices. Di Maggio and Kermani (2017) use the federal preemption of national banks from local laws against predatory lending in 2004 to show that the supply of credit has significant consequences on house prices. Favara and Imbs (2015) show the significant effects of credit expansion on house prices by exploiting the US branching deregulations during 1994 and 2005 as an exogenous expansion in mortgage credit. This article adds to this literature by examining the effect of negative liquidity shocks on house prices through a reduction in the real estate loans during the last financial crisis. The rest of the article is organized as follows. Section 2 introduces the data. Section 3 presents the empirical analysis and main results. In Section 4, I introduce robustness tests and additional controls. Section 5 concludes. 2. Data In the first part of my empirical analysis, I use data on the bank balance sheet, income statement, bank capital structure, and branch deposits for US bank holding companies (BHC). The data are collected from SNL Financial. The balance sheet and income statement on BHCs are taken from the Consolidated Financial Statements of Bank Holding Companies (FR Y-9 C) for the time period from 2006 to 2009.4 The bank branch data come from the Summary of Deposits filings of the Federal Deposit Insurance Corporation (FDIC). In the second part of my empirical analysis, I use data on population, income per capita, house price index (HPI), median age in the population, owner occupied housing, population with different levels of education, and unemployment rate at the MSA level. The data on population and income per capita are collected by the Bureau of Economic Analysis. The HPI data come from the Federal Housing Finance Agency (FHFA). The data on median age, owner occupied housing, levels of education, and unemployment rate are taken from the U.S. Census Bureau and collected through SNL Financial. To eliminate outliers, I exclude bank-years with negative values of total equity ratios. I focus on deposit taking institutions that have positive deposit holdings and positive loans outstanding. The final unbalanced sample consists of 913 bank-year observations and 242 banks. To measure a bank’s financial condition, several bank characteristics that proxy for the components of the CAMELS rating are used. The CAMELS rating is a supervisory rating used by federal banking regulators to provide a convenient summary of bank conditions. The acronym CAMELS refers to the six components of a bank’s condition: capital adequacy, asset quality, management, earnings, liquidity, and sensitivity to market risk. Banking regulators use this rating to provide a comprehensive assessment of a bank’s overall condition (Lopez, 1999). In this article, I use total equity to proxy capital adequacy, loan loss reserves for asset quality, net interest income for management, return on assets (ROA) for earnings, cash for liquidity, and total deposits for the sensitivity to the market risk. These six bank characteristics are included as controls in each regression to control for differences. In addition, I control for the size of the banks and their long-term debt ratios. Table I reports summary statistics for the variables used in the analysis. The average bank in my sample has $27.3 billion in total assets, and the standard variation is $178.2 billion. The smallest bank is Community Capital Bancshares with $0.19 billion in total assets, whereas the largest bank in the sample is Citigroup with total assets of $2.19 trillion. In my sample, the largest asset class is total loans which constitutes on average 71.7% of total assets. On average, 53% of the total loans are real estate loans, followed by C&I loans that constitute almost 21%. The third largest type of loans is the consumer loans that represent 13% of the total loans. On the liability side, a bank holds total equity that is on average 8.9% of its total assets. Total deposits are the largest liability class which is 74.5% of total assets. The total debt other than deposits (non-depository debt) is 14.7% of the total assets. 68.1% of this debt is long-term debt with at least 1 year maturity. The remaining 31.9% is short-term debt with a maturity shorter than 1 year. Table I. Definitions and summary statistics for variables Variable Definition Mean Standard deviation Panel A: Bank characteristics Total equity Total equity divided by total assets to proxy Capital adequacy (C) 0.089 0.023 Loan loss reserves Loan loss reserves divided by total loans to proxy Asset quality (A) 0.011 0.005 Net interest income Net interest income divided by total assets to proxy Management quality (M) 0.032 0.007 ROA Net income divided by total assets to proxy Earnings (E) 0.003 0.015 Cash Cash and noninterest-bearing deposits divided by total assets to proxy Liquidity (L) 0.055 0.052 Deposits Total deposits divided by total assets to proxy Sensitivity to market risk (S) 0.745 0.092 Total assets (billions) Total assets 27.338 178.200 Total loans Total loans divided by total assets 0.717 0.132 Real-estate loans Real-estate loans divided by total loans 0.530 0.147 C&I loans Commercial and industrial loans divided by total loans 0.209 0.122 Consumer loans Consumer loans divided by total loans 0.130 0.096 Total debt Total debt other than deposits 0.147 0.080 Short-term debt Total debt with a maturity of 1 year or less divided by total debt 0.319 0.234 Long-term debt Total debt minus short-term debt divided by total debt 0.681 0.234 Debt Total long-term debt issued from the beginning of 2000 until the end of 2006 divided by total assets 0.036 0.021 Maturing debt The fraction of long-term debt issued from the beginning of 2000 until the end of 2006 that matured during 2008–09 0.348 0.336 Bank-year observations 913 Panel B: MSA characteristics HPI HPI at the MSA level 185.834 36.779 Population (million) Population at the MSA level 0.476 0.694 Income per capita (thousand) Income divided by total population 35.075 6.502 Median age Population median age at the MSA level 36.333 3.969 Owner occupied houses Owner occupied housing units divided by total housing units 62.187 6.197 Population with a college degree Population aged 25+ with at least a college degree divided by total population aged 25+ 25.173 8.120 Unemployment rate Unemployment rate at the MSA level 6.116 2.214 MSA-year observations 1480 Variable Definition Mean Standard deviation Panel A: Bank characteristics Total equity Total equity divided by total assets to proxy Capital adequacy (C) 0.089 0.023 Loan loss reserves Loan loss reserves divided by total loans to proxy Asset quality (A) 0.011 0.005 Net interest income Net interest income divided by total assets to proxy Management quality (M) 0.032 0.007 ROA Net income divided by total assets to proxy Earnings (E) 0.003 0.015 Cash Cash and noninterest-bearing deposits divided by total assets to proxy Liquidity (L) 0.055 0.052 Deposits Total deposits divided by total assets to proxy Sensitivity to market risk (S) 0.745 0.092 Total assets (billions) Total assets 27.338 178.200 Total loans Total loans divided by total assets 0.717 0.132 Real-estate loans Real-estate loans divided by total loans 0.530 0.147 C&I loans Commercial and industrial loans divided by total loans 0.209 0.122 Consumer loans Consumer loans divided by total loans 0.130 0.096 Total debt Total debt other than deposits 0.147 0.080 Short-term debt Total debt with a maturity of 1 year or less divided by total debt 0.319 0.234 Long-term debt Total debt minus short-term debt divided by total debt 0.681 0.234 Debt Total long-term debt issued from the beginning of 2000 until the end of 2006 divided by total assets 0.036 0.021 Maturing debt The fraction of long-term debt issued from the beginning of 2000 until the end of 2006 that matured during 2008–09 0.348 0.336 Bank-year observations 913 Panel B: MSA characteristics HPI HPI at the MSA level 185.834 36.779 Population (million) Population at the MSA level 0.476 0.694 Income per capita (thousand) Income divided by total population 35.075 6.502 Median age Population median age at the MSA level 36.333 3.969 Owner occupied houses Owner occupied housing units divided by total housing units 62.187 6.197 Population with a college degree Population aged 25+ with at least a college degree divided by total population aged 25+ 25.173 8.120 Unemployment rate Unemployment rate at the MSA level 6.116 2.214 MSA-year observations 1480 Table I. Definitions and summary statistics for variables Variable Definition Mean Standard deviation Panel A: Bank characteristics Total equity Total equity divided by total assets to proxy Capital adequacy (C) 0.089 0.023 Loan loss reserves Loan loss reserves divided by total loans to proxy Asset quality (A) 0.011 0.005 Net interest income Net interest income divided by total assets to proxy Management quality (M) 0.032 0.007 ROA Net income divided by total assets to proxy Earnings (E) 0.003 0.015 Cash Cash and noninterest-bearing deposits divided by total assets to proxy Liquidity (L) 0.055 0.052 Deposits Total deposits divided by total assets to proxy Sensitivity to market risk (S) 0.745 0.092 Total assets (billions) Total assets 27.338 178.200 Total loans Total loans divided by total assets 0.717 0.132 Real-estate loans Real-estate loans divided by total loans 0.530 0.147 C&I loans Commercial and industrial loans divided by total loans 0.209 0.122 Consumer loans Consumer loans divided by total loans 0.130 0.096 Total debt Total debt other than deposits 0.147 0.080 Short-term debt Total debt with a maturity of 1 year or less divided by total debt 0.319 0.234 Long-term debt Total debt minus short-term debt divided by total debt 0.681 0.234 Debt Total long-term debt issued from the beginning of 2000 until the end of 2006 divided by total assets 0.036 0.021 Maturing debt The fraction of long-term debt issued from the beginning of 2000 until the end of 2006 that matured during 2008–09 0.348 0.336 Bank-year observations 913 Panel B: MSA characteristics HPI HPI at the MSA level 185.834 36.779 Population (million) Population at the MSA level 0.476 0.694 Income per capita (thousand) Income divided by total population 35.075 6.502 Median age Population median age at the MSA level 36.333 3.969 Owner occupied houses Owner occupied housing units divided by total housing units 62.187 6.197 Population with a college degree Population aged 25+ with at least a college degree divided by total population aged 25+ 25.173 8.120 Unemployment rate Unemployment rate at the MSA level 6.116 2.214 MSA-year observations 1480 Variable Definition Mean Standard deviation Panel A: Bank characteristics Total equity Total equity divided by total assets to proxy Capital adequacy (C) 0.089 0.023 Loan loss reserves Loan loss reserves divided by total loans to proxy Asset quality (A) 0.011 0.005 Net interest income Net interest income divided by total assets to proxy Management quality (M) 0.032 0.007 ROA Net income divided by total assets to proxy Earnings (E) 0.003 0.015 Cash Cash and noninterest-bearing deposits divided by total assets to proxy Liquidity (L) 0.055 0.052 Deposits Total deposits divided by total assets to proxy Sensitivity to market risk (S) 0.745 0.092 Total assets (billions) Total assets 27.338 178.200 Total loans Total loans divided by total assets 0.717 0.132 Real-estate loans Real-estate loans divided by total loans 0.530 0.147 C&I loans Commercial and industrial loans divided by total loans 0.209 0.122 Consumer loans Consumer loans divided by total loans 0.130 0.096 Total debt Total debt other than deposits 0.147 0.080 Short-term debt Total debt with a maturity of 1 year or less divided by total debt 0.319 0.234 Long-term debt Total debt minus short-term debt divided by total debt 0.681 0.234 Debt Total long-term debt issued from the beginning of 2000 until the end of 2006 divided by total assets 0.036 0.021 Maturing debt The fraction of long-term debt issued from the beginning of 2000 until the end of 2006 that matured during 2008–09 0.348 0.336 Bank-year observations 913 Panel B: MSA characteristics HPI HPI at the MSA level 185.834 36.779 Population (million) Population at the MSA level 0.476 0.694 Income per capita (thousand) Income divided by total population 35.075 6.502 Median age Population median age at the MSA level 36.333 3.969 Owner occupied houses Owner occupied housing units divided by total housing units 62.187 6.197 Population with a college degree Population aged 25+ with at least a college degree divided by total population aged 25+ 25.173 8.120 Unemployment rate Unemployment rate at the MSA level 6.116 2.214 MSA-year observations 1480 To examine whether any specific pre-crisis bank characteristics are correlated with the amount of maturing long-term debt during the crisis, I examine the effect of an increase in the maturing long-term debt on pre-crisis bank characteristics that proxy CAMELS rating. According to the results presented in Table II, the banks that had more maturing long-term debt during the crisis held significantly higher net interest income. More precisely, a bank with a 1 percentage point higher fraction of long-term debt that matured during the crisis had significantly more net interest income by 0.13 of a percentage point. Table II. The relationship between maturing long-term debt and pre-crisis bank characteristics The regressions in this table investigate the effect of an increase in the long-term debt maturing during the crisis on pre-crisis bank characteristics. Dependent variables are the pre-crisis bank characteristics, measured in 2006 and 2007, that proxy CAMELS supervisory rating. The variable “Maturing debt” is the fraction of long-term debt that matured during the crisis. All other variables are defined in Table I. All columns include controls for year fixed effects and 1 year lagged bank characteristics. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Dependent variable Total equity Loan loss reserves Net interest income ROA Cash Deposits Maturing debt 0.070 –0.002 0.131** 0.132 1.997 1.682 (0.170) (0.036) (0.065) (0.139) (1.517) (1.241) Total equity 0.927*** 0.000 0.022** –0.047 0.105 0.089 (0.024) (0.004) (0.009) (0.042) (0.073) (0.092) Loan loss reserves 0.001 0.903*** 0.036 –0.114 –2.524** 0.213 (0.202) (0.037) (0.063) (0.081) (1.036) (0.725) Net interest income 0.055 0.031* 0.930*** 0.026 0.145 –0.346 (0.098) (0.017) (0.035) (0.066) (0.917) (0.409) ROA –0.102 –0.018 –0.081* 0.793*** –0.893 –0.581 (0.139) (0.026) (0.042) (0.118) (0.788) (0.577) Cash –0.008 –0.004*** 0.001 0.004 0.866*** 0.076 (0.009) (0.002) (0.003) (0.005) (0.076) (0.047) Deposits 0.003 –0.002 0.002 0.006 0.008 0.933*** (0.006) (0.001) (0.002) (0.004) (0.042) (0.039) Size –0.004 0.005 –0.003 0.017 0.361** –0.265 (0.034) (0.007) (0.011) (0.019) (0.145) (0.161) Debt –0.048* 0.010* –0.001 –0.020 –0.007 –0.309* (0.028) (0.006) (0.010) (0.017) (0.120) (0.184) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.873 0.779 0.862 0.398 0.672 0.869 Observations 394 394 394 394 394 394 Dependent variable Total equity Loan loss reserves Net interest income ROA Cash Deposits Maturing debt 0.070 –0.002 0.131** 0.132 1.997 1.682 (0.170) (0.036) (0.065) (0.139) (1.517) (1.241) Total equity 0.927*** 0.000 0.022** –0.047 0.105 0.089 (0.024) (0.004) (0.009) (0.042) (0.073) (0.092) Loan loss reserves 0.001 0.903*** 0.036 –0.114 –2.524** 0.213 (0.202) (0.037) (0.063) (0.081) (1.036) (0.725) Net interest income 0.055 0.031* 0.930*** 0.026 0.145 –0.346 (0.098) (0.017) (0.035) (0.066) (0.917) (0.409) ROA –0.102 –0.018 –0.081* 0.793*** –0.893 –0.581 (0.139) (0.026) (0.042) (0.118) (0.788) (0.577) Cash –0.008 –0.004*** 0.001 0.004 0.866*** 0.076 (0.009) (0.002) (0.003) (0.005) (0.076) (0.047) Deposits 0.003 –0.002 0.002 0.006 0.008 0.933*** (0.006) (0.001) (0.002) (0.004) (0.042) (0.039) Size –0.004 0.005 –0.003 0.017 0.361** –0.265 (0.034) (0.007) (0.011) (0.019) (0.145) (0.161) Debt –0.048* 0.010* –0.001 –0.020 –0.007 –0.309* (0.028) (0.006) (0.010) (0.017) (0.120) (0.184) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.873 0.779 0.862 0.398 0.672 0.869 Observations 394 394 394 394 394 394 Table II. The relationship between maturing long-term debt and pre-crisis bank characteristics The regressions in this table investigate the effect of an increase in the long-term debt maturing during the crisis on pre-crisis bank characteristics. Dependent variables are the pre-crisis bank characteristics, measured in 2006 and 2007, that proxy CAMELS supervisory rating. The variable “Maturing debt” is the fraction of long-term debt that matured during the crisis. All other variables are defined in Table I. All columns include controls for year fixed effects and 1 year lagged bank characteristics. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Dependent variable Total equity Loan loss reserves Net interest income ROA Cash Deposits Maturing debt 0.070 –0.002 0.131** 0.132 1.997 1.682 (0.170) (0.036) (0.065) (0.139) (1.517) (1.241) Total equity 0.927*** 0.000 0.022** –0.047 0.105 0.089 (0.024) (0.004) (0.009) (0.042) (0.073) (0.092) Loan loss reserves 0.001 0.903*** 0.036 –0.114 –2.524** 0.213 (0.202) (0.037) (0.063) (0.081) (1.036) (0.725) Net interest income 0.055 0.031* 0.930*** 0.026 0.145 –0.346 (0.098) (0.017) (0.035) (0.066) (0.917) (0.409) ROA –0.102 –0.018 –0.081* 0.793*** –0.893 –0.581 (0.139) (0.026) (0.042) (0.118) (0.788) (0.577) Cash –0.008 –0.004*** 0.001 0.004 0.866*** 0.076 (0.009) (0.002) (0.003) (0.005) (0.076) (0.047) Deposits 0.003 –0.002 0.002 0.006 0.008 0.933*** (0.006) (0.001) (0.002) (0.004) (0.042) (0.039) Size –0.004 0.005 –0.003 0.017 0.361** –0.265 (0.034) (0.007) (0.011) (0.019) (0.145) (0.161) Debt –0.048* 0.010* –0.001 –0.020 –0.007 –0.309* (0.028) (0.006) (0.010) (0.017) (0.120) (0.184) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.873 0.779 0.862 0.398 0.672 0.869 Observations 394 394 394 394 394 394 Dependent variable Total equity Loan loss reserves Net interest income ROA Cash Deposits Maturing debt 0.070 –0.002 0.131** 0.132 1.997 1.682 (0.170) (0.036) (0.065) (0.139) (1.517) (1.241) Total equity 0.927*** 0.000 0.022** –0.047 0.105 0.089 (0.024) (0.004) (0.009) (0.042) (0.073) (0.092) Loan loss reserves 0.001 0.903*** 0.036 –0.114 –2.524** 0.213 (0.202) (0.037) (0.063) (0.081) (1.036) (0.725) Net interest income 0.055 0.031* 0.930*** 0.026 0.145 –0.346 (0.098) (0.017) (0.035) (0.066) (0.917) (0.409) ROA –0.102 –0.018 –0.081* 0.793*** –0.893 –0.581 (0.139) (0.026) (0.042) (0.118) (0.788) (0.577) Cash –0.008 –0.004*** 0.001 0.004 0.866*** 0.076 (0.009) (0.002) (0.003) (0.005) (0.076) (0.047) Deposits 0.003 –0.002 0.002 0.006 0.008 0.933*** (0.006) (0.001) (0.002) (0.004) (0.042) (0.039) Size –0.004 0.005 –0.003 0.017 0.361** –0.265 (0.034) (0.007) (0.011) (0.019) (0.145) (0.161) Debt –0.048* 0.010* –0.001 –0.020 –0.007 –0.309* (0.028) (0.006) (0.010) (0.017) (0.120) (0.184) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.873 0.779 0.862 0.398 0.672 0.869 Observations 394 394 394 394 394 394 Summary statistics for the variables at the MSA level are reported in Panel B in Table I. The mean HPI is 185.834 with Naples–Immokalee–Marco Island (FL) having the maximum HPI at 363.33 and Merced (CA) having the minimum HPI at 117.68. In my sample, the average MSA has 0.476 million people. The highest population belongs to New York–Newark–Jersey City (NY–NJ–PA) with 19.60 million people. The lowest belongs to Carson City (NV) with 55.29 thousand people. The average income per capita is $35.08 thousand with a minimum value of $18.73 thousand in McAllen–Edinburg–Mission (TX) and a maximum value of $88.81 in Bridgeport–Stamford–Norwalk (CT). The average median age in my sample is 36.33. Provo–Orem (UT) has the lowest median age with 23.4, whereas Punta Gorda (FL) has the highest median age with 54.4. The mean ratio of owner occupied housing units is 62.19% where the minimum ratio, 34.37%, belongs to Ocean City (NJ) and the maximum ratio, 77.09%, belongs to Monroe (MI). On average, people with at least a college degree at an associate level5 constitute 25.17% of total population. Dalton (GA) has the lowest fraction of people with a college degree at 14.64% and Boulder (CO) has the highest fraction at 58.70%. The mean unemployment rate is 6.116% where the highest unemployment rate, 13.6%, is in Hanford–Corcoran (CA) and the lowest unemployment rate, 2.6%, is in Sheboygan (WI). 2.1 Long-term debt due during the financial crisis of 2007–09 The most important variable of my analysis is the amount of long-term debt that matured during the 2007–09 financial crisis. The years 2008 and 2009 are used as the crisis following the National Bureau of Economic Research (NBER). The variation in banks’ exposure to the liquidity shock is measured by the ex ante heterogeneity in the fraction of their long-term debt that matured during the crisis. As mentioned above, the main advantage of using the fraction of long-term debt due during the crisis to identify an exogenous variation in banks’ exposure to the negative liquidity shock is that it mitigates endogeneity concerns. However, it also requires assumptions that can lead to a potential measurement error in this exposure. I collect the information on the issuance and the due date of the long-term debt for each bank from SNL Financial capital structure. I define the amount of long-term debt that matured during the crisis as the debt issued before the end of 2006 and due between the end of 2007 and 2009. There is one limitation with these data: the database SNL Financial has the data on capital structure from the end of 1999 onwards. By using these data, I neglect the long-term debt issued before the end of 1999 and due during the crisis. To understand how large this part of the debt is, I calculate the amount of long-term debt that was outstanding at the end of 2006 which was issued from the end of 1999 until the end of 2006. When I compare this amount with the total amount of long-term debt outstanding at the end of 2006, I find that approximately 50% of the total long-term debt was issued from the end of 1999 onwards given that these banks have debt due during the crisis. This means that half of the long-term debt was issued before the end of 1999. This might lead to a potential measurement error in this identification since the fraction of long-term debt that matured during the crisis which was issued after the end of 1999 does not take into account a bank that had a large amount of long-term debt that matured during the crisis where most of the debt was issued before the end of 1999. This introduces noise in the measurement of a bank’s exposure to the negative liquidity shock, so the estimation of the coefficient for this measure is expected to be biased downward unless there is a positive correlation between the long-term debt that was issued before the end of 1999 and the long-term debt that was issued after the end of 1999. In that case, positive correlation might lead to an upward bias in the coefficient estimate. The second concern related to this identification strategy is the importance of long-term debt relative to other types of liabilities in a bank’s balance sheet. In my sample, a bank’s total equity is on average 9% of its total assets. So the largest fraction, 91%, of its assets is financed by other types of liabilities. Out of these liabilities, deposits constitute 75% of the total assets. Other non-depository debt is 15% of the total assets. On average, 68% of the non-depository debt has a maturity of at least 1 year, defined as long-term debt, and the rest is short-term debt. One might argue that the amount of long-term debt is insignificant compared with other liabilities. So the effect of the long-term debt due during the crisis would also be insignificant. To understand the importance of the long-term debt that matured during the last financial crisis, one can think of the average bank in my sample with total assets of $27.3 billion. If we assume that the average bank in my sample has an average fraction of long-term debt that matured during the crisis (approximately 1.25%), the amount of this debt would be $341.3 million. This is an important amount of debt that matured during the crisis for one bank. If we look at the largest bank in my sample, the amount of long-term debt maturing during the crisis was $20.87 billion and this bank had $1.88 trillion total assets at the end of 2006.6 Although the fraction of long-term debt is quite small, the amount of long-term debt that matured during the crisis was still significantly important. 2.2 Matching banks to control for loan demand One of the key issues in determining the effects of a negative liquidity shock on banks’ loan supply is controlling for changes in the loan demand. The reason behind this concern is that the same economic conditions that lead to negative liquidity shocks can have a direct effect on the demand for loans. Employing a DID methodology helps to control for the systemic shock on loan demand during the crisis that affects all banks in my sample and for the possible differences in loan demand between banks with different exposure to the liquidity shock independent of the year. However, the effect of a negative liquidity shock on loan demand might differ geographically. To control for the differences in the effect of the last financial crisis on local loan demand, I use a matched bank approach that is proposed by Carlson, Shan, and Warusawitharana (2013) where each treated bank affected by the liquidity shock is matched with a set of control banks that are located in the same MSAs. Treated banks are defined as affected banks that had a positive amount of long-term debt matured during the crisis. Control banks are defined as unaffected neighboring banks that are located in the same MSAs with treated banks for each year. The reason behind this approach is that banks located in the same neighborhood face the same economic environment, so any change in the loan demand for the affected bank is expected to be the same for the matched unaffected bank. One important assumption in the DID estimation method is that the average change in the amount of loans would have been the same for both the treated and control banks in the absence of the negative liquidity shock. Table III compares the changes in the amount of loans over the period 2003–07 prior to the financial crisis. The Wilcoxon rank-sum test results show no statistical difference in the changes in the amount of loans for treated banks relative to control banks. This supports the identifying assumption of the empirical strategy. Figure 1 further presents the plots of the amount of loans for treated versus control banks before and after the crisis. The graph shows parallel trends in the amount of loans between treated and control banks before the beginning of the financial crisis. One can see the apparent difference in the trends between these groups after the crisis which provides further motivation for the DID estimation method. Table III. Trend comparison for treated and control banks before the crisis This table compares changes in total loans for treated and control banks over the period 2004–07 prior to the crisis. Treated banks are the banks with a positive amount of long-term debt that matured during the crisis. Control banks are defined as unaffected neighboring banks that are located in the same MSAs with treated banks. Change in total loans is calculated as the difference in total loans for 2 years divided by total loans in the previous year. The p-values of the Wilcoxon rank-sum test are presented in the rightmost column. Time period Treated Control Wilcoxon rank-sum test 2006–07 0.124 0.140 0.304 2005–06 0.168 0.183 0.674 2004–05 0.154 0.217 0.150 2003–04 0.182 0.260 0.500 Time period Treated Control Wilcoxon rank-sum test 2006–07 0.124 0.140 0.304 2005–06 0.168 0.183 0.674 2004–05 0.154 0.217 0.150 2003–04 0.182 0.260 0.500 Table III. Trend comparison for treated and control banks before the crisis This table compares changes in total loans for treated and control banks over the period 2004–07 prior to the crisis. Treated banks are the banks with a positive amount of long-term debt that matured during the crisis. Control banks are defined as unaffected neighboring banks that are located in the same MSAs with treated banks. Change in total loans is calculated as the difference in total loans for 2 years divided by total loans in the previous year. The p-values of the Wilcoxon rank-sum test are presented in the rightmost column. Time period Treated Control Wilcoxon rank-sum test 2006–07 0.124 0.140 0.304 2005–06 0.168 0.183 0.674 2004–05 0.154 0.217 0.150 2003–04 0.182 0.260 0.500 Time period Treated Control Wilcoxon rank-sum test 2006–07 0.124 0.140 0.304 2005–06 0.168 0.183 0.674 2004–05 0.154 0.217 0.150 2003–04 0.182 0.260 0.500 Figure 1. View largeDownload slide Trends in total loans before and after the crisis. This figure shows the trends in total loans before and after the crisis for treated and control banks. Treated banks are the banks with a positive amount of long-term debt that matured during the crisis. Control banks are defined as unaffected neighboring banks that are located in the same MSAs with treated banks. Total loans is calculated as the amount of total loans divided by 2005 total assets. Figure 1. View largeDownload slide Trends in total loans before and after the crisis. This figure shows the trends in total loans before and after the crisis for treated and control banks. Treated banks are the banks with a positive amount of long-term debt that matured during the crisis. Control banks are defined as unaffected neighboring banks that are located in the same MSAs with treated banks. Total loans is calculated as the amount of total loans divided by 2005 total assets. 3. The Transmission of Negative Liquidity Shocks In this section, I explore the transmission of negative liquidity shocks to the real economy through a reduction in banks’ loan supply in three steps: first, I show the effect of an increase in the fraction of long-term debt that matured during the crisis on the banks’ outstanding long-term debt. I expect that banks with more long-term debt that matured during the crisis would experience a larger reduction in their outstanding long-term debt since the cost of issuing long-term debt increased significantly in this time period. Second, I study whether the negative liquidity shock affects banks’ loan supply. I expect that banks with a higher exposure to the liquidity shock would reduce their lending more. Third, I analyze the impact of the negative liquidity shock on the real economy by studying the effect on house prices at the MSA level. I expect that house prices would decline more strongly in the MSAs where banks that are affected more by the negative liquidity shock have branches. 3.1 The effect on the outstanding long-term debt This section studies the effect of the negative liquidity shock on banks’ outstanding long-term debt. To study the effect of the shock, I use a DID estimation method with a continuous treatment variable, Di, and estimate LTDit=α0+α1Di×Pt+α2Di+α3Pt+θXi+δt+uit, (1) where LTDit is the outstanding long-term debt for bank i in time period t divided by total assets calculated in 2006.7 The time period captures 4 years: 2006, 2007, 2008, and 2009. Pt is the posttreatment indicator that is equal to one in 2008 and 2009, and zero in 2006 and 2007. Including the level Pt controls for trends common to all banks independent of their exposure to the liquidity shock. For example, if the debt holdings of banks are decreasing during the crisis due to uncertainty in the markets, α3 will capture this variation. Di is the continuous treatment variable and used to measure each treated bank’s exposure to the negative liquidity shock. It is the fraction of long-term debt issued before the end of 2006 and matured during the crisis, and calculated as the long-term debt issued before the end of 2006 with a maturity from the beginning of 2008 until the end of 2009 divided by the long-term debt issued before the end of 2006 with a maturity from the beginning of 2008. The level Di controls for permanent differences between banks with different exposure to the liquidity shock independent of the year. For example, if banks with more exposure to the liquidity shock have a larger amount of long-term debt than banks with less exposure, then α2 should capture this variation. The coefficient of interest is α1, which corresponds to the interaction term between Di and Pt. This variable measures whether banks with higher exposure to the negative liquidity shock experienced a greater reduction in their outstanding long-term debt from the pre-crisis years 2006 and 2007 to the crisis years 2008 and 2009 (Roberts and Whited, 2013). Xi includes the 2006 values of the size of the banks, the long-term debt issued before the end of 2006, and a set of bank characteristics that proxy the CAMELS supervisory rating, for example, total equity, loan loss reserves, net interest income, ROA, cash, and deposits, calculated in 2006.8 δt is the year fixed effects. Table IV presents the results. Column (1) shows that a bank with 1 percentage point increase in its long-term debt ratio that matured during the crisis decreased its long-term debt by almost 0.04% of its total assets. To study the effect of bank capital ratios on the transmission of the liquidity shock to the real economy, I split the sample into under-capitalized and well-capitalized banks. A bank is defined as under-capitalized if its total equity ratio was below the median in 2006 and as well-capitalized otherwise. I expect that the negative liquidity shock affects an under-capitalized bank significantly more relative to a well-capitalized bank. Columns (3) and (5) present the results for well-capitalized and under-capitalized banks, respectively. Column (3) shows that a well-capitalized bank with a higher long-term debt ratio that matured during the crisis does not experience a significant change in its outstanding long-term debt. On the contrary, the decrease in the outstanding long-term debt is stronger for under-capitalized banks according to Column (5). An under-capitalized bank with 1 percentage point higher fraction of long-term debt that matured during the crisis decreased its long-term debt by 0.05% of its total assets. When I repeat the analysis using the natural logarithm of long-term debt rather than normalizing it by total assets, the results are very similar as shown in Table V. Table IV. The effect of the negative liquidity shock on banks’ outstanding long-term debt and their loan supply The regressions in this table examine the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. Columns (1) and (2) show results for all banks and Columns (3)–(6) show results for well-capitalized and under-capitalized banks. A bank is defined as well-capitalized if its total equity ratio was above the median in 2006, and as under-capitalized otherwise. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.037** –0.085** –0.011 –0.035 –0.053** –0.121** (0.016) (0.038) (0.027) (0.044) (0.023) (0.054) Maturing debt –0.022 –0.140*** –0.008 –0.079* –0.017 –0.180** (0.020) (0.034) (0.029) (0.045) (0.032) (0.068) Crisis 0.017*** 0.019 0.043*** 0.069** 0.043*** 0.016 (0.004) (0.019) (0.006) (0.031) (0.007) (0.025) Total equity –0.003** –0.009* –0.005** –0.016*** –0.017* –0.014 (0.001) (0.005) (0.002) (0.006) (0.009) (0.015) Loan loss reserves –0.021 0.015 –0.011 –0.037 –0.021 0.113 (0.021) (0.045) (0.021) (0.067) (0.029) (0.073) Net interest income –0.009 0.040* 0.012 0.069** –0.023 0.014 (0.010) (0.024) (0.010) (0.028) (0.015) (0.053) ROA 0.008 0.021 –0.016 –0.032 0.019 0.073 (0.013) (0.036) (0.017) (0.057) (0.022) (0.051) Cash –0.003** –0.005*** –0.007*** –0.011*** –0.003** –0.004 (0.001) (0.002) (0.002) (0.004) (0.001) (0.003) Deposits –0.004*** 0.004** –0.005*** –0.001 –0.003*** 0.006** (0.001) (0.001) (0.001) (0.002) (0.001) (0.003) Size –0.007* –0.001 –0.003 0.003 –0.010 –0.005 (0.004) (0.009) (0.004) (0.013) (0.007) (0.016) Debt 0.009*** 0.011 0.014*** 0.009 0.003 0.008 (0.003) (0.008) (0.004) (0.009) (0.004) (0.012) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R-squared 0.351 0.145 0.442 0.073 0.358 0.242 Observations 613 712 319 372 294 340 All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.037** –0.085** –0.011 –0.035 –0.053** –0.121** (0.016) (0.038) (0.027) (0.044) (0.023) (0.054) Maturing debt –0.022 –0.140*** –0.008 –0.079* –0.017 –0.180** (0.020) (0.034) (0.029) (0.045) (0.032) (0.068) Crisis 0.017*** 0.019 0.043*** 0.069** 0.043*** 0.016 (0.004) (0.019) (0.006) (0.031) (0.007) (0.025) Total equity –0.003** –0.009* –0.005** –0.016*** –0.017* –0.014 (0.001) (0.005) (0.002) (0.006) (0.009) (0.015) Loan loss reserves –0.021 0.015 –0.011 –0.037 –0.021 0.113 (0.021) (0.045) (0.021) (0.067) (0.029) (0.073) Net interest income –0.009 0.040* 0.012 0.069** –0.023 0.014 (0.010) (0.024) (0.010) (0.028) (0.015) (0.053) ROA 0.008 0.021 –0.016 –0.032 0.019 0.073 (0.013) (0.036) (0.017) (0.057) (0.022) (0.051) Cash –0.003** –0.005*** –0.007*** –0.011*** –0.003** –0.004 (0.001) (0.002) (0.002) (0.004) (0.001) (0.003) Deposits –0.004*** 0.004** –0.005*** –0.001 –0.003*** 0.006** (0.001) (0.001) (0.001) (0.002) (0.001) (0.003) Size –0.007* –0.001 –0.003 0.003 –0.010 –0.005 (0.004) (0.009) (0.004) (0.013) (0.007) (0.016) Debt 0.009*** 0.011 0.014*** 0.009 0.003 0.008 (0.003) (0.008) (0.004) (0.009) (0.004) (0.012) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R-squared 0.351 0.145 0.442 0.073 0.358 0.242 Observations 613 712 319 372 294 340 Table IV. The effect of the negative liquidity shock on banks’ outstanding long-term debt and their loan supply The regressions in this table examine the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. Columns (1) and (2) show results for all banks and Columns (3)–(6) show results for well-capitalized and under-capitalized banks. A bank is defined as well-capitalized if its total equity ratio was above the median in 2006, and as under-capitalized otherwise. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.037** –0.085** –0.011 –0.035 –0.053** –0.121** (0.016) (0.038) (0.027) (0.044) (0.023) (0.054) Maturing debt –0.022 –0.140*** –0.008 –0.079* –0.017 –0.180** (0.020) (0.034) (0.029) (0.045) (0.032) (0.068) Crisis 0.017*** 0.019 0.043*** 0.069** 0.043*** 0.016 (0.004) (0.019) (0.006) (0.031) (0.007) (0.025) Total equity –0.003** –0.009* –0.005** –0.016*** –0.017* –0.014 (0.001) (0.005) (0.002) (0.006) (0.009) (0.015) Loan loss reserves –0.021 0.015 –0.011 –0.037 –0.021 0.113 (0.021) (0.045) (0.021) (0.067) (0.029) (0.073) Net interest income –0.009 0.040* 0.012 0.069** –0.023 0.014 (0.010) (0.024) (0.010) (0.028) (0.015) (0.053) ROA 0.008 0.021 –0.016 –0.032 0.019 0.073 (0.013) (0.036) (0.017) (0.057) (0.022) (0.051) Cash –0.003** –0.005*** –0.007*** –0.011*** –0.003** –0.004 (0.001) (0.002) (0.002) (0.004) (0.001) (0.003) Deposits –0.004*** 0.004** –0.005*** –0.001 –0.003*** 0.006** (0.001) (0.001) (0.001) (0.002) (0.001) (0.003) Size –0.007* –0.001 –0.003 0.003 –0.010 –0.005 (0.004) (0.009) (0.004) (0.013) (0.007) (0.016) Debt 0.009*** 0.011 0.014*** 0.009 0.003 0.008 (0.003) (0.008) (0.004) (0.009) (0.004) (0.012) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R-squared 0.351 0.145 0.442 0.073 0.358 0.242 Observations 613 712 319 372 294 340 All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.037** –0.085** –0.011 –0.035 –0.053** –0.121** (0.016) (0.038) (0.027) (0.044) (0.023) (0.054) Maturing debt –0.022 –0.140*** –0.008 –0.079* –0.017 –0.180** (0.020) (0.034) (0.029) (0.045) (0.032) (0.068) Crisis 0.017*** 0.019 0.043*** 0.069** 0.043*** 0.016 (0.004) (0.019) (0.006) (0.031) (0.007) (0.025) Total equity –0.003** –0.009* –0.005** –0.016*** –0.017* –0.014 (0.001) (0.005) (0.002) (0.006) (0.009) (0.015) Loan loss reserves –0.021 0.015 –0.011 –0.037 –0.021 0.113 (0.021) (0.045) (0.021) (0.067) (0.029) (0.073) Net interest income –0.009 0.040* 0.012 0.069** –0.023 0.014 (0.010) (0.024) (0.010) (0.028) (0.015) (0.053) ROA 0.008 0.021 –0.016 –0.032 0.019 0.073 (0.013) (0.036) (0.017) (0.057) (0.022) (0.051) Cash –0.003** –0.005*** –0.007*** –0.011*** –0.003** –0.004 (0.001) (0.002) (0.002) (0.004) (0.001) (0.003) Deposits –0.004*** 0.004** –0.005*** –0.001 –0.003*** 0.006** (0.001) (0.001) (0.001) (0.002) (0.001) (0.003) Size –0.007* –0.001 –0.003 0.003 –0.010 –0.005 (0.004) (0.009) (0.004) (0.013) (0.007) (0.016) Debt 0.009*** 0.011 0.014*** 0.009 0.003 0.008 (0.003) (0.008) (0.004) (0.009) (0.004) (0.012) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R-squared 0.351 0.145 0.442 0.073 0.358 0.242 Observations 613 712 319 372 294 340 Table V. The effect of the negative liquidity shock on the natural logarithm of banks’ outstanding long-term debt and their loan supply The regressions in this table explore the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply. Dependent variables are the natural logarithm of outstanding long-term debt and total loans. Columns (1) and (2) show results for all banks and Columns (3)–(6) show results for well-capitalized and under-capitalized banks. A bank is defined as well-capitalized if its total equity ratio was above the median in 2006, and as under-capitalized otherwise. The variable “Crisis” is the post-treatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.609*** –0.108* –0.428 –0.018 –0.804*** –0.180** (0.155) (0.055) (0.300) (0.057) (0.163) (0.073) Maturing debt –0.497* –0.181*** –0.418 –0.120* –0.442 –0.218** (0.279) (0.046) (0.415) (0.064) (0.412) (0.085) Crisis 0.168*** –0.006 0.219*** 0.061** 0.383*** –0.012 (0.041) (0.019) (0.073) (0.030) (0.063) (0.024) Total equity –0.028* –0.011* –0.058** –0.020** –0.088 –0.014 (0.017) (0.006) (0.026) (0.008) (0.081) (0.019) Loan loss reserves –0.346 –0.009 –0.470* –0.075 –0.017 0.116 (0.256) (0.064) (0.278) (0.103) (0.296) (0.088) Net interest income –0.058 0.061** 0.212 0.097*** –0.254* 0.042 (0.123) (0.029) (0.142) (0.035) (0.153) (0.058) ROA –0.083 0.027 –0.388 –0.030 0.164 0.070 (0.180) (0.051) (0.241) (0.083) (0.233) (0.063) Cash –0.034** –0.009*** –0.099*** –0.015*** –0.024** –0.006** (0.016) (0.002) (0.030) (0.005) (0.011) (0.003) Deposits –0.032*** 0.005** –0.049*** –0.001 –0.026** 0.007** (0.010) (0.002) (0.015) (0.002) (0.011) (0.003) Size 0.936*** 1.001*** 0.965*** 1.011*** 0.915*** 0.986*** (0.048) (0.011) (0.058) (0.015) (0.079) (0.018) Debt 0.163*** 0.011 0.236*** 0.009 0.081** 0.003 (0.042) (0.010) (0.064) (0.011) (0.039) (0.014) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.889 0.983 0.899 0.985 0.895 0.981 Observations 612 712 318 372 294 340 All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.609*** –0.108* –0.428 –0.018 –0.804*** –0.180** (0.155) (0.055) (0.300) (0.057) (0.163) (0.073) Maturing debt –0.497* –0.181*** –0.418 –0.120* –0.442 –0.218** (0.279) (0.046) (0.415) (0.064) (0.412) (0.085) Crisis 0.168*** –0.006 0.219*** 0.061** 0.383*** –0.012 (0.041) (0.019) (0.073) (0.030) (0.063) (0.024) Total equity –0.028* –0.011* –0.058** –0.020** –0.088 –0.014 (0.017) (0.006) (0.026) (0.008) (0.081) (0.019) Loan loss reserves –0.346 –0.009 –0.470* –0.075 –0.017 0.116 (0.256) (0.064) (0.278) (0.103) (0.296) (0.088) Net interest income –0.058 0.061** 0.212 0.097*** –0.254* 0.042 (0.123) (0.029) (0.142) (0.035) (0.153) (0.058) ROA –0.083 0.027 –0.388 –0.030 0.164 0.070 (0.180) (0.051) (0.241) (0.083) (0.233) (0.063) Cash –0.034** –0.009*** –0.099*** –0.015*** –0.024** –0.006** (0.016) (0.002) (0.030) (0.005) (0.011) (0.003) Deposits –0.032*** 0.005** –0.049*** –0.001 –0.026** 0.007** (0.010) (0.002) (0.015) (0.002) (0.011) (0.003) Size 0.936*** 1.001*** 0.965*** 1.011*** 0.915*** 0.986*** (0.048) (0.011) (0.058) (0.015) (0.079) (0.018) Debt 0.163*** 0.011 0.236*** 0.009 0.081** 0.003 (0.042) (0.010) (0.064) (0.011) (0.039) (0.014) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.889 0.983 0.899 0.985 0.895 0.981 Observations 612 712 318 372 294 340 Table V. The effect of the negative liquidity shock on the natural logarithm of banks’ outstanding long-term debt and their loan supply The regressions in this table explore the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply. Dependent variables are the natural logarithm of outstanding long-term debt and total loans. Columns (1) and (2) show results for all banks and Columns (3)–(6) show results for well-capitalized and under-capitalized banks. A bank is defined as well-capitalized if its total equity ratio was above the median in 2006, and as under-capitalized otherwise. The variable “Crisis” is the post-treatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.609*** –0.108* –0.428 –0.018 –0.804*** –0.180** (0.155) (0.055) (0.300) (0.057) (0.163) (0.073) Maturing debt –0.497* –0.181*** –0.418 –0.120* –0.442 –0.218** (0.279) (0.046) (0.415) (0.064) (0.412) (0.085) Crisis 0.168*** –0.006 0.219*** 0.061** 0.383*** –0.012 (0.041) (0.019) (0.073) (0.030) (0.063) (0.024) Total equity –0.028* –0.011* –0.058** –0.020** –0.088 –0.014 (0.017) (0.006) (0.026) (0.008) (0.081) (0.019) Loan loss reserves –0.346 –0.009 –0.470* –0.075 –0.017 0.116 (0.256) (0.064) (0.278) (0.103) (0.296) (0.088) Net interest income –0.058 0.061** 0.212 0.097*** –0.254* 0.042 (0.123) (0.029) (0.142) (0.035) (0.153) (0.058) ROA –0.083 0.027 –0.388 –0.030 0.164 0.070 (0.180) (0.051) (0.241) (0.083) (0.233) (0.063) Cash –0.034** –0.009*** –0.099*** –0.015*** –0.024** –0.006** (0.016) (0.002) (0.030) (0.005) (0.011) (0.003) Deposits –0.032*** 0.005** –0.049*** –0.001 –0.026** 0.007** (0.010) (0.002) (0.015) (0.002) (0.011) (0.003) Size 0.936*** 1.001*** 0.965*** 1.011*** 0.915*** 0.986*** (0.048) (0.011) (0.058) (0.015) (0.079) (0.018) Debt 0.163*** 0.011 0.236*** 0.009 0.081** 0.003 (0.042) (0.010) (0.064) (0.011) (0.039) (0.014) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.889 0.983 0.899 0.985 0.895 0.981 Observations 612 712 318 372 294 340 All banks Well-capitalized Under-capitalized Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.609*** –0.108* –0.428 –0.018 –0.804*** –0.180** (0.155) (0.055) (0.300) (0.057) (0.163) (0.073) Maturing debt –0.497* –0.181*** –0.418 –0.120* –0.442 –0.218** (0.279) (0.046) (0.415) (0.064) (0.412) (0.085) Crisis 0.168*** –0.006 0.219*** 0.061** 0.383*** –0.012 (0.041) (0.019) (0.073) (0.030) (0.063) (0.024) Total equity –0.028* –0.011* –0.058** –0.020** –0.088 –0.014 (0.017) (0.006) (0.026) (0.008) (0.081) (0.019) Loan loss reserves –0.346 –0.009 –0.470* –0.075 –0.017 0.116 (0.256) (0.064) (0.278) (0.103) (0.296) (0.088) Net interest income –0.058 0.061** 0.212 0.097*** –0.254* 0.042 (0.123) (0.029) (0.142) (0.035) (0.153) (0.058) ROA –0.083 0.027 –0.388 –0.030 0.164 0.070 (0.180) (0.051) (0.241) (0.083) (0.233) (0.063) Cash –0.034** –0.009*** –0.099*** –0.015*** –0.024** –0.006** (0.016) (0.002) (0.030) (0.005) (0.011) (0.003) Deposits –0.032*** 0.005** –0.049*** –0.001 –0.026** 0.007** (0.010) (0.002) (0.015) (0.002) (0.011) (0.003) Size 0.936*** 1.001*** 0.965*** 1.011*** 0.915*** 0.986*** (0.048) (0.011) (0.058) (0.015) (0.079) (0.018) Debt 0.163*** 0.011 0.236*** 0.009 0.081** 0.003 (0.042) (0.010) (0.064) (0.011) (0.039) (0.014) Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.889 0.983 0.899 0.985 0.895 0.981 Observations 612 712 318 372 294 340 These results indicate that banks with a higher exposure to the liquidity shock decreased their outstanding long-term debt significantly more. The results on the difference between well-capitalized and under-capitalized banks further adds to the literature by showing that the effect of higher financing costs is stronger for under-capitalized banks during contraction times. 3.2 Within bank transmission This section studies whether the liquidity shock is transmitted from the liability side to the asset side of a bank’s balance sheet by a reduction in its loan supply. To study the effect of the shock on the amount of loans, I again use a DID estimation method with a continuous treatment variable, Di, and estimate a similar regression to the Equation (1) Lit=β0+β1Di×Pt+β2Di+β3Pt+θXi+δt+uit, (2) where the only difference is the dependent variable Lit which is the total loans for bank i in time period t divided by total assets calculated in 2006. All other independent variables and controls are the same with Equation (1). The coefficient of interest is β1 which captures whether banks with higher exposure to the negative liquidity shock experienced a greater reduction in their loan supply from the pre-crisis years 2006 and 2007 to the crisis years 2008 and 2009. Table IV presents the results of the regression. Column (2) estimates the impact of the liquidity shock on the amount of total loans for all banks. I find that 1 percentage point increase in the fraction of a bank’s long-term debt that matured during the crisis led to a decrease in its loan supply by almost 0.09% of its total assets, which is economically and statistically highly significant. According to Column (4), the liquidity shock did not have any significant effect on well-capitalized banks’ loan supply during the crisis. On the contrary, the effect is stronger for under-capitalized banks as shown in Column (6). An under-capitalized bank with 1 percentage point higher long-term debt ratio that matured during the crisis decreased its loan supply by 0.12% of its total assets.9 These results indicate that banks that are affected more by the liquidity shock decreased their loan supply significantly more. In addition, the negative effect is stronger for under-capitalized banks, whereas it disappears for well-capitalized banks. I follow Puri, Rocholl, and Steffen (2011) and ask the question whether banks affected by the liquidity shock reduced their lending to reduce portfolio risk or to preserve liquidity. To investigate this, I first analyze the effect of the liquidity shock on banks’ nonperforming loans, loan loss reserves until the end of 2011 to capture the long-term effects. Column (1) in Table VI show that 1 percentage point increase in the long-term debt ratio that matured during the crisis led to a significant decrease in a bank’s nonperforming loans by almost 0.01% of its total assets. This suggests that banks affected by the liquidity shock reduced the risk of their loan portfolio. This is supported by the reduction in loan loss reserves: a bank with a 0.10 higher long-term debt ratio that matured during the crisis decreased its loan loss reserves by 0.03% of its total assets. I next analyze the effect on banks’ cash holdings to answer the question whether these affected banks preserved liquidity by reducing their lending activities. As shown in Column (3), I find that an increase in the long-term debt ratio matured during the crisis did not have any significant effect on the cash holdings of a bank. This indicates that affected banks did not preserve liquidity. Taken together, these results suggest that the banks affected heavily by the negative liquidity shock decreased their loan supply significantly, and this enabled them to reduce their portfolio risk. Table VI. The effect of the negative liquidity shock on nonperforming loans, loan loss reserves and cash holdings The regressions in this table investigate the effect of the liquidity shock on banks’ nonperforming loans, loan loss reserves and cash holdings. Dependent variables are the nonperforming loans, loan loss reserves and cash holdings divided by 2006 total assets. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. For results on nonperforming loans and loan loss reserves, as shown in Columns (1) and (2), the crisis dummy is extended 2 more years and includes 2010 and 2011 to capture the long-term effect. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Dependent variable Nonperforming loans Loan loss reserves Cash holdings Maturing debt × Crisis –0.008* –0.003* –0.020 (0.005) (0.002) (0.029) Maturing debt –0.001 –0.002* 0.021 (0.003) (0.001) (0.015) Crisis 0.027*** 0.008*** 0.032*** (0.002) (0.001) (0.004) Total equity –0.001 –0.000 0.000 (0.000) (0.000) (0.001) Loan loss reserves 0.001 0.009*** –0.035* (0.004) (0.001) (0.020) Net interest income 0.006*** 0.001* 0.020 (0.002) (0.001) (0.015) ROA –0.006* 0.001 –0.018 (0.004) (0.001) (0.015) Cash –0.000 –0.000** 0.009*** (0.000) (0.000) (0.001) Deposits 0.000 0.000 –0.000 (0.000) (0.000) (0.001) Size 0.001 0.001** 0.003 (0.001) (0.000) (0.002) Debt 0.001 0.000** 0.001 (0.001) (0.000) (0.002) Time fixed effects Yes Yes Yes Adjusted R2 0.342 0.432 0.500 Observations 1045 1058 712 Dependent variable Nonperforming loans Loan loss reserves Cash holdings Maturing debt × Crisis –0.008* –0.003* –0.020 (0.005) (0.002) (0.029) Maturing debt –0.001 –0.002* 0.021 (0.003) (0.001) (0.015) Crisis 0.027*** 0.008*** 0.032*** (0.002) (0.001) (0.004) Total equity –0.001 –0.000 0.000 (0.000) (0.000) (0.001) Loan loss reserves 0.001 0.009*** –0.035* (0.004) (0.001) (0.020) Net interest income 0.006*** 0.001* 0.020 (0.002) (0.001) (0.015) ROA –0.006* 0.001 –0.018 (0.004) (0.001) (0.015) Cash –0.000 –0.000** 0.009*** (0.000) (0.000) (0.001) Deposits 0.000 0.000 –0.000 (0.000) (0.000) (0.001) Size 0.001 0.001** 0.003 (0.001) (0.000) (0.002) Debt 0.001 0.000** 0.001 (0.001) (0.000) (0.002) Time fixed effects Yes Yes Yes Adjusted R2 0.342 0.432 0.500 Observations 1045 1058 712 Table VI. The effect of the negative liquidity shock on nonperforming loans, loan loss reserves and cash holdings The regressions in this table investigate the effect of the liquidity shock on banks’ nonperforming loans, loan loss reserves and cash holdings. Dependent variables are the nonperforming loans, loan loss reserves and cash holdings divided by 2006 total assets. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. For results on nonperforming loans and loan loss reserves, as shown in Columns (1) and (2), the crisis dummy is extended 2 more years and includes 2010 and 2011 to capture the long-term effect. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Dependent variable Nonperforming loans Loan loss reserves Cash holdings Maturing debt × Crisis –0.008* –0.003* –0.020 (0.005) (0.002) (0.029) Maturing debt –0.001 –0.002* 0.021 (0.003) (0.001) (0.015) Crisis 0.027*** 0.008*** 0.032*** (0.002) (0.001) (0.004) Total equity –0.001 –0.000 0.000 (0.000) (0.000) (0.001) Loan loss reserves 0.001 0.009*** –0.035* (0.004) (0.001) (0.020) Net interest income 0.006*** 0.001* 0.020 (0.002) (0.001) (0.015) ROA –0.006* 0.001 –0.018 (0.004) (0.001) (0.015) Cash –0.000 –0.000** 0.009*** (0.000) (0.000) (0.001) Deposits 0.000 0.000 –0.000 (0.000) (0.000) (0.001) Size 0.001 0.001** 0.003 (0.001) (0.000) (0.002) Debt 0.001 0.000** 0.001 (0.001) (0.000) (0.002) Time fixed effects Yes Yes Yes Adjusted R2 0.342 0.432 0.500 Observations 1045 1058 712 Dependent variable Nonperforming loans Loan loss reserves Cash holdings Maturing debt × Crisis –0.008* –0.003* –0.020 (0.005) (0.002) (0.029) Maturing debt –0.001 –0.002* 0.021 (0.003) (0.001) (0.015) Crisis 0.027*** 0.008*** 0.032*** (0.002) (0.001) (0.004) Total equity –0.001 –0.000 0.000 (0.000) (0.000) (0.001) Loan loss reserves 0.001 0.009*** –0.035* (0.004) (0.001) (0.020) Net interest income 0.006*** 0.001* 0.020 (0.002) (0.001) (0.015) ROA –0.006* 0.001 –0.018 (0.004) (0.001) (0.015) Cash –0.000 –0.000** 0.009*** (0.000) (0.000) (0.001) Deposits 0.000 0.000 –0.000 (0.000) (0.000) (0.001) Size 0.001 0.001** 0.003 (0.001) (0.000) (0.002) Debt 0.001 0.000** 0.001 (0.001) (0.000) (0.002) Time fixed effects Yes Yes Yes Adjusted R2 0.342 0.432 0.500 Observations 1045 1058 712 3.2.a. Cross-sectional results An important question is whether bank-specific characteristics affect the impact of the liquidity shock on banks’ loan supply. Following the literature, I am interested in two bank characteristics: deposits and short-term debt. First, I estimate the regressions separately for banks with higher deposits and banks with lower deposits. A bank is defined to have higher deposits if its total deposits ratio was above the median in 2006 and lower deposits otherwise. Recent studies show that banks with higher deposits cut their overall lending less severely during the last financial crisis (see, e.g., Ivashina and Scharfstein, 2010; Cornett et al., 2011; Dagher and Kazimov, 2015). They argue that banks with better access to deposit financing are more robust to liquidity shocks because deposits are a more stable source of funding (Gatev and Strahan, 2006). Thus, I expect a smaller effect of the liquidity shock on banks with higher deposits relative to banks with lower deposits. Table VII shows that the liquidity shock does not have any significant effect on the change in the amount of total loans for banks with higher deposits although they decreased their long-term debt significantly. On the contrary, banks with lower deposits experienced a significant decline in their loans when they were hit by the liquidity shock. A bank with 1 percentage point higher long-term debt ratio that matured during the crisis had a significant reduction in its loan supply by almost 0.10% of its total assets and in its long-term debt by almost 0.04% of its total assets.10 These results support the recent literature on the conclusion that banks with larger amount of deposits are more robust to negative liquidity shocks. Table VII. Cross-sectional results: deposits Regressions show the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply by deposits. A bank is defined to have higher deposits if its total deposits ratio was above the median in 2006 and lower deposits otherwise. Columns (1) and (2) show results for banks with higher deposits and Columns (3) and (4) show results for banks with lower deposits. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Higher deposits Lower deposits Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.036** –0.055 –0.035 –0.095** (0.017) (0.057) (0.022) (0.038) Maturing debt –0.003 –0.105* –0.030 –0.192*** (0.020) (0.053) (0.033) (0.048) Crisis 0.022*** 0.005 0.043*** 0.055*** (0.005) (0.018) (0.006) (0.018) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.375 0.043 0.211 0.210 Observations 285 353 328 359 Higher deposits Lower deposits Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.036** –0.055 –0.035 –0.095** (0.017) (0.057) (0.022) (0.038) Maturing debt –0.003 –0.105* –0.030 –0.192*** (0.020) (0.053) (0.033) (0.048) Crisis 0.022*** 0.005 0.043*** 0.055*** (0.005) (0.018) (0.006) (0.018) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.375 0.043 0.211 0.210 Observations 285 353 328 359 Table VII. Cross-sectional results: deposits Regressions show the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply by deposits. A bank is defined to have higher deposits if its total deposits ratio was above the median in 2006 and lower deposits otherwise. Columns (1) and (2) show results for banks with higher deposits and Columns (3) and (4) show results for banks with lower deposits. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Higher deposits Lower deposits Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.036** –0.055 –0.035 –0.095** (0.017) (0.057) (0.022) (0.038) Maturing debt –0.003 –0.105* –0.030 –0.192*** (0.020) (0.053) (0.033) (0.048) Crisis 0.022*** 0.005 0.043*** 0.055*** (0.005) (0.018) (0.006) (0.018) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.375 0.043 0.211 0.210 Observations 285 353 328 359 Higher deposits Lower deposits Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.036** –0.055 –0.035 –0.095** (0.017) (0.057) (0.022) (0.038) Maturing debt –0.003 –0.105* –0.030 –0.192*** (0.020) (0.053) (0.033) (0.048) Crisis 0.022*** 0.005 0.043*** 0.055*** (0.005) (0.018) (0.006) (0.018) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.375 0.043 0.211 0.210 Observations 285 353 328 359 Second, I estimate the main specification separately for banks with higher short-term debt holdings and banks with lower short-term debt holdings. I define a bank with higher short-term debt if its short-term debt ratio was above the median in 2006, and as a bank with lower short-term debt otherwise. A bank with more short-term debt holdings is expected to be affected more strongly by the negative liquidity shock since these banks faced higher roll-over risk during the 2007–09 financial crisis (see, e.g., Cornett et al., 2011; Huang and Ratnovski, 2011; Dagher and Kazimov, 2015). This is due to the fact that short-term financiers are uninsured creditors and more at risk of realizing losses. Banks that rely more on short-term debt had to find alternative financing sources to cover their shortfalls during the crisis. Thus, a bank with a high amount of short-term debt was affected more strongly from its maturing long-term debt than a bank with a low amount of short-term debt. Table VIII reports the results. I find that banks with higher short-term debt ratios decreased their long-term debt by 0.04% of their total assets and their loan supply by almost 0.12% of their total assets if there is 1 percentage point increase in the fraction of their long-term debt that matured during the crisis. The liquidity shock does not have any significant effect on the banks with lower short-term debt ratios. According to my results, banks with higher short-term debt holdings were more prone to the negative liquidity shock which is consistent with the literature. Table VIII. Cross-sectional results: short-term debt Regressions present the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply by short-term debt holdings. A bank is defined to have higher short-term debt if its short-term debt ratio was above the median in 2006, and lower short-term debt otherwise. Columns (1) and (2) show results for banks with higher short-term debt and Columns (3) and (4) show results for banks with lower short-term debt. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Higher short-term debt Lower short-term debt Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.040** –0.118** –0.039 0.052 (0.019) (0.046) (0.024) (0.079) Maturing debt –0.031* –0.146*** 0.059* –0.151*** (0.017) (0.042) (0.032) (0.057) Crisis 0.043*** 0.065** 0.025*** 0.072*** (0.006) (0.032) (0.005) (0.016) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R-squared 0.344 0.147 0.438 0.136 Observations 333 407 280 305 Higher short-term debt Lower short-term debt Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.040** –0.118** –0.039 0.052 (0.019) (0.046) (0.024) (0.079) Maturing debt –0.031* –0.146*** 0.059* –0.151*** (0.017) (0.042) (0.032) (0.057) Crisis 0.043*** 0.065** 0.025*** 0.072*** (0.006) (0.032) (0.005) (0.016) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R-squared 0.344 0.147 0.438 0.136 Observations 333 407 280 305 Table VIII. Cross-sectional results: short-term debt Regressions present the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply by short-term debt holdings. A bank is defined to have higher short-term debt if its short-term debt ratio was above the median in 2006, and lower short-term debt otherwise. Columns (1) and (2) show results for banks with higher short-term debt and Columns (3) and (4) show results for banks with lower short-term debt. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. The variable “Crisis” is the posttreatment indicator equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Higher short-term debt Lower short-term debt Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.040** –0.118** –0.039 0.052 (0.019) (0.046) (0.024) (0.079) Maturing debt –0.031* –0.146*** 0.059* –0.151*** (0.017) (0.042) (0.032) (0.057) Crisis 0.043*** 0.065** 0.025*** 0.072*** (0.006) (0.032) (0.005) (0.016) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R-squared 0.344 0.147 0.438 0.136 Observations 333 407 280 305 Higher short-term debt Lower short-term debt Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.040** –0.118** –0.039 0.052 (0.019) (0.046) (0.024) (0.079) Maturing debt –0.031* –0.146*** 0.059* –0.151*** (0.017) (0.042) (0.032) (0.057) Crisis 0.043*** 0.065** 0.025*** 0.072*** (0.006) (0.032) (0.005) (0.016) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R-squared 0.344 0.147 0.438 0.136 Observations 333 407 280 305 3.2.b. Results for different types of loans I next investigate the impact of the liquidity shock on different types of loans. I consider three types of loans which constitute 86.9% of total loans. The largest one is the real estate loans which are 53% of total loans. The second one is the C&I loans with 20.9% followed by consumer loans that represent 13% of total loans. Table IX presents the results of the regressions for three types of loans. Column (1) estimates the impact of the liquidity shock on real estate loans for all banks. I find that 1 percentage point increase in the fraction of a bank’s long-term debt that matured during the crisis generated a decrease in its real estate loans by 0.06% of its total assets, which is economically and statistically highly significant. Column (2) shows that a bank with 1 percentage point higher long-term debt ratio that matured during the crisis decreased its consumer loans by 0.02% of its total assets. According to Column (3), the effect of the liquidity shock on C&I loans is insignificant. I then repeat the analysis for well-capitalized and under-capitalized banks. The results for well-capitalized banks, shown in Columns (1), (2), and (3) in Table X, indicate that the liquidity shock did not have a significant effect on any of the three types of loans. However, the negative effect of the liquidity shock on the real estate loans and consumer loans is stronger for under-capitalized banks. As shown in Columns (4) and (5), an under-capitalized bank with 1 percentage point increase in its long-term debt ratio that matured during the crisis had a significant decrease in its real estate loans by 0.08% of its total assets and in its consumer loans by 0.02% of its total assets. Table IX. The effect of the negative liquidity shock on different types of loans This table shows the regression results for examining the effect of the liquidity shock on different types of loans: real-estate loans, consumer loans and C&I loans. Dependent variables are these three types of loans divided by 2006 total assets. The variable “Crisis” is the post-treatment indicator which is equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Dependent variable Real estate loans Consumer loans C&I loans Maturing debt × Crisis –0.063** –0.023*** 0.009 (0.028) (0.009) (0.016) Maturing debt –0.015 0.002 –0.090*** (0.034) (0.026) (0.025) Crisis 0.043*** 0.004 0.010 (0.009) (0.004) (0.007) Total equity –0.002 –0.005** –0.001 (0.005) (0.002) (0.003) Loan loss reserves –0.074 0.065** 0.021 (0.047) (0.027) (0.035) Net interest income 0.003 –0.009 0.017 (0.024) (0.011) (0.017) ROA 0.000 0.021 0.023 (0.034) (0.018) (0.025) Cash –0.003** –0.001** –0.001 (0.001) (0.001) (0.001) Deposits –0.000 0.001 0.002** (0.002) (0.001) (0.001) Size –0.033*** 0.015*** 0.009 (0.009) (0.005) (0.007) Debt –0.006 0.007 0.014** (0.009) (0.004) (0.006) Time fixed effects Yes Yes Yes Adjusted R-squared 0.169 0.157 0.070 Observations 611 646 667 All banks Dependent variable Real estate loans Consumer loans C&I loans Maturing debt × Crisis –0.063** –0.023*** 0.009 (0.028) (0.009) (0.016) Maturing debt –0.015 0.002 –0.090*** (0.034) (0.026) (0.025) Crisis 0.043*** 0.004 0.010 (0.009) (0.004) (0.007) Total equity –0.002 –0.005** –0.001 (0.005) (0.002) (0.003) Loan loss reserves –0.074 0.065** 0.021 (0.047) (0.027) (0.035) Net interest income 0.003 –0.009 0.017 (0.024) (0.011) (0.017) ROA 0.000 0.021 0.023 (0.034) (0.018) (0.025) Cash –0.003** –0.001** –0.001 (0.001) (0.001) (0.001) Deposits –0.000 0.001 0.002** (0.002) (0.001) (0.001) Size –0.033*** 0.015*** 0.009 (0.009) (0.005) (0.007) Debt –0.006 0.007 0.014** (0.009) (0.004) (0.006) Time fixed effects Yes Yes Yes Adjusted R-squared 0.169 0.157 0.070 Observations 611 646 667 Table IX. The effect of the negative liquidity shock on different types of loans This table shows the regression results for examining the effect of the liquidity shock on different types of loans: real-estate loans, consumer loans and C&I loans. Dependent variables are these three types of loans divided by 2006 total assets. The variable “Crisis” is the post-treatment indicator which is equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. The variable “Size” is the natural logarithm of total assets. All other variables are defined in Table I. All columns include controls for year fixed effects and bank characteristics calculated in 2006. Intercept and year fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Dependent variable Real estate loans Consumer loans C&I loans Maturing debt × Crisis –0.063** –0.023*** 0.009 (0.028) (0.009) (0.016) Maturing debt –0.015 0.002 –0.090*** (0.034) (0.026) (0.025) Crisis 0.043*** 0.004 0.010 (0.009) (0.004) (0.007) Total equity –0.002 –0.005** –0.001 (0.005) (0.002) (0.003) Loan loss reserves –0.074 0.065** 0.021 (0.047) (0.027) (0.035) Net interest income 0.003 –0.009 0.017 (0.024) (0.011) (0.017) ROA 0.000 0.021 0.023 (0.034) (0.018) (0.025) Cash –0.003** –0.001** –0.001 (0.001) (0.001) (0.001) Deposits –0.000 0.001 0.002** (0.002) (0.001) (0.001) Size –0.033*** 0.015*** 0.009 (0.009) (0.005) (0.007) Debt –0.006 0.007 0.014** (0.009) (0.004) (0.006) Time fixed effects Yes Yes Yes Adjusted R-squared 0.169 0.157 0.070 Observations 611 646 667 All banks Dependent variable Real estate loans Consumer loans C&I loans Maturing debt × Crisis –0.063** –0.023*** 0.009 (0.028) (0.009) (0.016) Maturing debt –0.015 0.002 –0.090*** (0.034) (0.026) (0.025) Crisis 0.043*** 0.004 0.010 (0.009) (0.004) (0.007) Total equity –0.002 –0.005** –0.001 (0.005) (0.002) (0.003) Loan loss reserves –0.074 0.065** 0.021 (0.047) (0.027) (0.035) Net interest income 0.003 –0.009 0.017 (0.024) (0.011) (0.017) ROA 0.000 0.021 0.023 (0.034) (0.018) (0.025) Cash –0.003** –0.001** –0.001 (0.001) (0.001) (0.001) Deposits –0.000 0.001 0.002** (0.002) (0.001) (0.001) Size –0.033*** 0.015*** 0.009 (0.009) (0.005) (0.007) Debt –0.006 0.007 0.014** (0.009) (0.004) (0.006) Time fixed effects Yes Yes Yes Adjusted R-squared 0.169 0.157 0.070 Observations 611 646 667 Table X. The effect of the negative liquidity shock on different types of loans for well-capitalized and under-capitalized banks The regressions in this table examine the effect of the liquidity shock on different types of loans, namely real-estate loans, consumer loans and C&I loans, for well-capitalized and under-capitalized banks, separately. A bank is defined as well-capitalized if its total equity ratio was above the median in 2006, and as under-capitalized otherwise. Dependent variables are these three types of loans divided by 2006 total assets. The variable “Crisis” is the post-treatment indicator which is equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Well-capitalized banks Under-capitalized banks Dependent variable Real estate Consumer Commercial Real estate Consumer Commercial Maturing debt × Crisis –0.061 –0.006 0.012 –0.079** –0.024** 0.009 (0.040) (0.018) (0.016) (0.036) (0.011) (0.032) Maturing debt 0.041 0.054 –0.118*** –0.041 –0.010 –0.070 (0.056) (0.038) (0.038) (0.066) (0.030) (0.058) Crisis 0.041*** 0.014** –0.008 0.042*** 0.001 0.016 (0.011) (0.006) (0.006) (0.013) (0.003) (0.013) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.201 0.260 0.118 0.170 0.092 0.056 Observations 331 345 345 280 301 322 Well-capitalized banks Under-capitalized banks Dependent variable Real estate Consumer Commercial Real estate Consumer Commercial Maturing debt × Crisis –0.061 –0.006 0.012 –0.079** –0.024** 0.009 (0.040) (0.018) (0.016) (0.036) (0.011) (0.032) Maturing debt 0.041 0.054 –0.118*** –0.041 –0.010 –0.070 (0.056) (0.038) (0.038) (0.066) (0.030) (0.058) Crisis 0.041*** 0.014** –0.008 0.042*** 0.001 0.016 (0.011) (0.006) (0.006) (0.013) (0.003) (0.013) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.201 0.260 0.118 0.170 0.092 0.056 Observations 331 345 345 280 301 322 Table X. The effect of the negative liquidity shock on different types of loans for well-capitalized and under-capitalized banks The regressions in this table examine the effect of the liquidity shock on different types of loans, namely real-estate loans, consumer loans and C&I loans, for well-capitalized and under-capitalized banks, separately. A bank is defined as well-capitalized if its total equity ratio was above the median in 2006, and as under-capitalized otherwise. Dependent variables are these three types of loans divided by 2006 total assets. The variable “Crisis” is the post-treatment indicator which is equal to one in 2008 and 2009, and zero in 2006 and 2007. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Well-capitalized banks Under-capitalized banks Dependent variable Real estate Consumer Commercial Real estate Consumer Commercial Maturing debt × Crisis –0.061 –0.006 0.012 –0.079** –0.024** 0.009 (0.040) (0.018) (0.016) (0.036) (0.011) (0.032) Maturing debt 0.041 0.054 –0.118*** –0.041 –0.010 –0.070 (0.056) (0.038) (0.038) (0.066) (0.030) (0.058) Crisis 0.041*** 0.014** –0.008 0.042*** 0.001 0.016 (0.011) (0.006) (0.006) (0.013) (0.003) (0.013) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.201 0.260 0.118 0.170 0.092 0.056 Observations 331 345 345 280 301 322 Well-capitalized banks Under-capitalized banks Dependent variable Real estate Consumer Commercial Real estate Consumer Commercial Maturing debt × Crisis –0.061 –0.006 0.012 –0.079** –0.024** 0.009 (0.040) (0.018) (0.016) (0.036) (0.011) (0.032) Maturing debt 0.041 0.054 –0.118*** –0.041 –0.010 –0.070 (0.056) (0.038) (0.038) (0.066) (0.030) (0.058) Crisis 0.041*** 0.014** –0.008 0.042*** 0.001 0.016 (0.011) (0.006) (0.006) (0.013) (0.003) (0.013) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.201 0.260 0.118 0.170 0.092 0.056 Observations 331 345 345 280 301 322 According to these results, banks with a higher fraction of long-term debt that matured during the crisis decreased their real estate loans and consumer loans, whereas there are no significant reduction in their C&I loans. The reduction is, particularly, strong for real estate loans as compared with consumer loans. This supports the results of Puri, Rocholl, and Steffen (2011) on German savings banks. They find that banks affected by the US financial crisis rejected substantially more loan applications than non-affected banks, and this result is particularly strong for mortgage loans relative to consumer loans. One possible explanation for this result could be that banks are willing to continue their lending relationships with existing borrower firms so that they are more likely to cut on real estate and consumer loans relative to C&I loans in distress times. In relation to this, Iyer et al. (2014) show that higher strength of an existing lending relationship lowers the negative effect of liquidity shocks on banks’ loan supply. Their results suggest that banking relationships mitigate credit supply restrictions. Another explanation could be that borrower firms draw funds from existing credit lines with their banks when the supply of market liquidity falls. Ivashina and Scharfstein (2010) present that firms drew on their credit lines during the last financial crisis and this led to a spike in C&I loans reported on bank balance sheets. This might indicate that banks are more likely to reduce lending in real estate and consumer loans when liquidity becomes scarce because they might have to continue to provide liquidity for borrower firms through the existing credit lines. 3.3 Transmission to the real economy: evidence from house prices In normal times, when there are no frictions in lending, the effect of the negative liquidity shock on house prices through a reduction in real estate loans can be offset by switching across banks. However, during crisis times, frictions in the economy prevent switching across banks, and the liquidity shock is expected to affect the allocation of lending, and as a result house prices might be affected. As banks reduce their real estate loans, the demand for housing decreases and this pushes house prices down. I use the HPI data at the MSA level to study the transmission of the liquidity shock to the real economy by exploring the effect on house prices. To test for the effect, I again use a DID estimation method with a continuous treatment variable, Djt, and estimate △Hjt=γ0+γ1Djt×Pt+γ2Djt+γ3Pt+θYjt+μZjt+λj+δt+ujt, (3) where Hjt is the natural logarithm of the MSA HPI as given by FHFA for MSA j in time period t. △Hjt which is the change in the natural logarithm of HPIs, the growth rate of house prices, is used as the dependent variable because it is argued in the literature that house prices in the USA display heterogeneous trends (Favara and Imbs, 2015) and that HPI has no economic interpretation (Himmelberg, Mayer, and Sinai, 2005). The time period captures 4 years: 2006, 2007, 2008, and 2009. Pt is the posttreatment indicator that is equal to one in 2008 and 2009, and zero in 2006 and 2007. Djt is the weighted average fraction of long-term debt that was issued before the end of 2006 and that matured during the crisis at the MSA level. Following Berger and Bouwman (2009), Flannery and Lin (2015), and Dursun-de Neef and Schandlbauer (2017), the amount of deposits that each bank holds in an MSA is used as a weight that reflects how important an affected bank is for an MSA. Djt is used to measure the exposure of an MSA to the liquidity shock through the branches of the affected banks in an MSA.11 The intuition is that an MSA is affected by the liquidity shock if affected banks have branches in this MSA. Yjt includes the weighted average 2006 values of the size of the banks in each MSA and the long-term debt issued before the end of 2006. Zjt summarizes additional determinants of house prices documented in the literature. I follow Lamont and Stein (1999) and Favara and Imbs (2015), and include contemporaneous and lagged income per capita and population to control for local influences on the real estate market. Following Case and Shiller (1989), I additionally include a lagged value of the house prices and the change in the house prices. I also control for socioeconomic factors at the MSA level by including median age, ratio of owner occupied housing units, fraction of the population above 25 years old with at least a college degree and unemployment rate. λj is the MSA fixed effects and δt is the year fixed effects. The coefficient of interest is γ1, which measures whether MSAs with higher exposure to the negative liquidity shock, through the branches of affected banks that are located in these MSAs, experienced a greater reduction in the house prices from the pre-crisis years 2006 and 2007 to the crisis years 2008 and 2009. To study the effect of the negative liquidity shock on house prices, I use the fraction of deposits that each bank holds in its branches in an MSA as weights to measure the effect of a reduction in the amount of that bank’s real estate loans on house prices in that MSA. I cannot use the fraction of real estate loans due to the lack of data on the outstanding real estate loans at the branch level for each bank. This requires the assumption that the distribution of deposits is similar to the distribution of real estate loans among banks in an MSA. To provide evidence on this assumption, I collect data on the volume of mortgage loans originated by each bank in each MSA from SNL Financial. These data are taken from the Home Mortgage Disclosure Act (HMDA) database. Using the HMDA, data enable me to calculate the market share of each bank in the mortgage loans market for newly issued loans in each MSA and compare this with the deposits market share of that bank in that MSA.12Table XI presents that the correlation between the fraction of mortgage loans originated by a bank in an MSA and the fraction of that bank’s deposits in that MSA is 0.209, and it is highly significant at 1% level. This supports the argument that deposit distribution matches housing loan distribution among banks in an MSA. So the fraction of deposits that each bank holds in an MSA can be used to measure how important that bank is in the real estate loans market in that MSA. Table XI. Correlations: mortgage loans and deposits This table shows the correlation between the fraction of mortgage loans originated by each bank in an MSA with the fraction of deposits that the bank holds in that MSA for 4 years: 2006, 2007, 2008, and 2009. *p < 0.10, **p < 0.05, ***p < 0.01. Correlations Variable Mortgage loans Deposits Mortgage loans 1.000 Deposits 0.209*** 1.000 Correlations Variable Mortgage loans Deposits Mortgage loans 1.000 Deposits 0.209*** 1.000 Table XI. Correlations: mortgage loans and deposits This table shows the correlation between the fraction of mortgage loans originated by each bank in an MSA with the fraction of deposits that the bank holds in that MSA for 4 years: 2006, 2007, 2008, and 2009. *p < 0.10, **p < 0.05, ***p < 0.01. Correlations Variable Mortgage loans Deposits Mortgage loans 1.000 Deposits 0.209*** 1.000 Correlations Variable Mortgage loans Deposits Mortgage loans 1.000 Deposits 0.209*** 1.000 Table XII reports the results of the effect of the negative liquidity shock on the growth rate of house prices. The results show that 1 percentage point increase in the weighted average fraction of long-term debt that matured during the crisis resulted in a 0.13 percentage point decrease in the growth rate of house prices. I next repeat the analysis for MSAs with more well-capitalized banks and MSAs with more under-capitalized banks, separately. I define an MSA with more well-capitalized banks if the weighted average total equity ratio was above median total equity ratio in 2006, and as an MSA with more under-capitalized banks otherwise. I expect that the negative effect of the liquidity shock is stronger for MSAs with more under-capitalized banks since these are the banks that cut their real estate loans significantly. According to the results for MSAs with more well-capitalized banks, 1 percentage point increase in the weighted average fraction of long-term debt that matured during the crisis led to 0.12 percentage point decrease in the growth rate of house prices, whereas the decrease is 0.18 percentage point in the growth rate for MSAs with more under-capitalized banks. Table XII. The transmission of the negative liquidity shock to the real economy: evidence from MSA house prices The regressions in this table investigate the effect of the liquidity shock on MSA house prices. The dependent variable is the change in the natural logarithm of HPIs. The variable “Maturing debt” measures each MSA’s exposure to the negative liquidity shock and is calculated as the weighted average fraction of long-term debt that matured during the crisis, where the amount of deposits that each bank holds in an MSA is used as weights. All other variables are defined in Table I. All columns include controls for year fixed effects, MSA fixed effects and 1 year lagged MSA characteristics. Intercept, year fixed effects and MSA fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the MSA level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All MSAs Well-capitalized Under-capitalized Dependent variable △House price △House price △House price Maturing debt × Crisis –0.130*** –0.120* –0.178*** (0.045) (0.064) (0.061) Maturing debt –0.037 –0.174** 0.083 (0.072) (0.074) (0.101) Crisis –0.026** –0.015 –0.001 (0.012) (0.016) (0.012) Weighted size –0.014** –0.017** –0.011 (0.007) (0.007) (0.009) Weighted total debt 0.330*** 0.255 0.391** (0.120) (0.163) (0.196) Population 0.155 –0.002 0.825 (0.224) (0.202) (0.571) Lagged population 0.010 0.155 –0.644 (0.191) (0.135) (0.480) Income per capita 0.297*** 0.223*** 0.368*** (0.053) (0.079) (0.070) Lagged income per capita 0.147** 0.159* 0.131 (0.063) (0.092) (0.090) Lagged house price –0.574*** –0.535*** –0.624*** (0.031) (0.040) (0.050) Lagged house price growth 0.602*** 0.594*** 0.604*** (0.038) (0.052) (0.058) Median age 0.007 0.016*** 0.001 (0.009) (0.006) (0.013) Owner occupied houses –0.004 –0.001 –0.005* (0.003) (0.020) (0.003) Population with a college degree 0.002 0.007 –0.004 (0.005) (0.006) (0.010) Unemployment rate 0.026 0.048 –0.001 (0.024) (0.036) (0.045) Time fixed effects Yes Yes Yes MSA fixed effects Yes Yes Yes Adjusted R2 0.793 0.813 0.775 Observations 1450 718 732 All MSAs Well-capitalized Under-capitalized Dependent variable △House price △House price △House price Maturing debt × Crisis –0.130*** –0.120* –0.178*** (0.045) (0.064) (0.061) Maturing debt –0.037 –0.174** 0.083 (0.072) (0.074) (0.101) Crisis –0.026** –0.015 –0.001 (0.012) (0.016) (0.012) Weighted size –0.014** –0.017** –0.011 (0.007) (0.007) (0.009) Weighted total debt 0.330*** 0.255 0.391** (0.120) (0.163) (0.196) Population 0.155 –0.002 0.825 (0.224) (0.202) (0.571) Lagged population 0.010 0.155 –0.644 (0.191) (0.135) (0.480) Income per capita 0.297*** 0.223*** 0.368*** (0.053) (0.079) (0.070) Lagged income per capita 0.147** 0.159* 0.131 (0.063) (0.092) (0.090) Lagged house price –0.574*** –0.535*** –0.624*** (0.031) (0.040) (0.050) Lagged house price growth 0.602*** 0.594*** 0.604*** (0.038) (0.052) (0.058) Median age 0.007 0.016*** 0.001 (0.009) (0.006) (0.013) Owner occupied houses –0.004 –0.001 –0.005* (0.003) (0.020) (0.003) Population with a college degree 0.002 0.007 –0.004 (0.005) (0.006) (0.010) Unemployment rate 0.026 0.048 –0.001 (0.024) (0.036) (0.045) Time fixed effects Yes Yes Yes MSA fixed effects Yes Yes Yes Adjusted R2 0.793 0.813 0.775 Observations 1450 718 732 Table XII. The transmission of the negative liquidity shock to the real economy: evidence from MSA house prices The regressions in this table investigate the effect of the liquidity shock on MSA house prices. The dependent variable is the change in the natural logarithm of HPIs. The variable “Maturing debt” measures each MSA’s exposure to the negative liquidity shock and is calculated as the weighted average fraction of long-term debt that matured during the crisis, where the amount of deposits that each bank holds in an MSA is used as weights. All other variables are defined in Table I. All columns include controls for year fixed effects, MSA fixed effects and 1 year lagged MSA characteristics. Intercept, year fixed effects and MSA fixed effects are not shown. Heteroskedasticity consistent standard errors clustered at the MSA level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All MSAs Well-capitalized Under-capitalized Dependent variable △House price △House price △House price Maturing debt × Crisis –0.130*** –0.120* –0.178*** (0.045) (0.064) (0.061) Maturing debt –0.037 –0.174** 0.083 (0.072) (0.074) (0.101) Crisis –0.026** –0.015 –0.001 (0.012) (0.016) (0.012) Weighted size –0.014** –0.017** –0.011 (0.007) (0.007) (0.009) Weighted total debt 0.330*** 0.255 0.391** (0.120) (0.163) (0.196) Population 0.155 –0.002 0.825 (0.224) (0.202) (0.571) Lagged population 0.010 0.155 –0.644 (0.191) (0.135) (0.480) Income per capita 0.297*** 0.223*** 0.368*** (0.053) (0.079) (0.070) Lagged income per capita 0.147** 0.159* 0.131 (0.063) (0.092) (0.090) Lagged house price –0.574*** –0.535*** –0.624*** (0.031) (0.040) (0.050) Lagged house price growth 0.602*** 0.594*** 0.604*** (0.038) (0.052) (0.058) Median age 0.007 0.016*** 0.001 (0.009) (0.006) (0.013) Owner occupied houses –0.004 –0.001 –0.005* (0.003) (0.020) (0.003) Population with a college degree 0.002 0.007 –0.004 (0.005) (0.006) (0.010) Unemployment rate 0.026 0.048 –0.001 (0.024) (0.036) (0.045) Time fixed effects Yes Yes Yes MSA fixed effects Yes Yes Yes Adjusted R2 0.793 0.813 0.775 Observations 1450 718 732 All MSAs Well-capitalized Under-capitalized Dependent variable △House price △House price △House price Maturing debt × Crisis –0.130*** –0.120* –0.178*** (0.045) (0.064) (0.061) Maturing debt –0.037 –0.174** 0.083 (0.072) (0.074) (0.101) Crisis –0.026** –0.015 –0.001 (0.012) (0.016) (0.012) Weighted size –0.014** –0.017** –0.011 (0.007) (0.007) (0.009) Weighted total debt 0.330*** 0.255 0.391** (0.120) (0.163) (0.196) Population 0.155 –0.002 0.825 (0.224) (0.202) (0.571) Lagged population 0.010 0.155 –0.644 (0.191) (0.135) (0.480) Income per capita 0.297*** 0.223*** 0.368*** (0.053) (0.079) (0.070) Lagged income per capita 0.147** 0.159* 0.131 (0.063) (0.092) (0.090) Lagged house price –0.574*** –0.535*** –0.624*** (0.031) (0.040) (0.050) Lagged house price growth 0.602*** 0.594*** 0.604*** (0.038) (0.052) (0.058) Median age 0.007 0.016*** 0.001 (0.009) (0.006) (0.013) Owner occupied houses –0.004 –0.001 –0.005* (0.003) (0.020) (0.003) Population with a college degree 0.002 0.007 –0.004 (0.005) (0.006) (0.010) Unemployment rate 0.026 0.048 –0.001 (0.024) (0.036) (0.045) Time fixed effects Yes Yes Yes MSA fixed effects Yes Yes Yes Adjusted R2 0.793 0.813 0.775 Observations 1450 718 732 The results show that the negative liquidity shock is transmitted to the real economy by a significant reduction in house prices. A natural interpretation is that the negative liquidity shock decreased the supply of real estate loans, which in turn declined the demand for houses, and as a result house prices declined. Although the difference between MSAs with more under-capitalized banks and MSAs with more well-capitalized banks is not significant with a p-value of 0.44, house prices declined more in MSAs with more under-capitalized banks. This is consistent with my previous result that well-capitalized banks could avoid the transmission of the negative liquidity shock. 3.4 Economic magnitude It is important to evaluate whether the effect of maturing long-term debt on lending and house prices during the crisis was of a significant economic magnitude. To address this question, I estimate the hypothetical increase in total loans and house prices if all banks had a lower fraction of long-term debt that matured during the crisis by 10 percentage points. Based on the coefficient of the interaction term (–0.085) reported in Column (2) in Table IV, this would have led to an average relative increase (i.e., smaller decrease) of around $46.5 billion in total loans because total assets of all banks in my sample was around $5.47 trillion in 2006. This implies that total loans would have decreased less by around 31% in this hypothetical scenario because the average reduction in total lending during the crisis was approximately $150 billion in my sample. For house prices, the coefficient of the interaction term (–0.13) shown in Column (1) in Table XII indicates that this hypothetical scenario would have led to an average relative increase in the growth rate of house prices (i.e., smaller decrease) by around 1.3 percentage points. Given that the average drop in the growth rate of house prices during the crisis was 8.6 percentage points in my sample, this hypothetical decrease in long-term debt would have led to 15% less reduction in house prices during the crisis. These estimates indicate that 10 percentage points decrease in the fraction of long-term debt that matured during the crisis could have led to 31% less reduction in total loans and 15% less reduction in house prices, where both effects are economically significant. 4. Robustness 4.1 Falsification test The assumption of the DID methodology is that banks with a larger fraction of long-term debt that matured during the 2007–09 financial crisis were affected more severely by the negative liquidity shock because it was difficult for them to refinance their obligations through alternative financing sources in this contraction period of time. This assumption does not hold in times with easier credit opportunities. This implies that the effect of the maturing long-term debt must be insignificant if the same experiment is repeated for a period of time without a crisis. To verify whether this holds, the same experiment is replicated for the time period of 2003–06, where there was no crisis and it was easier to find alternative financing sources to refinance the maturing long-term debt. I falsely assume that 2005–06 is the crisis and 2003–04 is the pre-crisis period of time, and repeat the same empirical analysis. As reported in Table XIII, the results show that the effect of an increase in the fraction of long-term debt maturing during 2005–06 on the outstanding total long-term debt and the loan supply is indistinguishable from zero. This replication ensures that the observed change in the amount of loans is more likely due to the negative liquidity shock during the 2007–09 financial crisis. Table XIII. Robustness: falsification test assuming 2005–06 is the crisis The regressions in this table falsely assume that 2005–06 is a crisis period of time and show the effect of the fraction of long-term debt that matured during this period on banks’ outstanding long-term debt and their loan supply. Dependent variables are the outstanding long-term debt and total loans divided by 2003 total assets. Columns (1) and (2) show results for all banks and Columns (3)–(6) show results for well-capitalized and under-capitalized banks, where a bank is defined as well-capitalized if its total equity ratio was above the median in 2003, and as under-capitalized otherwise. The variable “Crisis” is the post-treatment indicator equal to one in 2005 and 2006, and zero in 2003 and 2004. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Well-capitalized banks Under-capitalized banks Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.000 –0.032 0.000 0.009 0.003 –0.055 (0.007) (0.028) (0.008) (0.036) (0.010) (0.039) Maturing debt –0.002 0.076* 0.003 0.092 –0.014 0.049 (0.008) (0.043) (0.011) (0.059) (0.009) (0.065) Crisis 0.001 0.132*** 0.006 0.262*** 0.006 0.239*** (0.006) (0.027) (0.007) (0.038) (0.012) (0.045) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.613 0.297 0.725 0.220 0.557 0.339 Observations 438 455 201 214 237 241 All banks Well-capitalized banks Under-capitalized banks Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.000 –0.032 0.000 0.009 0.003 –0.055 (0.007) (0.028) (0.008) (0.036) (0.010) (0.039) Maturing debt –0.002 0.076* 0.003 0.092 –0.014 0.049 (0.008) (0.043) (0.011) (0.059) (0.009) (0.065) Crisis 0.001 0.132*** 0.006 0.262*** 0.006 0.239*** (0.006) (0.027) (0.007) (0.038) (0.012) (0.045) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.613 0.297 0.725 0.220 0.557 0.339 Observations 438 455 201 214 237 241 Table XIII. Robustness: falsification test assuming 2005–06 is the crisis The regressions in this table falsely assume that 2005–06 is a crisis period of time and show the effect of the fraction of long-term debt that matured during this period on banks’ outstanding long-term debt and their loan supply. Dependent variables are the outstanding long-term debt and total loans divided by 2003 total assets. Columns (1) and (2) show results for all banks and Columns (3)–(6) show results for well-capitalized and under-capitalized banks, where a bank is defined as well-capitalized if its total equity ratio was above the median in 2003, and as under-capitalized otherwise. The variable “Crisis” is the post-treatment indicator equal to one in 2005 and 2006, and zero in 2003 and 2004. The variable “Maturing debt” measures each affected bank’s exposure to the negative liquidity shock and is calculated as the fraction of long-term debt that matured during the crisis. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. All banks Well-capitalized banks Under-capitalized banks Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.000 –0.032 0.000 0.009 0.003 –0.055 (0.007) (0.028) (0.008) (0.036) (0.010) (0.039) Maturing debt –0.002 0.076* 0.003 0.092 –0.014 0.049 (0.008) (0.043) (0.011) (0.059) (0.009) (0.065) Crisis 0.001 0.132*** 0.006 0.262*** 0.006 0.239*** (0.006) (0.027) (0.007) (0.038) (0.012) (0.045) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.613 0.297 0.725 0.220 0.557 0.339 Observations 438 455 201 214 237 241 All banks Well-capitalized banks Under-capitalized banks Dependent variable Long-term debt Total loans Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.000 –0.032 0.000 0.009 0.003 –0.055 (0.007) (0.028) (0.008) (0.036) (0.010) (0.039) Maturing debt –0.002 0.076* 0.003 0.092 –0.014 0.049 (0.008) (0.043) (0.011) (0.059) (0.009) (0.065) Crisis 0.001 0.132*** 0.006 0.262*** 0.006 0.239*** (0.006) (0.027) (0.007) (0.038) (0.012) (0.045) Bank controls Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.613 0.297 0.725 0.220 0.557 0.339 Observations 438 455 201 214 237 241 Table XIV. Robustness: TLGP The regressions in this table investigate the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply for TLGP banks versus non-TLGP banks using interaction terms. A bank is defined to be a TLGP bank if the bank issued TLGP debt under the DGP, and TLGP amount is the amount of debt that was issued by the TLGP bank divided by its 2006 total assets. Columns (1) and (2) show results for TLGP banks and Columns (3) and (4) show results for TLGP amounts. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.043*** –0.078** –0.042** –0.093** (0.015) (0.038) (0.016) (0.039) Maturing debt –0.016 –0.134*** –0.016 –0.131*** (0.020) (0.033) (0.022) (0.034) Crisis 0.017*** 0.024 0.017*** 0.020 (0.004) (0.020) (0.004) (0.019) Maturing debt × Crisis × TLGP 0.072 0.099 (0.093) (0.170) Maturing debt × TLGP –0.099 –0.277* (0.062) (0.167) Crisis × TLGP –0.012 –0.061 (0.024) (0.051) TLGP 0.031** 0.068** (0.012) (0.034) Maturing debt × Crisis × TLGP amount 0.020 0.060 (0.013) (0.051) Maturing debt × TLGP amount –0.027* –0.056 (0.015) (0.051) Crisis × TLGP amount –0.002 –0.013 (0.002) (0.008) TLGP amount 0.005*** 0.012* (0.002) (0.007) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.355 0.145 0.353 0.142 Observations 613 712 613 712 Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.043*** –0.078** –0.042** –0.093** (0.015) (0.038) (0.016) (0.039) Maturing debt –0.016 –0.134*** –0.016 –0.131*** (0.020) (0.033) (0.022) (0.034) Crisis 0.017*** 0.024 0.017*** 0.020 (0.004) (0.020) (0.004) (0.019) Maturing debt × Crisis × TLGP 0.072 0.099 (0.093) (0.170) Maturing debt × TLGP –0.099 –0.277* (0.062) (0.167) Crisis × TLGP –0.012 –0.061 (0.024) (0.051) TLGP 0.031** 0.068** (0.012) (0.034) Maturing debt × Crisis × TLGP amount 0.020 0.060 (0.013) (0.051) Maturing debt × TLGP amount –0.027* –0.056 (0.015) (0.051) Crisis × TLGP amount –0.002 –0.013 (0.002) (0.008) TLGP amount 0.005*** 0.012* (0.002) (0.007) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.355 0.145 0.353 0.142 Observations 613 712 613 712 Table XIV. Robustness: TLGP The regressions in this table investigate the effect of the liquidity shock on banks’ outstanding long-term debt and their loan supply for TLGP banks versus non-TLGP banks using interaction terms. A bank is defined to be a TLGP bank if the bank issued TLGP debt under the DGP, and TLGP amount is the amount of debt that was issued by the TLGP bank divided by its 2006 total assets. Columns (1) and (2) show results for TLGP banks and Columns (3) and (4) show results for TLGP amounts. Dependent variables are the outstanding long-term debt and total loans divided by 2006 total assets. All columns include all exogenous variables from the full specification in Table IV. Intercept, year fixed effects and bank controls are not shown in the interest of parsimony. Heteroskedasticity consistent standard errors clustered at the bank level are shown in parentheses. Standard errors in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01. Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.043*** –0.078** –0.042** –0.093** (0.015) (0.038) (0.016) (0.039) Maturing debt –0.016 –0.134*** –0.016 –0.131*** (0.020) (0.033) (0.022) (0.034) Crisis 0.017*** 0.024 0.017*** 0.020 (0.004) (0.020) (0.004) (0.019) Maturing debt × Crisis × TLGP 0.072 0.099 (0.093) (0.170) Maturing debt × TLGP –0.099 –0.277* (0.062) (0.167) Crisis × TLGP –0.012 –0.061 (0.024) (0.051) TLGP 0.031** 0.068** (0.012) (0.034) Maturing debt × Crisis × TLGP amount 0.020 0.060 (0.013) (0.051) Maturing debt × TLGP amount –0.027* –0.056 (0.015) (0.051) Crisis × TLGP amount –0.002 –0.013 (0.002) (0.008) TLGP amount 0.005*** 0.012* (0.002) (0.007) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.355 0.145 0.353 0.142 Observations 613 712 613 712 Dependent variable Long-term debt Total loans Long-term debt Total loans Maturing debt × Crisis –0.043*** –0.078** –0.042** –0.093** (0.015) (0.038) (0.016) (0.039) Maturing debt –0.016 –0.134*** –0.016 –0.131*** (0.020) (0.033) (0.022) (0.034) Crisis 0.017*** 0.024 0.017*** 0.020 (0.004) (0.020) (0.004) (0.019) Maturing debt × Crisis × TLGP 0.072 0.099 (0.093) (0.170) Maturing debt × TLGP –0.099 –0.277* (0.062) (0.167) Crisis × TLGP –0.012 –0.061 (0.024) (0.051) TLGP 0.031** 0.068** (0.012) (0.034) Maturing debt × Crisis × TLGP amount 0.020 0.060 (0.013) (0.051) Maturing debt × TLGP amount –0.027* –0.056 (0.015) (0.051) Crisis × TLGP amount –0.002 –0.013 (0.002) (0.008) TLGP amount 0.005*** 0.012* (0.002) (0.007) Bank controls Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Adjusted R2 0.355 0.145 0.353 0.142 Observations 613 712 613 712 4.2 Temporary liquidity guarantee program On October 2008, the FDIC implemented the Temporary Liquidity Guarantee Program (TLGP). Under the Debt Guarantee Program (DGP), which was one of the components of the TLGP, the FDIC guaranteed in full the senior unsecured debt issued by a participating financial institution until the end of 2012. This enabled the banks and their holding companies to roll-over their short-term liabilities and issue longer-term debt during the financial crisis. In this section, I study whether the DGP has an impact on the transmission of the negative liquidity shock by enabling banks to roll over their maturing long-term debt at affordable costs. A bank is defined to be a TLGP bank if the bank issued debt under the guarantee of the DGP. Table XIV presents the results. Columns (1) and (2) show that being a TLGP bank did not have a significant effect on the reduction of banks’ outstanding long-term debt and their loan supply. I further examine whether the amount of debt that was issued under the DGP has an effect. According to the results reported in Columns (3) and (4), the reduction in the outstanding long-term debt and the loan supply is not affected by the amount of debt that each TLGP bank issued under the DGP. 5. Conclusion This article analyzes the transmission of negative liquidity shocks to the real economy through a reduction in banks’ loan supply by using the 2007–09 financial crisis as a negative liquidity shock on banks. The ex ante heterogeneity in the amount of long-term debt that matured during the crisis is used to measure bank-specific exposure to the negative liquidity shock. The banks that had high amounts of long-term debt that matured during the crisis had a hard time finding affordable financing sources to cover their shortfalls due to the increased costs and financial frictions. As a result, 1 percentage point increase in the fraction of long-term debt that matured during the crisis led to a significant decrease in banks’ outstanding long-term debt by almost 0.04% of their total assets. This is transmitted to the asset side of the banks’ balance sheets by a reduction in their loan supply by 0.09% of their total assets. This reduction is particularly strong for real estate loans as compared with consumer loans whereas there is no significant effect on C&I loans. I further analyze the transmission of this liquidity shock to the real economy through the reduction in the real estate loans by examining house prices in the MSAs where affected banks have branches. My findings show that 1 percentage point increase in the weighted average fraction of long-term debt that matured during the crisis resulted in a 0.13 percentage point decrease in the growth rate of house prices. According to my results, the liquidity shock did not have any significant effect on well-capitalized banks’ loan supply, whereas the effect is stronger for under-capitalized banks. An under-capitalized bank with 1 percentage point increase in its long-term debt ratio that matured during the crisis experienced a reduction in its loan supply by 0.12% of its total assets. This finding is consistent with the literature which shows that financial frictions prevent under-capitalized banks from finding alternative financing sources to continue financing their loans during contraction times since these banks are perceived as riskier than well-capitalized banks (see, e.g., Kashyap and Stein, 2000; Kishan and Opiela, 2000; Meh and Moran, 2010). Moreover, the effect of the liquidity shock is stronger on house prices in the MSAs with more under-capitalized banks relative to MSAs with more well-capitalized banks although the difference is not significant. Further cross-sectional results show that the effect of the negative liquidity shock is stronger for banks with lower deposits and banks with higher short-term debt holdings. The effect disappears for banks with higher deposits and banks with lower short-term debt holdings. The former result is consistent with the recent literature that argues that banks with better access to deposit financing are more robust to liquidity shocks (see, e.g., Gatev and Strahan, 2006; Ivashina and Scharfstein, 2010; Cornett et al., 2011; Dagher and Kazimov, 2015). The latter result is consistent with the recent studies that show that banks with higher short-term debt ratios faced higher roll-over risk during the last financial crisis (see, e.g., Cornett et al., 2011; Huang and Ratnovski, 2011; Dagher and Kazimov, 2015). Overall, these results suggest that bank lending establishes a transmission channel where negative liquidity shocks are transmitted from banks’ balance sheets to the real economy, and that holding higher bank capital ratios, higher deposit ratios, and lower short-term debt ratios mitigates the transmission through this channel. Footnotes 1 Almeida et al. (2011) use this identification to analyze the effect of financial contracting on firms’ investment decisions. They find that firms whose long-term debt largely matured after the third quarter of 2007 cut their investment-to-capital ratio by 2.5 percentage points more than otherwise similar firms. 2 Long-term debt is defined as the debt with at least 1 year maturity. 3 Following Berger and Bouwman (2009), Flannery and Lin (2015), and Dursun-de Neef and Schandlbauer (2017), the amount of deposits that each bank holds in an MSA is used as a weight that reflects how important that bank is for that MSA. Accordingly, to calculate an MSA’s exposure to the negative liquidity shock, a deposits-weighted average fraction of long-term debt that matured during the crisis is calculated for each MSA. 4 For the falsification test in the robustness section, I use data for the time period from 2003 to 2006. 5 This includes people above 25 years old with one of the following degrees: associate, bachelor, master, professional, and doctorate. 6 The GDP in 2006 was $13.86 trillion which means that the largest bank in my sample held total assets that were 13.6% of the US GDP. 7 By using a continuous treatment, I study the effect of a change in the exposure to the negative liquidity shock instead of defining a binary variable for treated and control groups. This is similar to Acemoglu, Autor, and Lyle (2004). 8 I use bank controls that are fixed in the year 2006 instead of 1 year lagged ones to control for possible endogeneity due to the possible effects of the crisis on these bank characteristics. 9 When I repeat the analysis using the natural logarithm of total loans rather than normalizing it by total assets, the results are very similar as shown in Table V. 10 The reduction in the long-term debt is not significant with a p-value of 12%. 11 It is calculated as the long-term debt issued before the end of 2006 with a maturity from the beginning of 2008 until the end of 2009 divided by the long-term debt issued before the end of 2006 with a maturity from the beginning of 2008 for each bank in an MSA. Afterwards, the amount of deposits that each bank holds in this MSA is used as weights to calculate the weighted average fraction of long-term debt that matured during the crisis. 12 HMDA reports information on mortgages originated by banks and independent mortgage companies, so I can collect data on the volume of mortgage loans that are originated in an MSA in a given year rather than outstanding mortgage loans that each bank holds in that MSA. References Acemoglu D. , Autor D. H. , Lyle D. ( 2004 ): Women, war, and wages: the effect of female labor supply on the wage structure at midcentury , Journal of Political Economy 112 , 497 – 551 . Google Scholar CrossRef Search ADS Adelino M. , Schoar A. , Severino F. ( 2012 ): Credit supply and house prices: evidence from mortgage market segmentation. Technical report, National Bureau of Economic Research. Almeida H. , Campello M. , Laranjeira B. , Weisbenner S. ( 2011 ): Corporate debt maturity and the real effects of the 2007 credit crisis , Critical Finance Review 1 , 3 – 58 . Google Scholar CrossRef Search ADS Berger A. N. , Bouwman C. H. ( 2009 ): Bank liquidity creation , Review of Financial Studies 22 , 3779 – 3837 . Google Scholar CrossRef Search ADS Carlson M. , Shan H. , Warusawitharana M. ( 2013 ): Capital ratios and bank lending: a matched bank approach , Journal of Financial Intermediation 22 , 663 – 687 . Google Scholar CrossRef Search ADS Case K. E. , Shiller R. J. ( 1989 ): The efficiency of the market for single-family homes , American Economic Review 79 , 125 – 137 . Chava S. , Purnanandam A. ( 2011 ): The effect of banking crisis on bank-dependent borrowers , Journal of Financial Economics 99 , 116 – 135 . Google Scholar CrossRef Search ADS Cingano F. , Manaresi F. , Sette E. ( 2016 ): Does credit crunch investment down? New evidence on the real effects of the bank-lending channel , Review of Financial Studies 29 , 2737 – 2773 . Google Scholar CrossRef Search ADS Cornett M. M. , McNutt J. J. , Strahan P. E. , Tehranian H. ( 2011 ): Liquidity risk management and credit supply in the financial crisis , Journal of Financial Economics 101 , 297 – 312 . Google Scholar CrossRef Search ADS Dagher J. , Kazimov K. ( 2015 ): Banks’ liability structure and mortgage lending during the financial crisis , Journal of Financial Economics 116 , 565 – 582 . Google Scholar CrossRef Search ADS Dell’Ariccia G. , Detragiache E. , Rajan R. ( 2008 ): The real effect of banking crises , Journal of Financial Intermediation 17 , 89 – 112 . Google Scholar CrossRef Search ADS Di Maggio M. , Kermani A. ( 2017 ): Credit-induced boom and bust , Review of Financial Studies 30 , 3711 – 3758 . Google Scholar CrossRef Search ADS Dursun-de Neef H. Ö. , Schandlbauer A. ( 2017 ): The causal relationship between bank loans and bank capital. Unpublished working paper. Favara G. , Imbs J. ( 2015 ): Credit supply and the price of housing , American Economic Review 105 , 958 – 992 . Google Scholar CrossRef Search ADS Flannery M. J. , Lin L. ( 2015 ): House prices, bank balance sheets, and bank credit supply. Unpublished working paper. Gan J. ( 2007 ): The real effects of asset market bubbles: loan-and firm-level evidence of a lending channel , Review of Financial Studies 20 , 1941 – 1973 . Google Scholar CrossRef Search ADS Gatev E. , Strahan P. E. ( 2006 ): Banks’ advantage in hedging liquidity risk: theory and evidence from the commercial paper market , Journal of Finance 61 , 867 – 892 . Google Scholar CrossRef Search ADS Himmelberg C. , Mayer C. , Sinai T. ( 2005 ): Assessing high house prices: bubbles, fundamentals and misperceptions , Journal of Economic Perspectives 19 , 67 – 92 . Google Scholar CrossRef Search ADS Huang R. , Ratnovski L. ( 2011 ): The dark side of bank wholesale funding , Journal of Financial Intermediation 20 , 248 – 263 . Google Scholar CrossRef Search ADS Ivashina V. , Scharfstein D. ( 2010 ): Bank lending during the financial crisis of 2008 , Journal of Financial Economics 97 , 319 – 338 . Google Scholar CrossRef Search ADS Iyer R. , Peydro J.-L. ( 2011 ): Interbank contagion at work: evidence from a natural experiment , Review of Financial Studies 24 , 1337 – 1377 . Google Scholar CrossRef Search ADS Iyer R. , Peydró J.-L. , da Rocha-Lopes S. , Schoar A. ( 2014 ): Interbank liquidity crunch and the firm credit crunch: evidence from the 2007–2009 crisis , Review of Financial Studies 27 , 347 – 372 . Google Scholar CrossRef Search ADS Kashyap A. K. , Stein J. C. ( 2000 ): What do a million observations on banks say about the transmission of monetary policy?, American Economic Review 90 , 407 – 428 . Google Scholar CrossRef Search ADS Khwaja A. I. , Mian A. ( 2008 ): Tracing the impact of bank liquidity shocks: evidence from an emerging market , American Economic Review 98 , 1413 – 1442 . Google Scholar CrossRef Search ADS Kishan R. P. , Opiela T. P. ( 2000 ): Bank size, bank capital, and the bank lending channel , Journal of Money, Credit and Banking 32 , 121 – 141 . Google Scholar CrossRef Search ADS Lamont O. , Stein J. C. ( 1999 ): Leverage and house-price dynamics in US cities , Journal of Economic Perspectives 30 , 498 – 514 . Lopez J. A. ( 1999 ): Using CAMELS ratings to monitor bank conditions, Federal Reserve Bank of San Francisco Economic Letter No. 19. Meh C. A. , Moran K. ( 2010 ): The role of bank capital in the propagation of shocks , Journal of Economic Dynamics and Control 34 , 555 – 576 . Google Scholar CrossRef Search ADS Mian A. , Sufi A. ( 2009 ): The consequences of mortgage credit expansion: evidence from the US mortgage default crisis , Quarterly Journal of Economics 124 , 1449 – 1496 . Google Scholar CrossRef Search ADS Ongena S. , Smith D. C. , Michalsen D. ( 2003 ): Firms and their distressed banks: lessons from the Norwegian banking crisis , Journal of Financial Economics 67 , 81 – 112 . Google Scholar CrossRef Search ADS Paravisini D. ( 2008 ): Local bank financial constraints and firm access to external finance , Journal of Finance 63 , 2161 – 2193 . Google Scholar CrossRef Search ADS Peek J. , Rosengren E. S. ( 2000 ): Collateral damage: effects of the Japanese bank crisis on real activity in the United States , American Economic Review 90 , 30 – 45 . Google Scholar CrossRef Search ADS Puri M. , Rocholl J. , Steffen S. ( 2011 ): Global retail lending in the aftermath of the US financial crisis: distinguishing between supply and demand effects , Journal of Financial Economics 100 , 556 – 578 . Google Scholar CrossRef Search ADS Roberts M. R. , Whited T. M. ( 2013 ): Endogeneity in empirical corporate finance , Handbook of the Economics of Finance 2 , 493 – 572 . Google Scholar CrossRef Search ADS Schnabl P. ( 2012 ): The international transmission of bank liquidity shocks: evidence from an emerging market , Journal of Finance 67 , 897 – 932 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the European Finance Association. All rights reserved. For Permissions, please email: journals.permissions@oup.com

Journal

Review of FinanceOxford University Press

Published: Jan 24, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off