Do Exposures to Sagging Real Estate, Subprime, or Conduits Abroad Lead to Contraction and Flight to Quality in Bank Lending at Home?

Do Exposures to Sagging Real Estate, Subprime, or Conduits Abroad Lead to Contraction and Flight... Abstract We investigate how differential exposures by German banks to the US real-estate market affect domestic lending in Germany when home prices started to decline in the USA. We find that banks with an exposure to the US real-estate sector and to conduits shift their domestic lending to industry–region combinations with lower insolvency ratios following a decrease in US home prices. These banks also contract their lending to German firms more than banks that do not have such exposure. We mainly document that possible losses abroad shift bank lending at home where the size of the effect depends on the type and the degree of exposure the bank has. 1. Introduction By mid-2006 real-estate prices in the US began to plummet, triggering the US subprime mortgage crisis that led to a global financial crisis. While the main focus was on the fragility of the financial system (and, to some extent, the regulatory focus is still on how to re-establish a healthy banking industry), the initial policy reaction relied mostly on monetary tools. These measures proved to be not entirely effective due to the presence of excessive household debt coupled with decreasing home prices. The crisis may not have been as severe, it is often argued now, if those underlying problems had have been addressed in a first and foremost step (Mian and Sufi, 2014).1 German banks, too, experienced considerable loan losses and given solvency considerations consequently more binding capital constraints. This was largely attributed to their various exposures to the US real-estate market. In addition to their direct lending to US firms in the real-estate sector and to major subprime lenders, German banks also became exposed by providing liquidity support in the form of credit lines to their asset-backed commercial paper (ABCP) conduits. The first banks that had to be bailed out by a government during the financial crisis were actually two German banks, IKB Deutsche Industriebank, and Sachsen Landesbank. The deteriorating quality of their assets and the panic on the ABCP market forced these German banks to write off the liquidity lines they had provided to their ABCP conduits. These write-offs resulted in considerable losses on their balance sheets. In general, the Landesbanks’ substantial exposures to US mortgage-backed securities through their ABCP conduits, which were higher than those of the big German banks, led to the collapse or bail-out of various Landesbanks. Given these differences, we therefore investigate how each type of exposure in the US real-estate market influenced domestic lending in Germany. We first suitably document the overall contraction in lending that occurred in Germany, then focus on the heterogeneity in the contraction taking place across banks and firms. We are mainly interested in studying whether—when home prices started to decline in the USA—differences in bank exposures to the US real-estate market started to determine bank lending in Germany according to firm risk.2 In other words: Is there a flight to quality in domestic lending depending on the degree of the German banks’ exposures in the USA—following the decline in US home prices? In terms of credit volume we find that banks that had a direct exposure to the US real-estate sector contracted their lending in Germany by more following a decrease in US home prices than banks that did not have such exposure. This effect is both statistically significant and economically relevant. For example, a bank with a €1 billion exposure to the US real-estate sector, and following a decrease by five-index points (which is equal to two standard deviations) in the S&P/Case–Shiller US National Home Price Index, is estimated to contract its quarterly lending in Germany by 1.19 percentage points more than a bank with no such exposure. This is a large effect given that the mean (median) quarterly loan growth during the sample period equals –2.48 (–0.71)%. And a bank with €1 billion conduit exposure is estimated to contract its quarterly lending in Germany by 1.47 percentage points more than banks without conduits in place following a decrease by five-index points in home prices. In terms of credit composition, we find clear evidence of a flight to quality. For example, a bank with a €1 billion exposure to the US real-estate sector, and following a decrease by five-index points in the S&P/Case–Shiller US National Home Price Index, is estimated to contract its quarterly lending in Germany to firms in riskier industry–region combinations (i.e., those with a 1 percentage point higher insolvency rate) by 4.01 percentage points more than a bank with no such exposure. A bank with €1 billion conduit exposure is estimated to contract its quarterly lending to such riskier industry–region pairs by 3.43 percentage points more than banks without conduits in place when US home prices decrease by five-index points. Overall, these findings imply that possible losses abroad may not only cut but also shift bank lending at home. Our results thereby vividly demonstrate how the recent globalization in banking activities may be inevitably linked to financial stability. In particular, economic fluctuations in one country are transmitted to other countries through this bank exposure channel. As the largest European economy and as a bank-based system, where bank financing plays a crucial role for corporations of all sizes, Germany is particularly interesting to pursue this analysis. These characteristics of the financial system enable us to focus on how affected banks change their lending toward domestic firms, which has an influence on the real economy as well. We are clearly not the first to study (and for identification purposes “exploit”) the international transmission of shocks through the banking sector. However, ours is the first paper to investigate the type and the level of the exposures to the US real-estate market by non-US banks and how these exposures influence both the volume and the composition of local lending depending on the US home price shock. In terms of identification, our approach is superior to extant work on this topic because we use the actual real-estate exposures to the USA, which is “ground zero” for the financial crisis, and because we focus on the initial exogenous shock in the US home prices in combination with those “ground zero” exposures. Additionally, we study lending taking place in another country (different from the crisis country) with the information from a comprehensive credit register providing strong controls for local credit demand. Our research follows the seminal work by Peek and Rosengren (1997) and Peek and Rosengren (2000) who show that when parent banks are faced with a (funding) shock, this can negatively affect lending by their foreign affiliates.3 In particular, Peek and Rosengren (1997) identify a supply shock to bank lending in the USA through US branches of Japanese banks, which was caused by the decline in Japanese stock market values. Unlike the previous studies trying to examine the relationship between capital ratios and the overall volume of lending, their study actually disentangles loan supply from loan demand by focusing on the transmission of the capital effects of the Japanese stock market declines. Their paper documents that the Japanese banks’ capital ratios significantly determine its commercial and industrial lending in the USA. Using similar data, Peek and Rosengren (2000) investigate the change in commercial real-estate loans in spatially separated markets, which enables them to examine the impact of this loan supply shock on the real economic activity in the USA. A recent paper (and closest to ours) is a paper by Puri, Rocholl, and Steffen (2011). They investigate the impact of the financial crises on the credit supply of German savings banks. Here, the transmission occurs through savings banks’ holdings in Landesbanks that were exposed to subprime mortgages.4 This mechanism is somewhat different from the one in Peek and Rosengren since an external financial shock is transmitted to a domestic market through the exposure of the domestic banks. For that reason, it becomes even more important and potentially more difficult to isolate the loan supply effect. Puri, Rocholl, and Steffen (2011) disentangle supply from demand effects by employing information coming from the loan application process. The authors find that affected savings banks reject more loan applications than non-affected banks. However, close bank–firm relationships help to mitigate the loan supply shock. A number of features distinguish our paper from theirs. First, having access to unique and confidential supervisory data, we know the actual time-varying exposures to the US real-estate market (direct lending to the US real-estate sector, to major subprime lenders, and conduit exposures) of all German banks. Combined with the US home price shock, these exposures to the US real-estate market allow us to identify possible (but at the time not yet publicly observable) bank losses. In contrast, Puri, Rocholl, and Steffen (2011) rely on ex post publicly reported distress at three Landesbanks that led to a decline in the value of equity held by savings banks present in their loan data set. Thus, our data set enables us to assess the time-varying effects of various types of German bank exposures to the US real-estate market throughout the entire crisis period, not just the presumed (though plausible) impact through indirect linkages within parts of the German banking system.5 Second, to identify the changes in lending we rely on credit register data that cover the entire banking sector in Germany, not just savings banks. We match this data with bank and market characteristics. Finally, we examine the resultant changes not only in the aggregate volume but also in the composition of bank lending in Germany across corporations, industries, and regions as we want to investigate whether there is a “flight to quality” in lending in Germany for those banks that were more exposed to the US real-estate market. In this respect, our paper also contributes to an extant literature that examines the flight to quality or loan strictness following negative shocks affecting banks (Lang and Nakamura, 1995; Bernanke, Gertler, and Gilchrist, 1996; Murfin, 2012; Becker and Ivashina, 2016), or documents bank risk-taking following expansionary monetary shocks (e.g., Jiménez et al., 2014; Ioannidou, Ongena, and Peydró, 2015; Dell’Ariccia, Laeven, and Suarez, 2017). These also include studies on bank security portfolios during the crisis where a similar pattern is observed through the purchase of high-quality assets by banks. Hildebrand, Rocholl, and Schulz (2012), for example, investigate banks’ investment strategies in Germany and find that banks prefer to purchase securities that are eligible as collateral when borrowing from the ECB. This behavior is more pronounced for less healthy or large banks, and for banks that are exposed to Greek bonds. Beber, Brandt, and Kavajecz (2009), on the other hand, examine whether investment decisions are driven by quality or liquidity concerns, and they find that during distressed times, investors prefer liquidity rather than credit quality. In that respect, both papers point to the importance of liquidity in investment decisions as well. However, changes in lending behavior are more likely to be driven by default risk only, especially if securitization is not an option. This feature distinguishes our paper from that strand of literature. The remainder of our paper is organized as follows. In Section 2, we discuss the various exposures German banks have in the USA and their lending to firms in Germany. In Section 3, we describe the data and the definition of the variables of interest. We discuss the methodology and present the main estimation results in Section 4. Section 5 concludes. 2. Portfolios of German Banks ABCP conduits set up by German banks performed maturity transformation by purchasing long-term assets and issuing ABCP, a short-term debt instrument which is often used to raise capital. The ABCP were primarily sold to money market funds (MMFs) and rolled over at regular intervals. Nevertheless, ABCP conduits used to be off-balance sheet vehicles and represented the agents of the “shadow banking” market, which appeared to be less regulated. Therefore, German banks could hold assets in their ABCP conduits without providing a sufficient amount of the required capital. ABCP conduits were designed to protect investors from declines in the market value of the underlying assets. Sponsoring banks provided liquidity support to their ABCP conduits. According to Moody’s (2007): “Most programmes have 100% committed liquidity support that can be drawn to repay ABCP up to the par value of non-defaulted assets, regardless of market value.” In cases where ABCP conduits experienced difficulties, credit risk attributed to the ABCP conduits effectively put a strain on their parent banks. Being a safe haven for investors before the crisis, ABCP conduits played a central role in the financial crisis 2007–09 when news about the deteriorating quality of US subprime mortgages roiled the financial markets and the market for ABCP froze with risk-averse investors being unwilling to purchase and roll over maturing ABCP (Acharya and Schnabl, 2010; Kacperczyk and Schnabl, 2010).6,7 At the end of July 2007, the US subprime mortgage crisis reached its first victim in Germany. On July 31, 2007, Frankfurter Allgemeine indeed reported about the failure of the IKB Deutsche Industriebank AG—a bank which financed mainly medium-sized enterprises. IKB failed on a large credit line provided to its conduit Rhineland Funding Capital which amounted to €12.7 billion. The IKB failure unfortunately did not remain an isolated incident. In the mid of August 2007 the next bank—SachenLB—made the headlines in the newspapers. On the August 22, 2007, the Financial Times wrote: “SachsenLB, a publicly-owned Landesbank, or a state bank, became the second German financial institution in three weeks to be forced to accept an emergency rescue, with fellow savings banks taking over a €17.3 billion credit facility.” As quoted also by the other newspapers such as Franfurter Allgemeine on August 10 and later by Handelsblatt on the November 9 the twenty-four biggest German bank SachsenLB had been running one of the world biggest conduits called Ormond Quay. The financing needs of SachsenLB raised to €17.5 billion and exceeded its own capital by more than ten times. The Financial Times titled its article “‘Not uncritical’ Subprime exposure drags down German banks” and characterized the situation in Germany as the following: “The crisis focuses on German public banks, in particular the Landesbanks … .”8 As the reasons for the Landesbanks to be involved in such risky activities the Financial Times pointed on the one hand to the effects of rising competition and capital requirements and on the other hand to the declining earnings opportunities in the home markets. Between August 2007 and the Lehman Brothers’ bankruptcy in September 2008, the ABCP market was seriously stressed. In this period, the total value of ABCP outstanding decreased by 37%, from $1.18 trillion to $745 billion. However, the cost of issuing overnight ABCP relative to the US Federal Reserve Funds rate also jumped from 10 to 150 basis points after the news of the withdrawals from BNP Paribas. On September 16, 2008, the Reserve Primary Fund—a large MMF—announced considerable losses on its holdings of Lehman Brothers’ CP. This, in turn triggered a run on the MMF industry and led to the reduction of holdings of all types of CP by MMFs. After Lehman Brothers’ collapse, German banks, which were already weakened by the need to meet their obligations on maturing ABCP, came under further pressure. Credit growth in Germany has been characterized by various fluctuations since 2002. The annual growth rate of lending to domestic firms rose sharply from –0.1% in May 2007 to 3.8% in July 2008, whereas a remarkable drop was observed starting only from the third quarter of 2008 onwards (Deutsche Bundesbank, Monthly Report, 9/2009). The annual growth in lending declined by 2.7 percentage points to 1.1% between July 2008 and July 2009. The rest of the euro area, on the other hand, witnessed a sharper decline in lending than Germany, and also at an earlier point in time—at the end of 2007. The slowdown in lending is found to be situated in the non-financial industry. Growth in lending declined sharply especially for those banking groups that were hit particularly hard by the global financial crisis. 3. Data and Variables 3.1 Data Sources We employ a unique matched firm–bank level dataset that contains quarterly information from the first quarter of 2005 to the fourth quarter of 2009. The data combine five databases: (1) the Deutsche Bundesbank’s credit register (MiMik), (2) Moody’s ABCP Query, (3) bank balance sheet data (Bista, BAKIS), (4) regional firm insolvencies per industry, and (5) home price indices (S&P/Case–Shiller). These data sources make it possible to observe the individual lending behavior of German banks to domestic firms, and to combine this information with the exposure of German banks to the real-estate sector in the USA, to subprime lenders in the USA, and to their conduits engaged in the US market, as well as to use the firm- and bank-specific information. The credit register (MiMik) is the main data source for the individual exposures of German banks to firms.9 The credit register contains information on large exposures of €1.5 million (formerly 3 million Deutsche Mark) and above.10 Therefore, exposures to small- and medium-sized firms might be underrepresented in this database. However, if the sum of the exposures to firms in a “borrower unit”, that is, a group of affiliated companies, exceeds the threshold of €1.5 million, the individual exposure to a firm in that group is reported, even if it is a very small exposure. This reporting partly abates the extant bias in the credit register toward medium- and large-sized firms. Moreover, large firms play a relatively more important role in our analysis since they represent a larger portion of the borrowing in total. At individual level, large firms potentially borrow from large banks in large volumes that may be adjusted swiftly. Therefore, we believe that the threshold helps to make our control group more comparable to the group of banks that were exposed to the US real-estate market, in effect similar to a matching exercise. Bank exposures to firms in the credit register are defined fairly broadly, for example, they include not only corporate loans but also corporate bonds and other securities which belong to the non-trading stock.11 In the credit register we are able to distinguish between on-balance sheet and off-balance sheet items.12 We choose to use only on-balance sheet positions, since the inclusion of off-balance sheet exposures leads to an overstatement of the actual exposures due to guarantees provided by banks (to other banks for exposures that were already covered in on-balance sheet items). Based on individual bank exposures to firms, banks, and other financial institutions, the credit register covers both domestic and foreign exposures and contains the information on country code and industry classification within a particular country. This structure of the credit register allows us to identify both individual bank exposures to the real-estate sector as well as to the top subprime lenders in the USA, and enables us to study the impact of these exposures on the lending of German banks to domestic firms. The credit register contains also information on firm quality, however for restricted periods in the sample. Therefore, we augment the available information with the industry-level number of firm defaults within particular German Federal States from the Federal Statistical Office. This information on the industry number of firm defaults (insolvency) makes it possible to differentiate between the lending behavior of German banks toward firms with high and low quality. The second database we use is the Moody’s ABCP Query. From this database, we take the information on all important conduits of German banks and on the amount of liquidity provision of German banks to their conduits. However, this information is available in the Moody’s ABCP Query only starting in 2007. For 2005 and 2006, we hand-collect this information from two Moody’s publications: “A Performance Overview for EMEA ABCP conduits” and “A Program Review for US conduits”. This information allows us to study the impact of the German bank exposure to their conduits on lending behavior to domestic firms. Similar to the real-estate exposure in the USA and exposure to the subprime lenders in the USA (taken from the credit register), the German bank exposure to their conduits is quarterly and is aggregated at the bank level. We borrow the majority of our bank-specific variables from the monthly balance sheet statistics (Bista) and some of our bank-specific variables from BAKIS. BAKIS is the Information System, which is shared between the Deutsche Bundesbank and BaFin (the German Federal Banking Supervisory Office). It contains the bank balance sheets for all German banks. We select the monthly balance sheet statistics and match them with the credit register on a quarterly basis. Some risk indicators, such as, for example, non-performing loans, are not available in Bista and we extract them on a yearly basis from BAKIS. Finally, to capture the price developments in the US real-estate market, we access the S&P/Case–Shiller Home Price Indices. 3.2 Sample Over the time period from the first quarter of 2005 to the fourth quarter of 2009, we consider 2,031 banks that provide domestic balance-sheet loans to 336,990 firms in Germany. In total, we have 3.9 million bank–firm–quarter observations of domestic on-balance-sheet lending. However, a number of bank mergers took place during this time period. We carry out a merger correction procedure by creating a new separate bank after the merger.13 Of the 2,031 banks involved, 90 banks have direct exposure to the US real-estate sector, 142 banks have direct exposure to subprime lenders, and 20 banks have conduit exposure. In a final step we match the datasets discussed in Section 3.1, and we end up with 1,664,262 bank–firm–quarter observations available for the whole set of variables for our main model. Tables I and II contain variable definitions and their summary statistics. The latter suggests that the sample selection is fortunately minimal. It is also worth noting that the change in log domestic lending has negative mean and median values: –0.025 and –0.007, respectively. In Figure 1 Panels A and B, we present the mean values of bank exposures, US home prices, and domestic lending growth over time. We observe that as home prices decrease lending growth in Germany contracts as well, which is more pronounced for banks that were exposed to the US real-estate sector. Table I. Variable definitions This table provides the variable name, definition, measurement, and data source. All variables are quarterly except ROA and NPL are available on a yearly basis. Variable name  Definition and measurement  Data source  Bank–firm level variable   ΔLog domestic lending  The change in the natural logarithm of domestic lending from time t–1 to time t by bank i to firm j  Bundesbank Credit Register  Bank-level variables   Log US real-estate exposure  Natural logarithm of total direct exposure to the US real-estate sector  Bundesbank Credit Register   Log US subprime exposure  Natural logarithm of total direct exposure to subprime lenders  Bundesbank Credit Register   Log conduit exposure  Natural logarithm of liquidity provided to asset-backed commerical paper (ABCP) conduits  Moody’s   Size  Natural logarithm of total assets  Bundesbank Bank Balance Sheet Data   Capital  Capital ratio  Bundesbank Bank Balance Sheet Data   Liquidity  (Cash + balances with central banks + securities) to total assets  Bundesbank Bank Balance Sheet Data   ROA  Return on assets  Bundesbank Bank Balance Sheet Data   NPL  Non-performing loans to total loans  Bundesbank Bank Balance Sheet Data   Deposits  Deposits to total liabilities  Bundesbank Bank Balance Sheet Data   CB funding  Central bank funding to total assets  Bundesbank Bank Balance Sheet Data  Macro variable   ΔUS homeprices  The change in the national US home price index  S&P/Case–Shiller Home Price Indices  Firm industry–region variable   Insolvency  The number of insolvencies divided by the total number of firms in an industry and region  Destatis Insolvency Data and Turnover Tax Statistics  Variable name  Definition and measurement  Data source  Bank–firm level variable   ΔLog domestic lending  The change in the natural logarithm of domestic lending from time t–1 to time t by bank i to firm j  Bundesbank Credit Register  Bank-level variables   Log US real-estate exposure  Natural logarithm of total direct exposure to the US real-estate sector  Bundesbank Credit Register   Log US subprime exposure  Natural logarithm of total direct exposure to subprime lenders  Bundesbank Credit Register   Log conduit exposure  Natural logarithm of liquidity provided to asset-backed commerical paper (ABCP) conduits  Moody’s   Size  Natural logarithm of total assets  Bundesbank Bank Balance Sheet Data   Capital  Capital ratio  Bundesbank Bank Balance Sheet Data   Liquidity  (Cash + balances with central banks + securities) to total assets  Bundesbank Bank Balance Sheet Data   ROA  Return on assets  Bundesbank Bank Balance Sheet Data   NPL  Non-performing loans to total loans  Bundesbank Bank Balance Sheet Data   Deposits  Deposits to total liabilities  Bundesbank Bank Balance Sheet Data   CB funding  Central bank funding to total assets  Bundesbank Bank Balance Sheet Data  Macro variable   ΔUS homeprices  The change in the national US home price index  S&P/Case–Shiller Home Price Indices  Firm industry–region variable   Insolvency  The number of insolvencies divided by the total number of firms in an industry and region  Destatis Insolvency Data and Turnover Tax Statistics  Table I. Variable definitions This table provides the variable name, definition, measurement, and data source. All variables are quarterly except ROA and NPL are available on a yearly basis. Variable name  Definition and measurement  Data source  Bank–firm level variable   ΔLog domestic lending  The change in the natural logarithm of domestic lending from time t–1 to time t by bank i to firm j  Bundesbank Credit Register  Bank-level variables   Log US real-estate exposure  Natural logarithm of total direct exposure to the US real-estate sector  Bundesbank Credit Register   Log US subprime exposure  Natural logarithm of total direct exposure to subprime lenders  Bundesbank Credit Register   Log conduit exposure  Natural logarithm of liquidity provided to asset-backed commerical paper (ABCP) conduits  Moody’s   Size  Natural logarithm of total assets  Bundesbank Bank Balance Sheet Data   Capital  Capital ratio  Bundesbank Bank Balance Sheet Data   Liquidity  (Cash + balances with central banks + securities) to total assets  Bundesbank Bank Balance Sheet Data   ROA  Return on assets  Bundesbank Bank Balance Sheet Data   NPL  Non-performing loans to total loans  Bundesbank Bank Balance Sheet Data   Deposits  Deposits to total liabilities  Bundesbank Bank Balance Sheet Data   CB funding  Central bank funding to total assets  Bundesbank Bank Balance Sheet Data  Macro variable   ΔUS homeprices  The change in the national US home price index  S&P/Case–Shiller Home Price Indices  Firm industry–region variable   Insolvency  The number of insolvencies divided by the total number of firms in an industry and region  Destatis Insolvency Data and Turnover Tax Statistics  Variable name  Definition and measurement  Data source  Bank–firm level variable   ΔLog domestic lending  The change in the natural logarithm of domestic lending from time t–1 to time t by bank i to firm j  Bundesbank Credit Register  Bank-level variables   Log US real-estate exposure  Natural logarithm of total direct exposure to the US real-estate sector  Bundesbank Credit Register   Log US subprime exposure  Natural logarithm of total direct exposure to subprime lenders  Bundesbank Credit Register   Log conduit exposure  Natural logarithm of liquidity provided to asset-backed commerical paper (ABCP) conduits  Moody’s   Size  Natural logarithm of total assets  Bundesbank Bank Balance Sheet Data   Capital  Capital ratio  Bundesbank Bank Balance Sheet Data   Liquidity  (Cash + balances with central banks + securities) to total assets  Bundesbank Bank Balance Sheet Data   ROA  Return on assets  Bundesbank Bank Balance Sheet Data   NPL  Non-performing loans to total loans  Bundesbank Bank Balance Sheet Data   Deposits  Deposits to total liabilities  Bundesbank Bank Balance Sheet Data   CB funding  Central bank funding to total assets  Bundesbank Bank Balance Sheet Data  Macro variable   ΔUS homeprices  The change in the national US home price index  S&P/Case–Shiller Home Price Indices  Firm industry–region variable   Insolvency  The number of insolvencies divided by the total number of firms in an industry and region  Destatis Insolvency Data and Turnover Tax Statistics  Table II. Descriptive statistics for the observations used in the regressions This table provides the mean, standard deviation, 10th percentile, 25th percentile, median, 75th percentile, and 90th percentile for all variables for the 1,664,262 observations used in the regressions. ΔLog domestic lending t is corrected for outliers at the one percentile level. Variable  Unit  Mean  Standard deviation  10th percentile  25th percentile  Median  75th percentile  90th percentile  All banks   Domestic lendingt  €million  3.83  14.67  0.11  0.49  1.58  3.24  7.58   ΔLog domestic lendingt  Logarithmic change  –0.025  0.341  –0.164  –0.036  –0.007  0.000  0.096   US real-estate exposuret−1  €million  370.30  981.40  0.00  0.00  0.00  86.74  1,309.28   Log US real-estate exposuret−1  Logarithm  6.75  9.30  0.00  0.00  0.00  18.28  20.99   US subprime exposuret−1  €million  31.34  115.65  0.00  0.00  0.00  1.00  63.24   Log US subprime exposuret−1  Logarithm  4.43  7.58  0.00  0.00  0.00  13.82  17.96   Conduit exposuret−1  €million  1,469.02  4,536.33  0.00  0.00  0.00  0.00  4,472.07   Log conduit exposuret−1  Logarithm  5.02  9.18  0.00  0.00  0.00  0.00  22.22   Total assetst−1  €million  88,691.14  160,680.90  857.53  2,081.79  7,388.21  137,105.90  307,803.90   Sizet−1  Logarithm  23.25  2.23  20.57  21.46  22.72  25.64  26.45   Capitalt−1  Percentage  14.15  3.50  10.30  11.67  13.60  15.78  18.83   Liquidityt−1  Percentage  7.98  5.35  2.08  4.03  7.12  10.69  14.92   ROAt−1  Percentage  0.77  0.63  0.25  0.51  0.74  1.02  1.25   NPLt−1  Percentage  4.55  3.56  1.23  2.40  4.04  6.03  8.19   Depositst−1  Percentage  23.75  15.24  0.61  9.98  25.97  35.19  42.59   CB fundingt−1  Percentage  1.61  2.58  0.00  0.00  0.00  2.39  5.20   ΔUS homepricest  Index change (%)  –2.24  2.45  –5.22  –4.86  –2.6  –0.06  1.40   Insolvencyt−1  Percentage  0.24  0.21  0.06  0.11  0.20  0.31  0.46  Variable  Unit  Mean  Standard deviation  10th percentile  25th percentile  Median  75th percentile  90th percentile  All banks   Domestic lendingt  €million  3.83  14.67  0.11  0.49  1.58  3.24  7.58   ΔLog domestic lendingt  Logarithmic change  –0.025  0.341  –0.164  –0.036  –0.007  0.000  0.096   US real-estate exposuret−1  €million  370.30  981.40  0.00  0.00  0.00  86.74  1,309.28   Log US real-estate exposuret−1  Logarithm  6.75  9.30  0.00  0.00  0.00  18.28  20.99   US subprime exposuret−1  €million  31.34  115.65  0.00  0.00  0.00  1.00  63.24   Log US subprime exposuret−1  Logarithm  4.43  7.58  0.00  0.00  0.00  13.82  17.96   Conduit exposuret−1  €million  1,469.02  4,536.33  0.00  0.00  0.00  0.00  4,472.07   Log conduit exposuret−1  Logarithm  5.02  9.18  0.00  0.00  0.00  0.00  22.22   Total assetst−1  €million  88,691.14  160,680.90  857.53  2,081.79  7,388.21  137,105.90  307,803.90   Sizet−1  Logarithm  23.25  2.23  20.57  21.46  22.72  25.64  26.45   Capitalt−1  Percentage  14.15  3.50  10.30  11.67  13.60  15.78  18.83   Liquidityt−1  Percentage  7.98  5.35  2.08  4.03  7.12  10.69  14.92   ROAt−1  Percentage  0.77  0.63  0.25  0.51  0.74  1.02  1.25   NPLt−1  Percentage  4.55  3.56  1.23  2.40  4.04  6.03  8.19   Depositst−1  Percentage  23.75  15.24  0.61  9.98  25.97  35.19  42.59   CB fundingt−1  Percentage  1.61  2.58  0.00  0.00  0.00  2.39  5.20   ΔUS homepricest  Index change (%)  –2.24  2.45  –5.22  –4.86  –2.6  –0.06  1.40   Insolvencyt−1  Percentage  0.24  0.21  0.06  0.11  0.20  0.31  0.46  Table II. Descriptive statistics for the observations used in the regressions This table provides the mean, standard deviation, 10th percentile, 25th percentile, median, 75th percentile, and 90th percentile for all variables for the 1,664,262 observations used in the regressions. ΔLog domestic lending t is corrected for outliers at the one percentile level. Variable  Unit  Mean  Standard deviation  10th percentile  25th percentile  Median  75th percentile  90th percentile  All banks   Domestic lendingt  €million  3.83  14.67  0.11  0.49  1.58  3.24  7.58   ΔLog domestic lendingt  Logarithmic change  –0.025  0.341  –0.164  –0.036  –0.007  0.000  0.096   US real-estate exposuret−1  €million  370.30  981.40  0.00  0.00  0.00  86.74  1,309.28   Log US real-estate exposuret−1  Logarithm  6.75  9.30  0.00  0.00  0.00  18.28  20.99   US subprime exposuret−1  €million  31.34  115.65  0.00  0.00  0.00  1.00  63.24   Log US subprime exposuret−1  Logarithm  4.43  7.58  0.00  0.00  0.00  13.82  17.96   Conduit exposuret−1  €million  1,469.02  4,536.33  0.00  0.00  0.00  0.00  4,472.07   Log conduit exposuret−1  Logarithm  5.02  9.18  0.00  0.00  0.00  0.00  22.22   Total assetst−1  €million  88,691.14  160,680.90  857.53  2,081.79  7,388.21  137,105.90  307,803.90   Sizet−1  Logarithm  23.25  2.23  20.57  21.46  22.72  25.64  26.45   Capitalt−1  Percentage  14.15  3.50  10.30  11.67  13.60  15.78  18.83   Liquidityt−1  Percentage  7.98  5.35  2.08  4.03  7.12  10.69  14.92   ROAt−1  Percentage  0.77  0.63  0.25  0.51  0.74  1.02  1.25   NPLt−1  Percentage  4.55  3.56  1.23  2.40  4.04  6.03  8.19   Depositst−1  Percentage  23.75  15.24  0.61  9.98  25.97  35.19  42.59   CB fundingt−1  Percentage  1.61  2.58  0.00  0.00  0.00  2.39  5.20   ΔUS homepricest  Index change (%)  –2.24  2.45  –5.22  –4.86  –2.6  –0.06  1.40   Insolvencyt−1  Percentage  0.24  0.21  0.06  0.11  0.20  0.31  0.46  Variable  Unit  Mean  Standard deviation  10th percentile  25th percentile  Median  75th percentile  90th percentile  All banks   Domestic lendingt  €million  3.83  14.67  0.11  0.49  1.58  3.24  7.58   ΔLog domestic lendingt  Logarithmic change  –0.025  0.341  –0.164  –0.036  –0.007  0.000  0.096   US real-estate exposuret−1  €million  370.30  981.40  0.00  0.00  0.00  86.74  1,309.28   Log US real-estate exposuret−1  Logarithm  6.75  9.30  0.00  0.00  0.00  18.28  20.99   US subprime exposuret−1  €million  31.34  115.65  0.00  0.00  0.00  1.00  63.24   Log US subprime exposuret−1  Logarithm  4.43  7.58  0.00  0.00  0.00  13.82  17.96   Conduit exposuret−1  €million  1,469.02  4,536.33  0.00  0.00  0.00  0.00  4,472.07   Log conduit exposuret−1  Logarithm  5.02  9.18  0.00  0.00  0.00  0.00  22.22   Total assetst−1  €million  88,691.14  160,680.90  857.53  2,081.79  7,388.21  137,105.90  307,803.90   Sizet−1  Logarithm  23.25  2.23  20.57  21.46  22.72  25.64  26.45   Capitalt−1  Percentage  14.15  3.50  10.30  11.67  13.60  15.78  18.83   Liquidityt−1  Percentage  7.98  5.35  2.08  4.03  7.12  10.69  14.92   ROAt−1  Percentage  0.77  0.63  0.25  0.51  0.74  1.02  1.25   NPLt−1  Percentage  4.55  3.56  1.23  2.40  4.04  6.03  8.19   Depositst−1  Percentage  23.75  15.24  0.61  9.98  25.97  35.19  42.59   CB fundingt−1  Percentage  1.61  2.58  0.00  0.00  0.00  2.39  5.20   ΔUS homepricest  Index change (%)  –2.24  2.45  –5.22  –4.86  –2.6  –0.06  1.40   Insolvencyt−1  Percentage  0.24  0.21  0.06  0.11  0.20  0.31  0.46  Figure 1. View largeDownload slide (A) Mean bank exposures over time. The figure displays the mean values for US real-estate exposure, US subprime exposure, and conduit exposure in €million (in logarithmic scale). (B) US home price changes and lending growth in Germany. Domestic lending growth exposed banks and unexposed banks relate to the right axis, in percent, while the Shiller index (with base year 2000) relates to the left axis, in index units. Figure 1. View largeDownload slide (A) Mean bank exposures over time. The figure displays the mean values for US real-estate exposure, US subprime exposure, and conduit exposure in €million (in logarithmic scale). (B) US home price changes and lending growth in Germany. Domestic lending growth exposed banks and unexposed banks relate to the right axis, in percent, while the Shiller index (with base year 2000) relates to the left axis, in index units. 3.3 Bank Exposure to Real Estate, Subprime, and Conduits German banks were engaged in at least three ways in the US real-estate market: Some banks had direct (regular) exposures to the US real-estate sector on their balance sheets, some banks had exposures to subprime lenders in the USA, and some banks had ABCP conduits in place.14 Table III gives a detailed overview of German banks’ portfolios based on the credit register data by distinguishing between on- and off-balance sheet exposures in 2007. The upper panel of the table provides the regional distribution of exposures to the real-estate sector, including Greece, Ireland, Portugal, Spain, offshore centers, the USA, and all foreign countries.15 We can clearly see that German banks’ exposure to US firms in the real-estate sector was significantly higher than exposures to any of the other countries’ firms in the same sector. Whereas US real-estate exposure exceeds €30 billion, the second highest exposure, which is to offshore centers, totals only €13.6 billion, while the third highest, which is to Spain, equals €10.6 billion. Moreover, US real-estate exposure was potentially the first to be struck by a collapse in home prices and this exposure will therefore experience the “cleanest” (identifiable) shock to real-estate during the crisis. We also note that we ended our sample period in 2009, in order to have a cleaner setting and not to include the beginning of the sovereign debt crisis. Table III. Regional Distribution for Exposures by German Banks in 2007Q2 This table provides the real-estate exposures taken from the Bundesbank credit register for Greece, Ireland, Portugal, Spain, offshore centers, the USA, and for all foreign countries in 2007Q2 for all German banks, including the big banks and the Landesbanks. Additionally, it provides the exposure toward banks in the USA, the exposure toward the offshore centers, and the banks’ total portfolio. The banks’ total portfolio does not comprise exposures to international organizations. The table lists the total exposure, the balance-sheet exposure, the off-balance-sheet exposure, and the derivatives calculated as the sum of all German banks in millions of euro. For offshore centers, the Bundesbank definition is used. According to this definition, offshore centers include twenty-three countries. Type of exposure  Unit  Total exposure  Balance-sheet exposure  Off-balance-sheet exposure  Derivatives  Real-estate exposure   Greece  €million  284.64  269.77  14.88  0.35   Ireland  €million  1,311.72  1,217.73  93.99  85.35   Portugal  €million  1,715.69  1,593.12  122.57  0.55   Spain  €million  10,613.89  9,215.53  1,398.36  28.21   Offshore centers  €million  13,625.11  12,193.87  1,431.23  982.51   USA  €million  31,041.15  27,337.69  3,703.47  265.95   All foreign countries  €million  140,789.47  128,757.77  12,031.70  2,166.93  Banks   USA  €million  156,898.31  84,979.48  71,918.83  56,381.09  Offshore centers  €million  147,931.56  96,858.99  51,072.58  27,016.96  Banks’ total portfolio  €million  5,350,873.61  3,707,425.18  1,643,448.43  651,867.36  Type of exposure  Unit  Total exposure  Balance-sheet exposure  Off-balance-sheet exposure  Derivatives  Real-estate exposure   Greece  €million  284.64  269.77  14.88  0.35   Ireland  €million  1,311.72  1,217.73  93.99  85.35   Portugal  €million  1,715.69  1,593.12  122.57  0.55   Spain  €million  10,613.89  9,215.53  1,398.36  28.21   Offshore centers  €million  13,625.11  12,193.87  1,431.23  982.51   USA  €million  31,041.15  27,337.69  3,703.47  265.95   All foreign countries  €million  140,789.47  128,757.77  12,031.70  2,166.93  Banks   USA  €million  156,898.31  84,979.48  71,918.83  56,381.09  Offshore centers  €million  147,931.56  96,858.99  51,072.58  27,016.96  Banks’ total portfolio  €million  5,350,873.61  3,707,425.18  1,643,448.43  651,867.36  Table III. Regional Distribution for Exposures by German Banks in 2007Q2 This table provides the real-estate exposures taken from the Bundesbank credit register for Greece, Ireland, Portugal, Spain, offshore centers, the USA, and for all foreign countries in 2007Q2 for all German banks, including the big banks and the Landesbanks. Additionally, it provides the exposure toward banks in the USA, the exposure toward the offshore centers, and the banks’ total portfolio. The banks’ total portfolio does not comprise exposures to international organizations. The table lists the total exposure, the balance-sheet exposure, the off-balance-sheet exposure, and the derivatives calculated as the sum of all German banks in millions of euro. For offshore centers, the Bundesbank definition is used. According to this definition, offshore centers include twenty-three countries. Type of exposure  Unit  Total exposure  Balance-sheet exposure  Off-balance-sheet exposure  Derivatives  Real-estate exposure   Greece  €million  284.64  269.77  14.88  0.35   Ireland  €million  1,311.72  1,217.73  93.99  85.35   Portugal  €million  1,715.69  1,593.12  122.57  0.55   Spain  €million  10,613.89  9,215.53  1,398.36  28.21   Offshore centers  €million  13,625.11  12,193.87  1,431.23  982.51   USA  €million  31,041.15  27,337.69  3,703.47  265.95   All foreign countries  €million  140,789.47  128,757.77  12,031.70  2,166.93  Banks   USA  €million  156,898.31  84,979.48  71,918.83  56,381.09  Offshore centers  €million  147,931.56  96,858.99  51,072.58  27,016.96  Banks’ total portfolio  €million  5,350,873.61  3,707,425.18  1,643,448.43  651,867.36  Type of exposure  Unit  Total exposure  Balance-sheet exposure  Off-balance-sheet exposure  Derivatives  Real-estate exposure   Greece  €million  284.64  269.77  14.88  0.35   Ireland  €million  1,311.72  1,217.73  93.99  85.35   Portugal  €million  1,715.69  1,593.12  122.57  0.55   Spain  €million  10,613.89  9,215.53  1,398.36  28.21   Offshore centers  €million  13,625.11  12,193.87  1,431.23  982.51   USA  €million  31,041.15  27,337.69  3,703.47  265.95   All foreign countries  €million  140,789.47  128,757.77  12,031.70  2,166.93  Banks   USA  €million  156,898.31  84,979.48  71,918.83  56,381.09  Offshore centers  €million  147,931.56  96,858.99  51,072.58  27,016.96  Banks’ total portfolio  €million  5,350,873.61  3,707,425.18  1,643,448.43  651,867.36  Table III also breaks down the exposures by three types, that is, on- and off-balance sheet exposures and derivatives. The fraction of the on-balance sheet exposures ranges from 87% to 92% for all three types of exposures, which explains our later approach of focusing on these on-balance sheet exposures. The rest of the table lists the outstanding loans to US banks (€157 billion) and the total exposure to offshore centers (€148 billion). It is not surprising that the structure of lending to banks differs from the direct exposure to the real-estate sector as the former consists of similar shares of on- and off-balance sheet exposures due to a higher share of derivatives. The bulk of the German banks’ engagement in the US subprime mortgage market took place through the investment activities of their ABCP conduits, however.16Table IV provides the US real-estate, subprime, and conduit exposures for all German banks in 2007Q2. For the mean bank among the 1,547 banks in our sample, direct real-estate exposure in 2007Q2, for example, was equal to €17.7 million, exposure to major subprime lenders equal to €1.3 million, and conduit exposure equal to €81.3 million. Among the 41 banks with real-estate exposures, the mean (median) exposure amounted to €666.8 (€161.8) million; for the 78 banks with subprime exposures, the mean (median) exposure was equal to €25 (€5) million, while for the 13 banks with conduits, the mean (median) exposure was equal to €9.7 (€5.7) billion. Relative to total assets, real-estate exposure averages to 0.92% whereas ratios for subprime exposure and conduit exposure are documented as 0.27% and 2.83%, respectively. While acknowledging that only a small number of banks were exposed, we also note that these banks are lenders to a much higher number of borrowers compared with the rest of the banks without exposure. This is observable in the number of observations of the exposed group to the total number as ranging from 25% to 35% (e.g., see the later Table VI). In other words, these exposed banks have an important role in overall domestic lending. Table IV. Descriptive statistics for exposures by German banks in the USA in 2007Q2 This table provides the real estate, subprime, and conduit exposures in the USA in 2007Q2 for all German banks, including the big banks and the Landesbanks. The table lists the number of observations, mean, standard deviation, 10th percentile, 25th percentile, median, 75th percentile, and 90th percentile for all exposures in millions of euro. Variable  Unit  Number of Observations  Mean  Standard Deviation  10th Percentile  25th Percentile  Median  75th Percentile  90th Percentile  All banks   US real-estate exposure  €million  1,547  17.67  195.49  0.00  0.00  0.00  0.00  0.00   US real-estate exposure/total assets  Percentage  1,547  0.02  0.31  0.00  0.00  0.00  0.00  0.00   US subprime exposure  €million  1,547  1.26  19.35  0.00  0.00  0.00  0.00  0.00   US subprime exposure/total assets  Percentage  1,547  0.01  0.08  0.00  0.00  0.00  0.00  0.00   Conduit exposure  €million  1,547  81.30  1,270.32  0.00  0.00  0.00  0.00  0.00   Conduit exposure/total assets  Percentage  1,547  0.02  0.32  0.00  0.00  0.00  0.00  0.00   Banks with US real-estate exposure  €million  41  666.77  1,016.60  3.08  16.45  161.81  845.19  1,729.27   Banks with US real-estate exposure/total assets  Percentage  41  0.92  1.69  0.02  0.06  0.33  1.03  2.20   Banks with US subprime exposure  €million  78  25.08  83.12  1.89  2.04  4.99  12.45  29.98   Banks with US subprime exposure/total assets  Percentage  78  0.27  0.27  0.02  0.05  0.18  0.42  0.68   Banks with conduit exposure  €million  13  9,674.29  10,361.48  55.89  2,868.42  5,686.36  15,048.98  23,711.96   Banks with conduit exposure/total assets  Percentage  13  2.83  2.09  0.11  1.24  2.90  3.45  5.51  Variable  Unit  Number of Observations  Mean  Standard Deviation  10th Percentile  25th Percentile  Median  75th Percentile  90th Percentile  All banks   US real-estate exposure  €million  1,547  17.67  195.49  0.00  0.00  0.00  0.00  0.00   US real-estate exposure/total assets  Percentage  1,547  0.02  0.31  0.00  0.00  0.00  0.00  0.00   US subprime exposure  €million  1,547  1.26  19.35  0.00  0.00  0.00  0.00  0.00   US subprime exposure/total assets  Percentage  1,547  0.01  0.08  0.00  0.00  0.00  0.00  0.00   Conduit exposure  €million  1,547  81.30  1,270.32  0.00  0.00  0.00  0.00  0.00   Conduit exposure/total assets  Percentage  1,547  0.02  0.32  0.00  0.00  0.00  0.00  0.00   Banks with US real-estate exposure  €million  41  666.77  1,016.60  3.08  16.45  161.81  845.19  1,729.27   Banks with US real-estate exposure/total assets  Percentage  41  0.92  1.69  0.02  0.06  0.33  1.03  2.20   Banks with US subprime exposure  €million  78  25.08  83.12  1.89  2.04  4.99  12.45  29.98   Banks with US subprime exposure/total assets  Percentage  78  0.27  0.27  0.02  0.05  0.18  0.42  0.68   Banks with conduit exposure  €million  13  9,674.29  10,361.48  55.89  2,868.42  5,686.36  15,048.98  23,711.96   Banks with conduit exposure/total assets  Percentage  13  2.83  2.09  0.11  1.24  2.90  3.45  5.51  Table IV. Descriptive statistics for exposures by German banks in the USA in 2007Q2 This table provides the real estate, subprime, and conduit exposures in the USA in 2007Q2 for all German banks, including the big banks and the Landesbanks. The table lists the number of observations, mean, standard deviation, 10th percentile, 25th percentile, median, 75th percentile, and 90th percentile for all exposures in millions of euro. Variable  Unit  Number of Observations  Mean  Standard Deviation  10th Percentile  25th Percentile  Median  75th Percentile  90th Percentile  All banks   US real-estate exposure  €million  1,547  17.67  195.49  0.00  0.00  0.00  0.00  0.00   US real-estate exposure/total assets  Percentage  1,547  0.02  0.31  0.00  0.00  0.00  0.00  0.00   US subprime exposure  €million  1,547  1.26  19.35  0.00  0.00  0.00  0.00  0.00   US subprime exposure/total assets  Percentage  1,547  0.01  0.08  0.00  0.00  0.00  0.00  0.00   Conduit exposure  €million  1,547  81.30  1,270.32  0.00  0.00  0.00  0.00  0.00   Conduit exposure/total assets  Percentage  1,547  0.02  0.32  0.00  0.00  0.00  0.00  0.00   Banks with US real-estate exposure  €million  41  666.77  1,016.60  3.08  16.45  161.81  845.19  1,729.27   Banks with US real-estate exposure/total assets  Percentage  41  0.92  1.69  0.02  0.06  0.33  1.03  2.20   Banks with US subprime exposure  €million  78  25.08  83.12  1.89  2.04  4.99  12.45  29.98   Banks with US subprime exposure/total assets  Percentage  78  0.27  0.27  0.02  0.05  0.18  0.42  0.68   Banks with conduit exposure  €million  13  9,674.29  10,361.48  55.89  2,868.42  5,686.36  15,048.98  23,711.96   Banks with conduit exposure/total assets  Percentage  13  2.83  2.09  0.11  1.24  2.90  3.45  5.51  Variable  Unit  Number of Observations  Mean  Standard Deviation  10th Percentile  25th Percentile  Median  75th Percentile  90th Percentile  All banks   US real-estate exposure  €million  1,547  17.67  195.49  0.00  0.00  0.00  0.00  0.00   US real-estate exposure/total assets  Percentage  1,547  0.02  0.31  0.00  0.00  0.00  0.00  0.00   US subprime exposure  €million  1,547  1.26  19.35  0.00  0.00  0.00  0.00  0.00   US subprime exposure/total assets  Percentage  1,547  0.01  0.08  0.00  0.00  0.00  0.00  0.00   Conduit exposure  €million  1,547  81.30  1,270.32  0.00  0.00  0.00  0.00  0.00   Conduit exposure/total assets  Percentage  1,547  0.02  0.32  0.00  0.00  0.00  0.00  0.00   Banks with US real-estate exposure  €million  41  666.77  1,016.60  3.08  16.45  161.81  845.19  1,729.27   Banks with US real-estate exposure/total assets  Percentage  41  0.92  1.69  0.02  0.06  0.33  1.03  2.20   Banks with US subprime exposure  €million  78  25.08  83.12  1.89  2.04  4.99  12.45  29.98   Banks with US subprime exposure/total assets  Percentage  78  0.27  0.27  0.02  0.05  0.18  0.42  0.68   Banks with conduit exposure  €million  13  9,674.29  10,361.48  55.89  2,868.42  5,686.36  15,048.98  23,711.96   Banks with conduit exposure/total assets  Percentage  13  2.83  2.09  0.11  1.24  2.90  3.45  5.51  In Table V Panel A, we present the correlations between US real-estate, subprime, and conduit exposures. They are all highly correlated but the highest correlation is observed between the US real-estate and conduit exposures. In Panel B, we classify banks based on banking group and exposure type in 2007Q2. We observe that all big banks were exposed to the US real-estate market, while eleven out of twelve Landesbanks were involved in lending to this sector. Surprisingly, even a few savings banks and cooperative banks had US real-estate exposure whereas thirty-three savings banks and twenty-three cooperative banks had lending exposures to US subprime lenders. As expected, those two groups were not involved in providing liquidity to conduits.17 Table V. (A) Correlations between exposures and (B) number of banks by exposure and bank type in 2007Q2 Panel A reports the correlations between exposure types, and the significance levels for a Pearson correlation test, while Panel B lists the number of observations of banks by exposure type and bank type in 2007Q2. *Significant at 10%. Panel A     Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log US real-estate exposure  1      Log US subprime exposure  0.4894*  1    Log conduit exposure  0.7735*  0.3984*  1    Panel B    Bank type  Number of observations      With US real-estate exposure     Without US real-estate exposure    All banks  41     1,506   Commerical banks  20     139    Big banks  5     0    Regional banks  5     130    Mortgage banks  10     9               Public sector banks  16     442    Landesbanks  11     1    Savings banks  5     441               All cooperative banks  5     925    Cooperative central banks  2     0    Cooperative banks  3     925       Wth US subprime exposure     Without US subprime exposure    All banks  78     1,469   Commerical banks  14     145    Big banks  1     4    Regional banks  8     127    Mortgage banks  5     14   Public sector banks  41     417    Landesbanks  8     4    Savings banks  33     413   All cooperative banks  23     907    Cooperative central banks  0     2    Cooperative banks  23     905  Panel A     Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log US real-estate exposure  1      Log US subprime exposure  0.4894*  1    Log conduit exposure  0.7735*  0.3984*  1    Panel B    Bank type  Number of observations      With US real-estate exposure     Without US real-estate exposure    All banks  41     1,506   Commerical banks  20     139    Big banks  5     0    Regional banks  5     130    Mortgage banks  10     9               Public sector banks  16     442    Landesbanks  11     1    Savings banks  5     441               All cooperative banks  5     925    Cooperative central banks  2     0    Cooperative banks  3     925       Wth US subprime exposure     Without US subprime exposure    All banks  78     1,469   Commerical banks  14     145    Big banks  1     4    Regional banks  8     127    Mortgage banks  5     14   Public sector banks  41     417    Landesbanks  8     4    Savings banks  33     413   All cooperative banks  23     907    Cooperative central banks  0     2    Cooperative banks  23     905     With conduit exposure     Without conduit exposure  All banks  13     1,534   Commerical banks  4     155    Big banks  4     1    Regional banks  0     135    Mortgage banks  0     19   Public sector banks  7     451    Landesbanks  7     5    Savings banks  0     446   All cooperative banks  2     928    Cooperative central banks  2     0    Cooperative banks  0     928     With conduit exposure     Without conduit exposure  All banks  13     1,534   Commerical banks  4     155    Big banks  4     1    Regional banks  0     135    Mortgage banks  0     19   Public sector banks  7     451    Landesbanks  7     5    Savings banks  0     446   All cooperative banks  2     928    Cooperative central banks  2     0    Cooperative banks  0     928  Table V. (A) Correlations between exposures and (B) number of banks by exposure and bank type in 2007Q2 Panel A reports the correlations between exposure types, and the significance levels for a Pearson correlation test, while Panel B lists the number of observations of banks by exposure type and bank type in 2007Q2. *Significant at 10%. Panel A     Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log US real-estate exposure  1      Log US subprime exposure  0.4894*  1    Log conduit exposure  0.7735*  0.3984*  1    Panel B    Bank type  Number of observations      With US real-estate exposure     Without US real-estate exposure    All banks  41     1,506   Commerical banks  20     139    Big banks  5     0    Regional banks  5     130    Mortgage banks  10     9               Public sector banks  16     442    Landesbanks  11     1    Savings banks  5     441               All cooperative banks  5     925    Cooperative central banks  2     0    Cooperative banks  3     925       Wth US subprime exposure     Without US subprime exposure    All banks  78     1,469   Commerical banks  14     145    Big banks  1     4    Regional banks  8     127    Mortgage banks  5     14   Public sector banks  41     417    Landesbanks  8     4    Savings banks  33     413   All cooperative banks  23     907    Cooperative central banks  0     2    Cooperative banks  23     905  Panel A     Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log US real-estate exposure  1      Log US subprime exposure  0.4894*  1    Log conduit exposure  0.7735*  0.3984*  1    Panel B    Bank type  Number of observations      With US real-estate exposure     Without US real-estate exposure    All banks  41     1,506   Commerical banks  20     139    Big banks  5     0    Regional banks  5     130    Mortgage banks  10     9               Public sector banks  16     442    Landesbanks  11     1    Savings banks  5     441               All cooperative banks  5     925    Cooperative central banks  2     0    Cooperative banks  3     925       Wth US subprime exposure     Without US subprime exposure    All banks  78     1,469   Commerical banks  14     145    Big banks  1     4    Regional banks  8     127    Mortgage banks  5     14   Public sector banks  41     417    Landesbanks  8     4    Savings banks  33     413   All cooperative banks  23     907    Cooperative central banks  0     2    Cooperative banks  23     905     With conduit exposure     Without conduit exposure  All banks  13     1,534   Commerical banks  4     155    Big banks  4     1    Regional banks  0     135    Mortgage banks  0     19   Public sector banks  7     451    Landesbanks  7     5    Savings banks  0     446   All cooperative banks  2     928    Cooperative central banks  2     0    Cooperative banks  0     928     With conduit exposure     Without conduit exposure  All banks  13     1,534   Commerical banks  4     155    Big banks  4     1    Regional banks  0     135    Mortgage banks  0     19   Public sector banks  7     451    Landesbanks  7     5    Savings banks  0     446   All cooperative banks  2     928    Cooperative central banks  2     0    Cooperative banks  0     928  When measuring bank exposures to financial shocks in the USA, we differentiate between direct and indirect bank exposures. Direct exposures are taken from the credit register and appear talis qualis on banks’ balance sheets. Under indirect exposure, we consider the amount of liquidity that German banks provided to their conduits before and in the aftermath of the US crisis. These exposures are considered to be off-balance sheet and thus do not directly appear on banks’ balance sheets in the event that banks’ conduits run into trouble. More precisely, we define the three different bank exposures in the USA as follows. The first one is the Log US real-estate exposureit which is the logarithm of the total exposure of bank i in a particular quarter t to the US real-estate sector.18 More specifically, the variable is defined as the logarithm of one plus the exposure in order to retain the banks in the sample that do not have any exposure to the US real-estate sector. We use the same approach for the other exposures as well. This information is taken directly from the credit register. US real-estate exposureit is accumulated across individual bank–firm level exposures to the US real-estate sector (and therefore varies at the bank level but does not vary across firms borrowing from the same bank when we assess the changes in lending at the bank–firm level). We also define Log US subprime exposureit as the logarithm of the total exposure of bank i in a particular quarter t to the subprime lenders in the USA. Again this information is taken directly from the credit register. We gathered information on the top twenty-five subprime lenders in the USA which experienced difficulties during the US mortgage crisis. However, the German banks have been exposed to eighteen of these top subprime lenders.19 US subprime exposureit is accumulated across the individual exposures to subprime lenders in the USA and therefore varies only by bank. Finally, we define the log conduit exposureit as the logarithm of the total amount of liquidity provided by bank i in a particular quarter t to its ABCP conduits. The information on the liquidity lines is taken from Moody’s ABCP Query. Log conduit exposureit is accumulated across individual exposures to ABCP conduits and therefore varies again only by bank. Recall that the definition and measurement of bank exposures to financial shocks in the USA and all other variables were summarized in Table I. 3.4 Assessing Domestic Bank–Firm Lending in Germany Our model explains the quarterly change in lending, representing the first difference of the logarithm of domestic lending:   Δ log domestic lendingijt= log(domestic lendingijt)− log(domestic lendingijt−1), (1) where domestic lendingijt represents exposure of bank i to firm j in Germany in a particular quarter t. If the exposures were fully repaid during the quarter, zero values are reported at the end of the quarter in the dataset. However, in our analysis we capture only non-zero exposures and therefore predominately focus on continuing changes in domestic lending, that is, the “internal margin”. We have around 3.9 million bank–firm–quarter observations to assess domestic lending by banks to firms. We note that average domestic lending in our sample appears to be much lower compared with the average US real-estate exposure, the average subprime exposure, and the average conduit exposure. This is the case because large banks in particular have such exposures. Table VI compares the number of observations for banks with and without exposures to the real-estate sector, subprime, or conduits. These sub-samples are not mutually exclusive. The sub-sample of banks with real-estate exposure has over 1.4 million bank–firm–quarter observations of domestic lending. The average bank–firm level amount of domestic lending in this sub-sample equals €6 million and is somewhat larger than average domestic lending for the total sample. It should be mentioned that banks with conduit exposures belong to the same sub-sample. However, a couple of banks have subprime exposures although they do not have real-estate exposures. Table VI. Descriptive statistics, by bank exposures This table provides the number of observations, mean, and standard deviation for all bank-specific variables, by bank exposures, for the full sample. Variable  Unit  Number of observations  Mean  Standard deviation  Number of observations  Mean  Standard deviation        Banks witd US real-estate exposure  Banks witdout US real-estate exposure  Domestic lending  €million  1,412,731  6.08  24.54  2,450,487  2.22  7.92  US subprime exposure  €million  1,827,003  90.74  185.82  2,890,361  0.81  3.45  Conduit exposure  €million  1,827,003  4,798.17  7,605.05  2,890,361  0.00  0.00  Total assets  €million  1,826,999  271,374.60  215,268.00  2,889,950  7,842.87  13,472.22  Capital  Percentage  1,777,098  13.98  3.48  2,869,058  14.01  3.63  Liquidity  Percentage  1,826,946  7.14  4.51  2,767,824  8.23  5.68  ROA  Percentage  1,810,322  0.43  0.46  2,881,715  1.02  1.20  NPL  Percentage  1,810,322  2.95  2.37  2,881,063  5.36  5.46  Deposits  Percentage  1,438,295  5.94  7.78  2,601,426  30.12  12.33  CB funding  Percentage  1,821,826  2.74  2.62  2,890,306  1.29  2.57      Banks with US subprime exposure  Banks without US subprime exposure  Domestic lending  €million  925,784  5.72  24.37  2,937,434  2.97  12.55  US real-estate exposure  €million  1,157,278  1,216.53  1,584.37  3,560,086  190.70  684.35  Conduit exposure  €million  1,157,278  5,549.67  8,876.52  3,560,086  658.34  2,331.93  Total assets  €million  1,157,278  257,923.10  268,741.50  3,559,671  61,797.27  113,370.30  Capital  Percentage  1,124,986  13.68  3.28  3,521,170  14.10  3.66  Liquidity  Percentage  1,157,278  7.23  4.68  3,437,492  7.99  5.45  ROA  Percentage  1,156,479  0.46  0.39  3,535,558  0.91  1.14  NPL  Percentage  1,156,479  2.83  2.35  3,534,906  4.95  5.11  Deposits  Percentage  877,500  10.04  13.31  3,162,221  24.69  15.10  CB funding  Percentage  1,157,278  2.32  2.68  3,554,854  1.70  2.66        Banks with conduit exposure  Banks without conduit exposure  Domestic lending  €million  914,750  5.89  25.82  2,948,468  2.93  11.68  US real-estate exposure  €million  1,249,293  1,175.60  1,420.14  3,468,071  178.23  770.38  US subprime exposure  €million  1,249,293  123.40  212.89  3,468,071  4.03  26.66  Total assets  €million  1,249,293  354,940.50  207,099.10  3,467,656  21,640.56  45,939.66  Capital  Percentage  1,241,699  14.89  3.44  3,404,457  13.67  3.56  Liquidity  Percentage  1,249,293  7.43  2.96  3,345,477  7.94  5.91  ROA  Percentage  1,241,258  0.38  0.39  3,450,779  0.94  1.14  NPL  Percentage  1,241,258  2.47  1.90  3,450,127  5.14  5.15  Deposits  Percentage  1,101,002  3.45  2.95  2,938,719  28.27  13.32  CB funding  Percentage  1,249,293  2.68  1.91  3,462,839  1.55  2.85  Variable  Unit  Number of observations  Mean  Standard deviation  Number of observations  Mean  Standard deviation        Banks witd US real-estate exposure  Banks witdout US real-estate exposure  Domestic lending  €million  1,412,731  6.08  24.54  2,450,487  2.22  7.92  US subprime exposure  €million  1,827,003  90.74  185.82  2,890,361  0.81  3.45  Conduit exposure  €million  1,827,003  4,798.17  7,605.05  2,890,361  0.00  0.00  Total assets  €million  1,826,999  271,374.60  215,268.00  2,889,950  7,842.87  13,472.22  Capital  Percentage  1,777,098  13.98  3.48  2,869,058  14.01  3.63  Liquidity  Percentage  1,826,946  7.14  4.51  2,767,824  8.23  5.68  ROA  Percentage  1,810,322  0.43  0.46  2,881,715  1.02  1.20  NPL  Percentage  1,810,322  2.95  2.37  2,881,063  5.36  5.46  Deposits  Percentage  1,438,295  5.94  7.78  2,601,426  30.12  12.33  CB funding  Percentage  1,821,826  2.74  2.62  2,890,306  1.29  2.57      Banks with US subprime exposure  Banks without US subprime exposure  Domestic lending  €million  925,784  5.72  24.37  2,937,434  2.97  12.55  US real-estate exposure  €million  1,157,278  1,216.53  1,584.37  3,560,086  190.70  684.35  Conduit exposure  €million  1,157,278  5,549.67  8,876.52  3,560,086  658.34  2,331.93  Total assets  €million  1,157,278  257,923.10  268,741.50  3,559,671  61,797.27  113,370.30  Capital  Percentage  1,124,986  13.68  3.28  3,521,170  14.10  3.66  Liquidity  Percentage  1,157,278  7.23  4.68  3,437,492  7.99  5.45  ROA  Percentage  1,156,479  0.46  0.39  3,535,558  0.91  1.14  NPL  Percentage  1,156,479  2.83  2.35  3,534,906  4.95  5.11  Deposits  Percentage  877,500  10.04  13.31  3,162,221  24.69  15.10  CB funding  Percentage  1,157,278  2.32  2.68  3,554,854  1.70  2.66        Banks with conduit exposure  Banks without conduit exposure  Domestic lending  €million  914,750  5.89  25.82  2,948,468  2.93  11.68  US real-estate exposure  €million  1,249,293  1,175.60  1,420.14  3,468,071  178.23  770.38  US subprime exposure  €million  1,249,293  123.40  212.89  3,468,071  4.03  26.66  Total assets  €million  1,249,293  354,940.50  207,099.10  3,467,656  21,640.56  45,939.66  Capital  Percentage  1,241,699  14.89  3.44  3,404,457  13.67  3.56  Liquidity  Percentage  1,249,293  7.43  2.96  3,345,477  7.94  5.91  ROA  Percentage  1,241,258  0.38  0.39  3,450,779  0.94  1.14  NPL  Percentage  1,241,258  2.47  1.90  3,450,127  5.14  5.15  Deposits  Percentage  1,101,002  3.45  2.95  2,938,719  28.27  13.32  CB funding  Percentage  1,249,293  2.68  1.91  3,462,839  1.55  2.85  Table VI. Descriptive statistics, by bank exposures This table provides the number of observations, mean, and standard deviation for all bank-specific variables, by bank exposures, for the full sample. Variable  Unit  Number of observations  Mean  Standard deviation  Number of observations  Mean  Standard deviation        Banks witd US real-estate exposure  Banks witdout US real-estate exposure  Domestic lending  €million  1,412,731  6.08  24.54  2,450,487  2.22  7.92  US subprime exposure  €million  1,827,003  90.74  185.82  2,890,361  0.81  3.45  Conduit exposure  €million  1,827,003  4,798.17  7,605.05  2,890,361  0.00  0.00  Total assets  €million  1,826,999  271,374.60  215,268.00  2,889,950  7,842.87  13,472.22  Capital  Percentage  1,777,098  13.98  3.48  2,869,058  14.01  3.63  Liquidity  Percentage  1,826,946  7.14  4.51  2,767,824  8.23  5.68  ROA  Percentage  1,810,322  0.43  0.46  2,881,715  1.02  1.20  NPL  Percentage  1,810,322  2.95  2.37  2,881,063  5.36  5.46  Deposits  Percentage  1,438,295  5.94  7.78  2,601,426  30.12  12.33  CB funding  Percentage  1,821,826  2.74  2.62  2,890,306  1.29  2.57      Banks with US subprime exposure  Banks without US subprime exposure  Domestic lending  €million  925,784  5.72  24.37  2,937,434  2.97  12.55  US real-estate exposure  €million  1,157,278  1,216.53  1,584.37  3,560,086  190.70  684.35  Conduit exposure  €million  1,157,278  5,549.67  8,876.52  3,560,086  658.34  2,331.93  Total assets  €million  1,157,278  257,923.10  268,741.50  3,559,671  61,797.27  113,370.30  Capital  Percentage  1,124,986  13.68  3.28  3,521,170  14.10  3.66  Liquidity  Percentage  1,157,278  7.23  4.68  3,437,492  7.99  5.45  ROA  Percentage  1,156,479  0.46  0.39  3,535,558  0.91  1.14  NPL  Percentage  1,156,479  2.83  2.35  3,534,906  4.95  5.11  Deposits  Percentage  877,500  10.04  13.31  3,162,221  24.69  15.10  CB funding  Percentage  1,157,278  2.32  2.68  3,554,854  1.70  2.66        Banks with conduit exposure  Banks without conduit exposure  Domestic lending  €million  914,750  5.89  25.82  2,948,468  2.93  11.68  US real-estate exposure  €million  1,249,293  1,175.60  1,420.14  3,468,071  178.23  770.38  US subprime exposure  €million  1,249,293  123.40  212.89  3,468,071  4.03  26.66  Total assets  €million  1,249,293  354,940.50  207,099.10  3,467,656  21,640.56  45,939.66  Capital  Percentage  1,241,699  14.89  3.44  3,404,457  13.67  3.56  Liquidity  Percentage  1,249,293  7.43  2.96  3,345,477  7.94  5.91  ROA  Percentage  1,241,258  0.38  0.39  3,450,779  0.94  1.14  NPL  Percentage  1,241,258  2.47  1.90  3,450,127  5.14  5.15  Deposits  Percentage  1,101,002  3.45  2.95  2,938,719  28.27  13.32  CB funding  Percentage  1,249,293  2.68  1.91  3,462,839  1.55  2.85  Variable  Unit  Number of observations  Mean  Standard deviation  Number of observations  Mean  Standard deviation        Banks witd US real-estate exposure  Banks witdout US real-estate exposure  Domestic lending  €million  1,412,731  6.08  24.54  2,450,487  2.22  7.92  US subprime exposure  €million  1,827,003  90.74  185.82  2,890,361  0.81  3.45  Conduit exposure  €million  1,827,003  4,798.17  7,605.05  2,890,361  0.00  0.00  Total assets  €million  1,826,999  271,374.60  215,268.00  2,889,950  7,842.87  13,472.22  Capital  Percentage  1,777,098  13.98  3.48  2,869,058  14.01  3.63  Liquidity  Percentage  1,826,946  7.14  4.51  2,767,824  8.23  5.68  ROA  Percentage  1,810,322  0.43  0.46  2,881,715  1.02  1.20  NPL  Percentage  1,810,322  2.95  2.37  2,881,063  5.36  5.46  Deposits  Percentage  1,438,295  5.94  7.78  2,601,426  30.12  12.33  CB funding  Percentage  1,821,826  2.74  2.62  2,890,306  1.29  2.57      Banks with US subprime exposure  Banks without US subprime exposure  Domestic lending  €million  925,784  5.72  24.37  2,937,434  2.97  12.55  US real-estate exposure  €million  1,157,278  1,216.53  1,584.37  3,560,086  190.70  684.35  Conduit exposure  €million  1,157,278  5,549.67  8,876.52  3,560,086  658.34  2,331.93  Total assets  €million  1,157,278  257,923.10  268,741.50  3,559,671  61,797.27  113,370.30  Capital  Percentage  1,124,986  13.68  3.28  3,521,170  14.10  3.66  Liquidity  Percentage  1,157,278  7.23  4.68  3,437,492  7.99  5.45  ROA  Percentage  1,156,479  0.46  0.39  3,535,558  0.91  1.14  NPL  Percentage  1,156,479  2.83  2.35  3,534,906  4.95  5.11  Deposits  Percentage  877,500  10.04  13.31  3,162,221  24.69  15.10  CB funding  Percentage  1,157,278  2.32  2.68  3,554,854  1.70  2.66        Banks with conduit exposure  Banks without conduit exposure  Domestic lending  €million  914,750  5.89  25.82  2,948,468  2.93  11.68  US real-estate exposure  €million  1,249,293  1,175.60  1,420.14  3,468,071  178.23  770.38  US subprime exposure  €million  1,249,293  123.40  212.89  3,468,071  4.03  26.66  Total assets  €million  1,249,293  354,940.50  207,099.10  3,467,656  21,640.56  45,939.66  Capital  Percentage  1,241,699  14.89  3.44  3,404,457  13.67  3.56  Liquidity  Percentage  1,249,293  7.43  2.96  3,345,477  7.94  5.91  ROA  Percentage  1,241,258  0.38  0.39  3,450,779  0.94  1.14  NPL  Percentage  1,241,258  2.47  1.90  3,450,127  5.14  5.15  Deposits  Percentage  1,101,002  3.45  2.95  2,938,719  28.27  13.32  CB funding  Percentage  1,249,293  2.68  1.91  3,462,839  1.55  2.85  The sub-sample of banks with subprime exposure provides us with 926,000 bank–firm–quarter observations of domestic lending. The average domestic lending for this sub-sample amounts to €5.7 million. Also a number of banks do have direct exposure to the US real-estate sector and provide liquidity to conduits although they do not offer loans to subprime lenders. We note that the number of banks that provide liquidity to conduits is significantly smaller compared with the sub-samples discussed before. The sub-sample of banks with conduit exposure provides us with only around 426,000 bank–firm–quarter observations of domestic lending. Similar to the sub-samples with direct exposure to the real-estate sector and the one with subprime exposure, the average for domestic lending in this sub-sample, at €5.9 million, tends to be larger than the average for the total sample. 4. Explaining Domestic Bank–Firm Lending in Germany 4.1 Specifications In Table VII we run the growth in domestic bank lending in Germany on different types of exposures, starting with total US exposure (total exposure to the US real-estate market), followed by a classification of our main variables of interest; US real-estate exposure, US subprime exposure and conduit exposure, and various interactions that are introduced in different models, for the sample that consists of the 1,664,262 bank–firm–quarter credit exposures.20 Table VII. Explaining the change in domestic bank lending in Germany following shocks to exposures in the USA The dependent variable is the quarter-on-quarter logarithmic change in domestic lending by banks to firms (ΔLog domestic lending) and an ordinary least squares estimation is used. Table I contains all variable definitions. In our estimations, the measurement for insolvency, capital, liquidity, ROA, NPL, deposits, and CB funding are in ratios. Coefficients are listed in the first row, robust standard errors clustered at bank level are reported in the row below, and the corresponding significance levels are adjacent to the coefficient. “Yes” indicates that the set of fixed effects is included. ***Significant at 1%. **Significant at 5%. *Significant at 10%. Model  1  2  3  4  5  6  7  8  9  10  11  12  Log exposure =  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  –0.006  0.003  0.001  –0.017  –0.007  –0.009  –0.003  0.001  0.001  –0.02  –0.006  –0.008  [0.013]  [0.014]  [0.014]  [0.019]  [0.021]  [0.021]  [0.009]  [0.010]  [0.010]  [0.015]  [0.015]  [0.015]  Log exposuret−1 * ΔUS homepricest     0.978***  1.035***     1.148***  1.204***     0.416  0.402     1.414***  1.499***     [0.288]  [0.292]     [0.332]  [0.343]     [0.338]  [0.350]     [0.434]  [0.437]  Log exposuret−1 * insolvencyt−1        –6.095*        –5.702        1.330        –7.396*        [3.593]        [4.177]        [4.317]        [4.169]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1        312.665**        386.937**        105.93        330.725*        [143.579]        [156.078]        [201.079]        [171.839]  ΔUS homepricest  –5.169  –4.476  –4.398  –5.018  –4.155  –4.058  –5.168  –5.215  –5.16  –5.234  –4.308  –4.221  [3.471]  [3.492]  [3.485]  [3.420]  [3.415]  [3.408]  [3.494]  [3.506]  [3.504]  [3.414]  [3.456]  [3.447]  Insolvencyt−1  –9.041  –8.259  –13.46  –9.061  –8.604  –13.76  –8.963  –8.785  –14.439  –9.206  –8.415  –12.453  [19.557]  [19.625]  [19.828]  [19.558]  [19.603]  [20.064]  [19.556]  [19.580]  [18.963]  [19.542]  [19.633]  [20.199]  ΔUS homepricest * insolvencyt–1        –766.254        –713.86        –962.887        –857.229        [789.980]        [792.499]        [750.649]        [783.057]  Sizet−1  –0.041  –0.037  –0.038  –0.015  –0.01  –0.011  –0.053  –0.054  –0.054  –0.007  –0.005  –0.005  [0.049]  [0.048]  [0.048]  [0.050]  [0.047]  [0.048]  [0.049]  [0.049]  [0.049]  [0.060]  [0.062]  [0.062]  Capitalt−1  7.679  7.749  7.705  7.293  7.261  7.250  7.798  7.885  7.890  8.057  8.105  8.062  [7.346]  [7.291]  [7.295]  [7.482]  [7.411]  [7.409]  [7.030]  [7.007]  [7.007]  [7.137]  [7.087]  [7.079]  Liquidityt−1  0.773  0.795  0.787  0.861  0.811  0.809  0.746  0.758  0.764  0.955  0.865  0.866  [1.150]  [1.148]  [1.149]  [1.159]  [1.159]  [1.160]  [1.155]  [1.152]  [1.152]  [1.162]  [1.156]  [1.156]  ROAt−1  16.157  15.478  15.811  15.157  14.343  14.56  16.941  17.113  17.222  14.452  14.187  14.536  [24.206]  [24.866]  [24.831]  [24.461]  [24.862]  [24.862]  [23.304]  [23.122]  [23.144]  [23.546]  [24.746]  [24.729]  NPLt−1  1.019  0.904  0.922  1.042  0.913  0.939  1.011  1.028  1.031  0.622  0.518  0.532  [3.194]  [3.226]  [3.229]  [3.182]  [3.223]  [3.229]  [3.179]  [3.164]  [3.165]  [3.211]  [3.300]  [3.302]  Depositst−1  3.970***  4.032***  4.037***  3.829***  3.890***  3.893***  4.027***  4.044***  4.046***  3.800***  3.892***  3.891***  [0.636]  [0.626]  [0.626]  [0.629]  [0.620]  [0.619]  [0.648]  [0.652]  [0.651]  [0.715]  [0.689]  [0.689]  CB fundingt−1  6.737***  6.801***  6.813***  6.688***  6.712***  6.713***  6.767***  6.811***  6.799***  6.455***  6.258***  6.265***  [1.950]  [1.977]  [1.977]  [1.939]  [1.953]  [1.955]  [1.941]  [1.965]  [1.966]  [2.012]  [2.022]  [2.021]  Constant  –2.003***  –1.986***  –1.978***  –2.011***  –1.993***  –1.986***  –2.000***  –2.000***  –1.992***  –2.010***  –1.980***  –1.972***  [0.062]  [0.062]  [0.062]  [0.063]  [0.064]  [0.064]  [0.061]  [0.061]  [0.061]  [0.062]  [0.063]  [0.063]  Bank-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Firm-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  R-squared  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  Difference in change in domestic bank lending in Germany between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1  –0.0012  0.0006  0.0002                               €1 billion in US real-estate exposuret−1           –0.0035  –0.0015  –0.0019                      €100 million in US subprime exposuret−1                    –0.0006  0.0002  0.0002             €1 billion in conduit exposuret−1                             –0.0041  –0.0012  –0.0017  Additional difference in change in domestic bank lending in Germany following a five-index point decrease in US homeprices between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1     –0.0101  –0.0107                               €1 billion in US real-estate exposuret−1              –0.0119  –0.0125                      €100 million in US subprime exposuret−1                       –0.0038  –0.0037             €1 billion in conduit exposuret−1                                –0.0147  –0.0155  Additional difference in change in domestic bank lending to firms in industries–regions with a 1%p higher insolvency rate in Germany between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1        –0.0126                               €1 billion in US real-estate exposuret−1                 –0.0118                      €100 million in US subprime exposuret−1                          0.0024             €1 billion in conduit exposuret−1                                   –0.0153  Additional difference in change in domestic bank lending to firms in industries–regions with a 1%p higher insolvency rate in Germany following a 5-index point decrease in US homeprices between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1        –0.0324                               €1 billion in US real-estate exposuret−1                 –0.0401                      €100 million in US subprime exposuret−1                          –0.0098             €1 billion in conduit exposuret−1                                   –0.0343  Model  1  2  3  4  5  6  7  8  9  10  11  12  Log exposure =  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  –0.006  0.003  0.001  –0.017  –0.007  –0.009  –0.003  0.001  0.001  –0.02  –0.006  –0.008  [0.013]  [0.014]  [0.014]  [0.019]  [0.021]  [0.021]  [0.009]  [0.010]  [0.010]  [0.015]  [0.015]  [0.015]  Log exposuret−1 * ΔUS homepricest     0.978***  1.035***     1.148***  1.204***     0.416  0.402     1.414***  1.499***     [0.288]  [0.292]     [0.332]  [0.343]     [0.338]  [0.350]     [0.434]  [0.437]  Log exposuret−1 * insolvencyt−1        –6.095*        –5.702        1.330        –7.396*        [3.593]        [4.177]        [4.317]        [4.169]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1        312.665**        386.937**        105.93        330.725*        [143.579]        [156.078]        [201.079]        [171.839]  ΔUS homepricest  –5.169  –4.476  –4.398  –5.018  –4.155  –4.058  –5.168  –5.215  –5.16  –5.234  –4.308  –4.221  [3.471]  [3.492]  [3.485]  [3.420]  [3.415]  [3.408]  [3.494]  [3.506]  [3.504]  [3.414]  [3.456]  [3.447]  Insolvencyt−1  –9.041  –8.259  –13.46  –9.061  –8.604  –13.76  –8.963  –8.785  –14.439  –9.206  –8.415  –12.453  [19.557]  [19.625]  [19.828]  [19.558]  [19.603]  [20.064]  [19.556]  [19.580]  [18.963]  [19.542]  [19.633]  [20.199]  ΔUS homepricest * insolvencyt–1        –766.254        –713.86        –962.887        –857.229        [789.980]        [792.499]        [750.649]        [783.057]  Sizet−1  –0.041  –0.037  –0.038  –0.015  –0.01  –0.011  –0.053  –0.054  –0.054  –0.007  –0.005  –0.005  [0.049]  [0.048]  [0.048]  [0.050]  [0.047]  [0.048]  [0.049]  [0.049]  [0.049]  [0.060]  [0.062]  [0.062]  Capitalt−1  7.679  7.749  7.705  7.293  7.261  7.250  7.798  7.885  7.890  8.057  8.105  8.062  [7.346]  [7.291]  [7.295]  [7.482]  [7.411]  [7.409]  [7.030]  [7.007]  [7.007]  [7.137]  [7.087]  [7.079]  Liquidityt−1  0.773  0.795  0.787  0.861  0.811  0.809  0.746  0.758  0.764  0.955  0.865  0.866  [1.150]  [1.148]  [1.149]  [1.159]  [1.159]  [1.160]  [1.155]  [1.152]  [1.152]  [1.162]  [1.156]  [1.156]  ROAt−1  16.157  15.478  15.811  15.157  14.343  14.56  16.941  17.113  17.222  14.452  14.187  14.536  [24.206]  [24.866]  [24.831]  [24.461]  [24.862]  [24.862]  [23.304]  [23.122]  [23.144]  [23.546]  [24.746]  [24.729]  NPLt−1  1.019  0.904  0.922  1.042  0.913  0.939  1.011  1.028  1.031  0.622  0.518  0.532  [3.194]  [3.226]  [3.229]  [3.182]  [3.223]  [3.229]  [3.179]  [3.164]  [3.165]  [3.211]  [3.300]  [3.302]  Depositst−1  3.970***  4.032***  4.037***  3.829***  3.890***  3.893***  4.027***  4.044***  4.046***  3.800***  3.892***  3.891***  [0.636]  [0.626]  [0.626]  [0.629]  [0.620]  [0.619]  [0.648]  [0.652]  [0.651]  [0.715]  [0.689]  [0.689]  CB fundingt−1  6.737***  6.801***  6.813***  6.688***  6.712***  6.713***  6.767***  6.811***  6.799***  6.455***  6.258***  6.265***  [1.950]  [1.977]  [1.977]  [1.939]  [1.953]  [1.955]  [1.941]  [1.965]  [1.966]  [2.012]  [2.022]  [2.021]  Constant  –2.003***  –1.986***  –1.978***  –2.011***  –1.993***  –1.986***  –2.000***  –2.000***  –1.992***  –2.010***  –1.980***  –1.972***  [0.062]  [0.062]  [0.062]  [0.063]  [0.064]  [0.064]  [0.061]  [0.061]  [0.061]  [0.062]  [0.063]  [0.063]  Bank-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Firm-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  R-squared  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  Difference in change in domestic bank lending in Germany between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1  –0.0012  0.0006  0.0002                               €1 billion in US real-estate exposuret−1           –0.0035  –0.0015  –0.0019                      €100 million in US subprime exposuret−1                    –0.0006  0.0002  0.0002             €1 billion in conduit exposuret−1                             –0.0041  –0.0012  –0.0017  Additional difference in change in domestic bank lending in Germany following a five-index point decrease in US homeprices between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1     –0.0101  –0.0107                               €1 billion in US real-estate exposuret−1              –0.0119  –0.0125                      €100 million in US subprime exposuret−1                       –0.0038  –0.0037             €1 billion in conduit exposuret−1                                –0.0147  –0.0155  Additional difference in change in domestic bank lending to firms in industries–regions with a 1%p higher insolvency rate in Germany between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1        –0.0126                               €1 billion in US real-estate exposuret−1                 –0.0118                      €100 million in US subprime exposuret−1                          0.0024             €1 billion in conduit exposuret−1                                   –0.0153  Additional difference in change in domestic bank lending to firms in industries–regions with a 1%p higher insolvency rate in Germany following a 5-index point decrease in US homeprices between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1        –0.0324                               €1 billion in US real-estate exposuret−1                 –0.0401                      €100 million in US subprime exposuret−1                          –0.0098             €1 billion in conduit exposuret−1                                   –0.0343  Table VII. Explaining the change in domestic bank lending in Germany following shocks to exposures in the USA The dependent variable is the quarter-on-quarter logarithmic change in domestic lending by banks to firms (ΔLog domestic lending) and an ordinary least squares estimation is used. Table I contains all variable definitions. In our estimations, the measurement for insolvency, capital, liquidity, ROA, NPL, deposits, and CB funding are in ratios. Coefficients are listed in the first row, robust standard errors clustered at bank level are reported in the row below, and the corresponding significance levels are adjacent to the coefficient. “Yes” indicates that the set of fixed effects is included. ***Significant at 1%. **Significant at 5%. *Significant at 10%. Model  1  2  3  4  5  6  7  8  9  10  11  12  Log exposure =  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  –0.006  0.003  0.001  –0.017  –0.007  –0.009  –0.003  0.001  0.001  –0.02  –0.006  –0.008  [0.013]  [0.014]  [0.014]  [0.019]  [0.021]  [0.021]  [0.009]  [0.010]  [0.010]  [0.015]  [0.015]  [0.015]  Log exposuret−1 * ΔUS homepricest     0.978***  1.035***     1.148***  1.204***     0.416  0.402     1.414***  1.499***     [0.288]  [0.292]     [0.332]  [0.343]     [0.338]  [0.350]     [0.434]  [0.437]  Log exposuret−1 * insolvencyt−1        –6.095*        –5.702        1.330        –7.396*        [3.593]        [4.177]        [4.317]        [4.169]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1        312.665**        386.937**        105.93        330.725*        [143.579]        [156.078]        [201.079]        [171.839]  ΔUS homepricest  –5.169  –4.476  –4.398  –5.018  –4.155  –4.058  –5.168  –5.215  –5.16  –5.234  –4.308  –4.221  [3.471]  [3.492]  [3.485]  [3.420]  [3.415]  [3.408]  [3.494]  [3.506]  [3.504]  [3.414]  [3.456]  [3.447]  Insolvencyt−1  –9.041  –8.259  –13.46  –9.061  –8.604  –13.76  –8.963  –8.785  –14.439  –9.206  –8.415  –12.453  [19.557]  [19.625]  [19.828]  [19.558]  [19.603]  [20.064]  [19.556]  [19.580]  [18.963]  [19.542]  [19.633]  [20.199]  ΔUS homepricest * insolvencyt–1        –766.254        –713.86        –962.887        –857.229        [789.980]        [792.499]        [750.649]        [783.057]  Sizet−1  –0.041  –0.037  –0.038  –0.015  –0.01  –0.011  –0.053  –0.054  –0.054  –0.007  –0.005  –0.005  [0.049]  [0.048]  [0.048]  [0.050]  [0.047]  [0.048]  [0.049]  [0.049]  [0.049]  [0.060]  [0.062]  [0.062]  Capitalt−1  7.679  7.749  7.705  7.293  7.261  7.250  7.798  7.885  7.890  8.057  8.105  8.062  [7.346]  [7.291]  [7.295]  [7.482]  [7.411]  [7.409]  [7.030]  [7.007]  [7.007]  [7.137]  [7.087]  [7.079]  Liquidityt−1  0.773  0.795  0.787  0.861  0.811  0.809  0.746  0.758  0.764  0.955  0.865  0.866  [1.150]  [1.148]  [1.149]  [1.159]  [1.159]  [1.160]  [1.155]  [1.152]  [1.152]  [1.162]  [1.156]  [1.156]  ROAt−1  16.157  15.478  15.811  15.157  14.343  14.56  16.941  17.113  17.222  14.452  14.187  14.536  [24.206]  [24.866]  [24.831]  [24.461]  [24.862]  [24.862]  [23.304]  [23.122]  [23.144]  [23.546]  [24.746]  [24.729]  NPLt−1  1.019  0.904  0.922  1.042  0.913  0.939  1.011  1.028  1.031  0.622  0.518  0.532  [3.194]  [3.226]  [3.229]  [3.182]  [3.223]  [3.229]  [3.179]  [3.164]  [3.165]  [3.211]  [3.300]  [3.302]  Depositst−1  3.970***  4.032***  4.037***  3.829***  3.890***  3.893***  4.027***  4.044***  4.046***  3.800***  3.892***  3.891***  [0.636]  [0.626]  [0.626]  [0.629]  [0.620]  [0.619]  [0.648]  [0.652]  [0.651]  [0.715]  [0.689]  [0.689]  CB fundingt−1  6.737***  6.801***  6.813***  6.688***  6.712***  6.713***  6.767***  6.811***  6.799***  6.455***  6.258***  6.265***  [1.950]  [1.977]  [1.977]  [1.939]  [1.953]  [1.955]  [1.941]  [1.965]  [1.966]  [2.012]  [2.022]  [2.021]  Constant  –2.003***  –1.986***  –1.978***  –2.011***  –1.993***  –1.986***  –2.000***  –2.000***  –1.992***  –2.010***  –1.980***  –1.972***  [0.062]  [0.062]  [0.062]  [0.063]  [0.064]  [0.064]  [0.061]  [0.061]  [0.061]  [0.062]  [0.063]  [0.063]  Bank-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Firm-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  R-squared  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  Difference in change in domestic bank lending in Germany between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1  –0.0012  0.0006  0.0002                               €1 billion in US real-estate exposuret−1           –0.0035  –0.0015  –0.0019                      €100 million in US subprime exposuret−1                    –0.0006  0.0002  0.0002             €1 billion in conduit exposuret−1                             –0.0041  –0.0012  –0.0017  Additional difference in change in domestic bank lending in Germany following a five-index point decrease in US homeprices between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1     –0.0101  –0.0107                               €1 billion in US real-estate exposuret−1              –0.0119  –0.0125                      €100 million in US subprime exposuret−1                       –0.0038  –0.0037             €1 billion in conduit exposuret−1                                –0.0147  –0.0155  Additional difference in change in domestic bank lending to firms in industries–regions with a 1%p higher insolvency rate in Germany between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1        –0.0126                               €1 billion in US real-estate exposuret−1                 –0.0118                      €100 million in US subprime exposuret−1                          0.0024             €1 billion in conduit exposuret−1                                   –0.0153  Additional difference in change in domestic bank lending to firms in industries–regions with a 1%p higher insolvency rate in Germany following a 5-index point decrease in US homeprices between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1        –0.0324                               €1 billion in US real-estate exposuret−1                 –0.0401                      €100 million in US subprime exposuret−1                          –0.0098             €1 billion in conduit exposuret−1                                   –0.0343  Model  1  2  3  4  5  6  7  8  9  10  11  12  Log exposure =  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  –0.006  0.003  0.001  –0.017  –0.007  –0.009  –0.003  0.001  0.001  –0.02  –0.006  –0.008  [0.013]  [0.014]  [0.014]  [0.019]  [0.021]  [0.021]  [0.009]  [0.010]  [0.010]  [0.015]  [0.015]  [0.015]  Log exposuret−1 * ΔUS homepricest     0.978***  1.035***     1.148***  1.204***     0.416  0.402     1.414***  1.499***     [0.288]  [0.292]     [0.332]  [0.343]     [0.338]  [0.350]     [0.434]  [0.437]  Log exposuret−1 * insolvencyt−1        –6.095*        –5.702        1.330        –7.396*        [3.593]        [4.177]        [4.317]        [4.169]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1        312.665**        386.937**        105.93        330.725*        [143.579]        [156.078]        [201.079]        [171.839]  ΔUS homepricest  –5.169  –4.476  –4.398  –5.018  –4.155  –4.058  –5.168  –5.215  –5.16  –5.234  –4.308  –4.221  [3.471]  [3.492]  [3.485]  [3.420]  [3.415]  [3.408]  [3.494]  [3.506]  [3.504]  [3.414]  [3.456]  [3.447]  Insolvencyt−1  –9.041  –8.259  –13.46  –9.061  –8.604  –13.76  –8.963  –8.785  –14.439  –9.206  –8.415  –12.453  [19.557]  [19.625]  [19.828]  [19.558]  [19.603]  [20.064]  [19.556]  [19.580]  [18.963]  [19.542]  [19.633]  [20.199]  ΔUS homepricest * insolvencyt–1        –766.254        –713.86        –962.887        –857.229        [789.980]        [792.499]        [750.649]        [783.057]  Sizet−1  –0.041  –0.037  –0.038  –0.015  –0.01  –0.011  –0.053  –0.054  –0.054  –0.007  –0.005  –0.005  [0.049]  [0.048]  [0.048]  [0.050]  [0.047]  [0.048]  [0.049]  [0.049]  [0.049]  [0.060]  [0.062]  [0.062]  Capitalt−1  7.679  7.749  7.705  7.293  7.261  7.250  7.798  7.885  7.890  8.057  8.105  8.062  [7.346]  [7.291]  [7.295]  [7.482]  [7.411]  [7.409]  [7.030]  [7.007]  [7.007]  [7.137]  [7.087]  [7.079]  Liquidityt−1  0.773  0.795  0.787  0.861  0.811  0.809  0.746  0.758  0.764  0.955  0.865  0.866  [1.150]  [1.148]  [1.149]  [1.159]  [1.159]  [1.160]  [1.155]  [1.152]  [1.152]  [1.162]  [1.156]  [1.156]  ROAt−1  16.157  15.478  15.811  15.157  14.343  14.56  16.941  17.113  17.222  14.452  14.187  14.536  [24.206]  [24.866]  [24.831]  [24.461]  [24.862]  [24.862]  [23.304]  [23.122]  [23.144]  [23.546]  [24.746]  [24.729]  NPLt−1  1.019  0.904  0.922  1.042  0.913  0.939  1.011  1.028  1.031  0.622  0.518  0.532  [3.194]  [3.226]  [3.229]  [3.182]  [3.223]  [3.229]  [3.179]  [3.164]  [3.165]  [3.211]  [3.300]  [3.302]  Depositst−1  3.970***  4.032***  4.037***  3.829***  3.890***  3.893***  4.027***  4.044***  4.046***  3.800***  3.892***  3.891***  [0.636]  [0.626]  [0.626]  [0.629]  [0.620]  [0.619]  [0.648]  [0.652]  [0.651]  [0.715]  [0.689]  [0.689]  CB fundingt−1  6.737***  6.801***  6.813***  6.688***  6.712***  6.713***  6.767***  6.811***  6.799***  6.455***  6.258***  6.265***  [1.950]  [1.977]  [1.977]  [1.939]  [1.953]  [1.955]  [1.941]  [1.965]  [1.966]  [2.012]  [2.022]  [2.021]  Constant  –2.003***  –1.986***  –1.978***  –2.011***  –1.993***  –1.986***  –2.000***  –2.000***  –1.992***  –2.010***  –1.980***  –1.972***  [0.062]  [0.062]  [0.062]  [0.063]  [0.064]  [0.064]  [0.061]  [0.061]  [0.061]  [0.062]  [0.063]  [0.063]  Bank-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Firm-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  R-squared  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  0.001  Difference in change in domestic bank lending in Germany between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1  –0.0012  0.0006  0.0002                               €1 billion in US real-estate exposuret−1           –0.0035  –0.0015  –0.0019                      €100 million in US subprime exposuret−1                    –0.0006  0.0002  0.0002             €1 billion in conduit exposuret−1                             –0.0041  –0.0012  –0.0017  Additional difference in change in domestic bank lending in Germany following a five-index point decrease in US homeprices between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1     –0.0101  –0.0107                               €1 billion in US real-estate exposuret−1              –0.0119  –0.0125                      €100 million in US subprime exposuret−1                       –0.0038  –0.0037             €1 billion in conduit exposuret−1                                –0.0147  –0.0155  Additional difference in change in domestic bank lending to firms in industries–regions with a 1%p higher insolvency rate in Germany between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1        –0.0126                               €1 billion in US real-estate exposuret−1                 –0.0118                      €100 million in US subprime exposuret−1                          0.0024             €1 billion in conduit exposuret−1                                   –0.0153  Additional difference in change in domestic bank lending to firms in industries–regions with a 1%p higher insolvency rate in Germany following a 5-index point decrease in US homeprices between banks that have the indicated exposure in the USA and that have no such exposure    €1 billion in total US exposuret−1        –0.0324                               €1 billion in US real-estate exposuret−1                 –0.0401                      €100 million in US subprime exposuret−1                          –0.0098             €1 billion in conduit exposuret−1                                   –0.0343  Table VIII. Explaining the change in domestic bank lending in Germany following shocks to exposures in the USA, including firm–size–year-fixed effects The dependent variable is the quarter-on-quarter logarithmic change in domestic lending by banks to firms (Δlog domestic lending) and an ordinary least squares estimation is used. Table I contains all variable definitions. Coefficients are listed in the first row, robust standard errors clustered at bank level are reported in the row below, and the corresponding significance levels are adjacent to the coefficient. “Yes” indicates that the set of fixed effects is included. ***Significant at 1%. **Significant at 5%. *Significant at 10%. Model  1  2  3  4  Log exposure=  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  0.002  0.003  –0.027  –0.349***  [0.019]  [0.033]  [0.025]  [0.081]  Log exposuret−1 * ΔUS homepricest  –0.398  –0.67  –2.683  29.996***  [0.788]  [1.435]  [1.731]  [4.409]  Log exposuret−1 * insolvencyt−1  –0.891  5.928  2.738  –48.725***  [7.595]  [13.393]  [5.870]  [11.480]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1  159.438  753.507  –363.427**  1,399.740  [258.788]  [527.603]  [158.834]  [976.151]  ΔUS homepricest  –19.936***  –19.995***  –18.408***  –25.327***  [3.057]  [3.025]  [2.919]  [3.629]  Insolvencyt−1  42.806***  41.395***  42.099**  51.491***  [15.614]  [15.413]  [16.428]  [15.823]  ΔUS homepricest * insolvencyt−1  –1,539.257**  –1,526.999**  –1,393.142**  –2,047.858***  [654.163]  [640.504]  [663.835]  [551.940]  Constant  Yes  Yes  Yes  Yes  Bank controls  Yes  Yes  Yes  Yes  Bank-fixed effects  Yes  Yes  Yes  Yes  Firm–size–year-fixed effects  Yes  Yes  Yes  Yes  Observations  1,664,262  1,664,262  1,664,262  1,664,262  R-squared  0.028  0.028  0.028  0.028  Model  1  2  3  4  Log exposure=  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  0.002  0.003  –0.027  –0.349***  [0.019]  [0.033]  [0.025]  [0.081]  Log exposuret−1 * ΔUS homepricest  –0.398  –0.67  –2.683  29.996***  [0.788]  [1.435]  [1.731]  [4.409]  Log exposuret−1 * insolvencyt−1  –0.891  5.928  2.738  –48.725***  [7.595]  [13.393]  [5.870]  [11.480]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1  159.438  753.507  –363.427**  1,399.740  [258.788]  [527.603]  [158.834]  [976.151]  ΔUS homepricest  –19.936***  –19.995***  –18.408***  –25.327***  [3.057]  [3.025]  [2.919]  [3.629]  Insolvencyt−1  42.806***  41.395***  42.099**  51.491***  [15.614]  [15.413]  [16.428]  [15.823]  ΔUS homepricest * insolvencyt−1  –1,539.257**  –1,526.999**  –1,393.142**  –2,047.858***  [654.163]  [640.504]  [663.835]  [551.940]  Constant  Yes  Yes  Yes  Yes  Bank controls  Yes  Yes  Yes  Yes  Bank-fixed effects  Yes  Yes  Yes  Yes  Firm–size–year-fixed effects  Yes  Yes  Yes  Yes  Observations  1,664,262  1,664,262  1,664,262  1,664,262  R-squared  0.028  0.028  0.028  0.028  Table VIII. Explaining the change in domestic bank lending in Germany following shocks to exposures in the USA, including firm–size–year-fixed effects The dependent variable is the quarter-on-quarter logarithmic change in domestic lending by banks to firms (Δlog domestic lending) and an ordinary least squares estimation is used. Table I contains all variable definitions. Coefficients are listed in the first row, robust standard errors clustered at bank level are reported in the row below, and the corresponding significance levels are adjacent to the coefficient. “Yes” indicates that the set of fixed effects is included. ***Significant at 1%. **Significant at 5%. *Significant at 10%. Model  1  2  3  4  Log exposure=  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  0.002  0.003  –0.027  –0.349***  [0.019]  [0.033]  [0.025]  [0.081]  Log exposuret−1 * ΔUS homepricest  –0.398  –0.67  –2.683  29.996***  [0.788]  [1.435]  [1.731]  [4.409]  Log exposuret−1 * insolvencyt−1  –0.891  5.928  2.738  –48.725***  [7.595]  [13.393]  [5.870]  [11.480]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1  159.438  753.507  –363.427**  1,399.740  [258.788]  [527.603]  [158.834]  [976.151]  ΔUS homepricest  –19.936***  –19.995***  –18.408***  –25.327***  [3.057]  [3.025]  [2.919]  [3.629]  Insolvencyt−1  42.806***  41.395***  42.099**  51.491***  [15.614]  [15.413]  [16.428]  [15.823]  ΔUS homepricest * insolvencyt−1  –1,539.257**  –1,526.999**  –1,393.142**  –2,047.858***  [654.163]  [640.504]  [663.835]  [551.940]  Constant  Yes  Yes  Yes  Yes  Bank controls  Yes  Yes  Yes  Yes  Bank-fixed effects  Yes  Yes  Yes  Yes  Firm–size–year-fixed effects  Yes  Yes  Yes  Yes  Observations  1,664,262  1,664,262  1,664,262  1,664,262  R-squared  0.028  0.028  0.028  0.028  Model  1  2  3  4  Log exposure=  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  0.002  0.003  –0.027  –0.349***  [0.019]  [0.033]  [0.025]  [0.081]  Log exposuret−1 * ΔUS homepricest  –0.398  –0.67  –2.683  29.996***  [0.788]  [1.435]  [1.731]  [4.409]  Log exposuret−1 * insolvencyt−1  –0.891  5.928  2.738  –48.725***  [7.595]  [13.393]  [5.870]  [11.480]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1  159.438  753.507  –363.427**  1,399.740  [258.788]  [527.603]  [158.834]  [976.151]  ΔUS homepricest  –19.936***  –19.995***  –18.408***  –25.327***  [3.057]  [3.025]  [2.919]  [3.629]  Insolvencyt−1  42.806***  41.395***  42.099**  51.491***  [15.614]  [15.413]  [16.428]  [15.823]  ΔUS homepricest * insolvencyt−1  –1,539.257**  –1,526.999**  –1,393.142**  –2,047.858***  [654.163]  [640.504]  [663.835]  [551.940]  Constant  Yes  Yes  Yes  Yes  Bank controls  Yes  Yes  Yes  Yes  Bank-fixed effects  Yes  Yes  Yes  Yes  Firm–size–year-fixed effects  Yes  Yes  Yes  Yes  Observations  1,664,262  1,664,262  1,664,262  1,664,262  R-squared  0.028  0.028  0.028  0.028  We are particularly interested in the interaction of the exposures with: (1) the change in US home prices to gauge the impact of this incoming shock on the volume of bank lending, on the one hand; and with (2) the change in US home prices and insolvency to gauge the impact of the incoming shock on the composition of bank lending, on the other hand. We estimate different forms of the following specification:   Δlog domestic lendingijt=β1 log exposureit−1+β2Δ US housepricest+β3 insolvencyjt−1+ β4 log exposureit−1 * Δ US housepricest+β5 log exposureit−1 * insolvencyjt−1+ β6 Δ US housepricest *insolvencyjt−1+β7 log exposureit−1 * Δ US housepricest *insolvencyjt−1+ ∑n=1Nβ8nbank controlsit−1n +αi+αj+εijt, (2) where Δlog domestic lendingijt is the growth of domestic lending (measured as the quarter-on-quarter logarithmic change in domestic lending by bank i to firm j in quarter t). In terms of exposure by German banks in the USA, we distinguish between total US exposureit–1, US real-estate exposureit–1, subprime exposureit–1, and conduit exposureit–1. ΔUS homepricest is the change in US home prices while insolvencyjt–1 defines the firm insolvency rate and proxies for firm quality at the industry–region level. The bank controls we feature are: size (log of total assets), capital (leverage ratio), liquidity (short-term assets to total assets), ROA (return on assets), NPL (non-performing loans to total loans), deposits (deposits to total liabilities), and CB funding (central bank funding to total assets).21 All specifications include comprehensive sets of bank and also firm-fixed effects ( αi and αj).22ɛijt is the error term. We employ sets of three specifications for each exposure type, always starting with a simple model without any interactions, a second model with the interaction of exposure and US home prices, and a third model with the double and triple interactions with insolvency. Each third specification, and also the final one which includes the triple interaction term, can help answer one of our main research questions: “Is there a flight to quality in bank lending in Germany when home prices in the US decline, and does the strength of this effect depend on the degree of the German banks’ exposure to real estate, subprime, and conduits in the USA?” 4.2 Control Variables We start by discussing the estimated coefficients on the control variables, after which we turn to the coefficients of main interest on the double and triple interaction terms that include exposures. Among bank control variables, most estimated coefficients have the expected sign but only the estimated coefficients on deposits and central bank funding is statistically significant. Representative estimates in this regard from Model 1, for example, equal 3.970*** and 6.737***, respectively.23 These estimates imply that a one percentage point increase in the deposit ratio increases the growth in lending by 0.04 percentage points, and that a one percentage point increase in central bank funding increases it by 0.07 percentage points.24 For comparison, we note that the mean growth of domestic loans across all bank–firm–quarter observations equals –2.5%. These findings suggest that banks that rely on deposits and central bank funding increase lending to corporations. This is not surprising when taking into account the difficulties certain institutions faced in accessing wholesale markets as well as the ECB’s willingness to provide liquidity. The rest of the control variables, although all imprecisely estimated, imply that smaller, capitalized, liquid, and profitable banks increase lending which all corresponds to priors. The positive coefficient for the non-performing loans (also insignificant) can be attributed to a mechanical relationship since non-performing loans are reported in the credit register data and continue to exist until they are written off. Next, we discuss the estimated coefficients on the variables the exposures will be interacted with, that is, the variable that captures the change in US home prices and the variable insolvency. The coefficient for the change in US home prices is negative in all models ranging between –4.058 and –5.234. This implies that a five-index point decrease in US home prices (which is the largest drop that is observed but one that occurs in more than one-quarter of the observations) increases the growth in domestic lending maximum by 0.20 percentage points (=0.05 ×4.058). These estimates suggest that a substitution effect may be taking place whereby home price declines in the USA per se may lead to more lending in Germany overall. However, the coefficient is imprecisely estimated. Although the magnitude of this effect is not that large, our estimates of coefficients on the double and triple interactions with this variable presented below may gain further credence as this substitution effect (as we will see) is overturned when banks have exposures in the USA. With respect to the insolvency ratio of the borrower’s region and industry, it decreases the growth in lending as expected. The interaction term of the change in US home prices and insolvency appears to have an inverse relationship with the change in lending. The coefficients for both terms are negative in all specifications but statistically insignificant. 4.3 Main Estimates on Exposures Let us now turn to the exposures and their interactions. Before investigating the three types that are directly related to the origins of the crisis, we estimate our model with the total exposure to the US real-estate market that is the sum of the three types. We observe in the first three models in Table VII that an increase in total US exposure by itself does not have an impact on the growth in domestic lending. Its interaction with changes in US home prices, for which the coefficient is estimated to equal 0.978***, however implies that following a decrease by five-index points in the S&P/Case–Shiller US National Home Price Index, a bank with a €1 billion total exposure to the USA is estimated to contract its quarterly lending in Germany by 1.01 percentage points more than a bank with no such exposure. This is a large effect given that the mean (median) quarterly loan growth during the sample period equals –2.48 (–0.71)% or when considering that going in 2 years from peak to trough in home prices in the USA (i.e., a drop by 60 index points or twelve times the five-index points assessment provided above) by itself would result in an extra cut in credit of more than 12 percentage points for firms borrowing from these exposed banks. The lower panel in Table VII further details the economic relevancy assessment calculations.25 We also provide a vivid graphical illustration of these economic effects in Figures 2 and 3. Figure 2. View largeDownload slide Change in the composition of domestic bank lending in Germany by bank exposure in the USA. The figure shows the difference in the change in domestic bank lending in Germany between banks that have the indicated exposure in the USA and that have no such exposure. The differences are based on the estimates of the coefficients in Table VII models 3 (total), 6 (real-estate), 9 (subprime), and 12 (conduit). Only arrows in the lower panel that are starred represent coefficients that are statistically significant at the 10% level. Figure 2. View largeDownload slide Change in the composition of domestic bank lending in Germany by bank exposure in the USA. The figure shows the difference in the change in domestic bank lending in Germany between banks that have the indicated exposure in the USA and that have no such exposure. The differences are based on the estimates of the coefficients in Table VII models 3 (total), 6 (real-estate), 9 (subprime), and 12 (conduit). Only arrows in the lower panel that are starred represent coefficients that are statistically significant at the 10% level. Figure 3. View largeDownload slide Change in the composition of domestic bank lending in Germany by bank exposure in the USA following a 5-index point decrease in US homeprices. The figure shows the difference in the change in domestic bank lending in Germany between banks that have the indicated exposure in the USA and that have no such exposure following a 5-index point decrease in US homeprices. The differences are based on the estimates of the coefficients in Table VII models 3 (total), 6 (real-estate), 9 (subprime), and 12 (conduit). Only arrows in the lower panel that are starred represent coefficients that are statistically significant at the 10% level. Figure 3. View largeDownload slide Change in the composition of domestic bank lending in Germany by bank exposure in the USA following a 5-index point decrease in US homeprices. The figure shows the difference in the change in domestic bank lending in Germany between banks that have the indicated exposure in the USA and that have no such exposure following a 5-index point decrease in US homeprices. The differences are based on the estimates of the coefficients in Table VII models 3 (total), 6 (real-estate), 9 (subprime), and 12 (conduit). Only arrows in the lower panel that are starred represent coefficients that are statistically significant at the 10% level. This is our first main finding: German bank exposure to the US real-estate market, and the possible losses emanating there as real-estate prices in the US sagged, substantially contracted bank lending in Germany. This presents a direct link in terms of credit volume. Next, in Model 3 we interact insolvency as a measure of ex ante credit risk at the industry–region level with the aforementioned terms. The estimated coefficient on the double interaction term of exposure and insolvency equals –6.095*, while the estimated coefficient on the triple interaction term of exposure, US home prices, and insolvency equals 312.665**. The triple interaction implies that, following a decrease by five-index points in US home prices, a bank with a €1 billion total US exposure contracts its quarterly lending to firms in Germany in riskier industry–region combinations (i.e., those with a 1 percentage point higher insolvency rate) by an additional 3.24 percentage points (=312.665 × Ln(1 billion) × 0.05 × 0.01/100) more than a bank with no such exposure.26 Or, going in 2 years from peak to trough in home prices in the USA by itself results in an extra cut in credit of almost 39 percentage points for riskier firms borrowing in Germany from these US real-estate exposed banks. This is a large cut in access to credit for these riskier firms that may be almost totally reliant on bank credit for their financing needs. Hence, this is our second main finding: German bank exposure to the US real-estate market overall and the possible losses emanating there as real-estate prices in the US sagged, substantially shifted bank lending in Germany. A direct link in terms of credit composition, and clear evidence for a flight to quality. Total US exposure is defined as the sum of the exposures to the US real-estate market and consists of direct lending to the real-estate sector and the subprime lenders, and the indirect conduit exposure. However, we would also like to focus on each particular type of exposure that is linked to different aspects of the problems in the real-estate sector in the USA. Therefore, it is of great interest to examine a direct exposure to the US real-estate sector in order to have a cleaner measure. We first observe in Models 4, 5, and 6 in Table VII that US real-estate exposure per se does not explain much of the changes in domestic lending, but that its interaction with US home prices in Model 5 strongly does. In the latter model, the estimated coefficient on the interaction equals 1.148***. This estimate implies that a bank with a €1 billion exposure to the US real-estate sector, and following a decrease by five-index points in the S&P/Case–Shiller US National Home Price Index, is estimated to contract its quarterly lending in Germany by 1.19 percentage points more than a bank with no such exposure.27 This effect is slightly larger than the estimated coefficient in the previous set of exercises with total US exposure. The coefficient on the double interaction term of exposure and insolvency has a larger magnitude (–5.702); however, it is imprecisely estimated. The coefficient on the triple interaction term of exposure, US home prices, and insolvency equals 386.937**. Following a decrease by five index points in US home prices, a bank with a €1 billion exposure to US real estate is estimated to contract its quarterly lending in Germany to riskier firms by an additional 4.01 percentage points more than a bank with no such exposure. This is clearly a larger economic effect compared with the one found for total exposure to the US real-estate market suggesting that the direct link in terms of credit composition exists and may comprise a large part of the exposure effect. In Models 7–9 in Table VII, we replace real estate with subprime exposure (which, as mentioned previously, is to subprime lenders and distinct from real-estate exposure). All relevant estimated coefficients are imprecisely estimated probably due to the smaller amounts of exposures involved. Yet, the signs of most coefficients are as expected. For example, the estimates in Model 9 imply that a bank with a €100 million exposure to subprime lenders, and following a decrease by five-index points in US home prices, contracts its quarterly lending in Germany by 0.38 percentage points overall, and to riskier firms by 0.98 percentage points more than a bank with no such exposure. Hence, once more credit volume and composition in Germany are affected by the possible losses that emanate from exposures combined with home price declines in the USA. Finally, in Models 10–12 in Table VII we introduce conduit exposure, which is very large on average. Indeed, the liquidity potentially provided to conduits is three times as high as US real-estate exposure on average, and much larger than the amount lent to subprime lenders in our sample. However, we do not find evidence to argue that conduit exposure itself has an impact on lending in Germany. Yet the estimates in Model 11 show that a contraction in domestic lending is again spurred by US home prices dropping. A bank with a €1 billion exposure cuts lending by 1.47 percentage points more following a decrease by five-index points in US home prices than a bank with zero exposure.28 Finally, in Model 12 the estimated additional coefficients further imply that a bank with €1 billion in US conduits contracts its quarterly lending to riskier firms in Germany by 1.53 percentage points more than banks without conduits in place, with the additional home price effect resulting in a contraction of 3.43 percentage points. In sum, credit volume and composition in Germany are affected by the possible losses that emanate from exposures, combined with US home price declines, on real-estate, subprime, and conduits in the USA. 4.4 Further Controlling for Demand Effects Our identification strategy relies on bank–time level variation in exposures in the USA, coupled with firm-fixed effects that account for firm-level demand in Germany. However, borrowers may potentially contract their expenditures and reduce their demand for loans over time. In order to show the change in the credit amount for the same firm borrowing from multiple banks as in Khwaja and Mian (2008), we have to control for time varying firm-level demand. Due to the lack of multiplicity in relationships, that is, few German firms engage multiple banks that are differentiated by their exposures in the USA, including firm–quarter-fixed effects removes all the variation we are interested in. We employ two exercises to account for time varying firm-level demand. In Table VIII, we generate “firm–size times year”-fixed effects (as in, e.g., Acharya et al., 2016; De Jonghe et al., 2016,, 2017). We proxy firm size with the sum of total bank borrowing at firm level, and use the distribution of this variable to generate ten different percentile dummies. For each size percentile we then generate a set of year-specific fixed effects. We note that we lose significance for the interaction terms of total exposure and US real-estate exposure. However, we do observe a much stronger effect for the conduit exposure. Conduit exposure itself actually leads to a contraction in lending in Germany. This finding likely results from the sudden realization at the onset of the financial crisis that conduits “could come crashing back on the banks’ balance sheets” (actually optimal given potential reputational losses in, e.g., Segura, 2018), and banks taking appropriate action in terms of lending in Germany. The estimated coefficient of –0.349*** in Model 4 implies that a bank with €1 billion in US conduits is estimated to contract its quarterly lending in Germany by 7.23 percentage points more than banks without conduits in place. The double and triple interactions point to the same direction with larger magnitudes than previously estimated. However, the coefficient for the triple interaction is no longer statistically significant. Following Schnabl (2012), we include only firms that borrow from at least three lenders and re-estimate our model with firm–year-fixed effects to control for time-varying firm demand. We present the results in Table IX. Table IX. Explaining the change in domestic bank lending in Germany following shocks to exposures in the USA, controlling for firm demand by employing only firms with multiple lenders The dependent variable is the quarter-on-quarter logarithmic change in domestic lending by banks to firms (Δlog domestic lending) and an ordinary least squares estimation is used. Table II contains all variable definitions. Coefficients are listed in the first row, robust standard errors clustered at bank level are reported in the row below, and the corresponding significance levels are adjacent to the coefficient. “Yes” indicates that the set of fixed effects is included. ***Significant at 1%. **Significant at 5%. *Significant at 10%. Model  1  2  3  4  Log exposure =  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  –0.042  –0.069**  0.028  –0.028  [0.026]  [0.029]  [0.019]  [0.047]  Log exposuret−1 * ΔUS homepricest  2.361*  2.630**  3.776*  2.311  [1.222]  [1.267]  [1.984]  [1.454]  Log exposuret−1 * insolvencyt−1  –24.889***  –25.747***  –12.198*  –20.306**  [7.391]  [8.042]  [7.237]  [8.080]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1  942.356***  1,042.101***  496.898  781.465*  [330.766]  [348.612]  [496.908]  [426.614]  ΔUS homepricest  –9.192  –9.02  –12.246  –10.238  [11.062]  [10.703]  [10.537]  [10.937]  Insolvencyt–1  56.455  63.568  75.819  62.51  [66.827]  [70.657]  [54.839]  [67.405]  ΔUS homepricest * insolvencyt−1  –775.084  –891.038  –1,379.58  –962.619  [3,295.780]  [3,390.294]  [3,268.091]  [3,435.433]  Constant  Yes  Yes  Yes  Yes  Bank controls  Yes  Yes  Yes  Yes  Bank-fixed effects  Yes  Yes  Yes  Yes  Firm–year-fixed effects  Yes  Yes  Yes  Yes  Observations  258,040  258,040  258,040  258,040  R-squared  0.003  0.003  0.003  0.003  Model  1  2  3  4  Log exposure =  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  –0.042  –0.069**  0.028  –0.028  [0.026]  [0.029]  [0.019]  [0.047]  Log exposuret−1 * ΔUS homepricest  2.361*  2.630**  3.776*  2.311  [1.222]  [1.267]  [1.984]  [1.454]  Log exposuret−1 * insolvencyt−1  –24.889***  –25.747***  –12.198*  –20.306**  [7.391]  [8.042]  [7.237]  [8.080]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1  942.356***  1,042.101***  496.898  781.465*  [330.766]  [348.612]  [496.908]  [426.614]  ΔUS homepricest  –9.192  –9.02  –12.246  –10.238  [11.062]  [10.703]  [10.537]  [10.937]  Insolvencyt–1  56.455  63.568  75.819  62.51  [66.827]  [70.657]  [54.839]  [67.405]  ΔUS homepricest * insolvencyt−1  –775.084  –891.038  –1,379.58  –962.619  [3,295.780]  [3,390.294]  [3,268.091]  [3,435.433]  Constant  Yes  Yes  Yes  Yes  Bank controls  Yes  Yes  Yes  Yes  Bank-fixed effects  Yes  Yes  Yes  Yes  Firm–year-fixed effects  Yes  Yes  Yes  Yes  Observations  258,040  258,040  258,040  258,040  R-squared  0.003  0.003  0.003  0.003  Table IX. Explaining the change in domestic bank lending in Germany following shocks to exposures in the USA, controlling for firm demand by employing only firms with multiple lenders The dependent variable is the quarter-on-quarter logarithmic change in domestic lending by banks to firms (Δlog domestic lending) and an ordinary least squares estimation is used. Table II contains all variable definitions. Coefficients are listed in the first row, robust standard errors clustered at bank level are reported in the row below, and the corresponding significance levels are adjacent to the coefficient. “Yes” indicates that the set of fixed effects is included. ***Significant at 1%. **Significant at 5%. *Significant at 10%. Model  1  2  3  4  Log exposure =  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  –0.042  –0.069**  0.028  –0.028  [0.026]  [0.029]  [0.019]  [0.047]  Log exposuret−1 * ΔUS homepricest  2.361*  2.630**  3.776*  2.311  [1.222]  [1.267]  [1.984]  [1.454]  Log exposuret−1 * insolvencyt−1  –24.889***  –25.747***  –12.198*  –20.306**  [7.391]  [8.042]  [7.237]  [8.080]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1  942.356***  1,042.101***  496.898  781.465*  [330.766]  [348.612]  [496.908]  [426.614]  ΔUS homepricest  –9.192  –9.02  –12.246  –10.238  [11.062]  [10.703]  [10.537]  [10.937]  Insolvencyt–1  56.455  63.568  75.819  62.51  [66.827]  [70.657]  [54.839]  [67.405]  ΔUS homepricest * insolvencyt−1  –775.084  –891.038  –1,379.58  –962.619  [3,295.780]  [3,390.294]  [3,268.091]  [3,435.433]  Constant  Yes  Yes  Yes  Yes  Bank controls  Yes  Yes  Yes  Yes  Bank-fixed effects  Yes  Yes  Yes  Yes  Firm–year-fixed effects  Yes  Yes  Yes  Yes  Observations  258,040  258,040  258,040  258,040  R-squared  0.003  0.003  0.003  0.003  Model  1  2  3  4  Log exposure =  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  –0.042  –0.069**  0.028  –0.028  [0.026]  [0.029]  [0.019]  [0.047]  Log exposuret−1 * ΔUS homepricest  2.361*  2.630**  3.776*  2.311  [1.222]  [1.267]  [1.984]  [1.454]  Log exposuret−1 * insolvencyt−1  –24.889***  –25.747***  –12.198*  –20.306**  [7.391]  [8.042]  [7.237]  [8.080]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1  942.356***  1,042.101***  496.898  781.465*  [330.766]  [348.612]  [496.908]  [426.614]  ΔUS homepricest  –9.192  –9.02  –12.246  –10.238  [11.062]  [10.703]  [10.537]  [10.937]  Insolvencyt–1  56.455  63.568  75.819  62.51  [66.827]  [70.657]  [54.839]  [67.405]  ΔUS homepricest * insolvencyt−1  –775.084  –891.038  –1,379.58  –962.619  [3,295.780]  [3,390.294]  [3,268.091]  [3,435.433]  Constant  Yes  Yes  Yes  Yes  Bank controls  Yes  Yes  Yes  Yes  Bank-fixed effects  Yes  Yes  Yes  Yes  Firm–year-fixed effects  Yes  Yes  Yes  Yes  Observations  258,040  258,040  258,040  258,040  R-squared  0.003  0.003  0.003  0.003  Table X. Explaining the change in domestic bank lending in Germany following shocks to exposures in the USA, for alternative shock measures The dependent variable is the quarter-on-quarter logarithmic change in domestic lending by banks to firms (Δlog domestic lending) and an ordinary least squares estimation is used. Dummy homeprices is equal to one after the second quarter of the year 2006, i.e., when home prices started to drop, and equal to zero otherwise. Dummy ABCP market is equal to one after the second quarter of 2007, and equal to zero otherwise, and captures the shock to the banks with conduit exposures. Table I contains all other variable definitions. In our estimations, the measurement for insolvency, capital, liquidity, ROA, NPL, deposits, and CB funding are in ratios. Coefficients are listed in the first row, robust standard errors clustered at bank level are reported in the row below, and the corresponding significance levels are adjacent to the coefficient. “Yes” indicates that the set of fixed effects is included. ***Significant at 1%. **Significant at 5%. *Significant at 10%. Model  1  2  3  4  5  6  7  8  9  Log exposure =  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure     Dummy shock =  Dummy homeprices     Dummy homeprices     Dummy ABCP market     Log exposuret−1  –0.033*  –0.02  –0.019  0.012  0.047***  0.050***  –0.017  –0.016  –0.016  [0.018]  [0.019]  [0.020]  [0.008]  [0.014]  [0.015]  [0.012]  [0.011]  [0.011]  Log exposuret−1 * dummy shockt     –0.062*  –0.067*     –0.152***  –0.160***     –0.012  –0.014     [0.034]  [0.035]     [0.041]  [0.045]     [0.028]  [0.028]  Log exposuret−1 * insolvencyt−1        –6.006        –6.29        –6.442*        [4.809]        [6.527]        [3.763]  Log exposuret−1 * dummy shockt * insolvencyt−1        7.618        31.018        1.258        [15.955]        [21.985]        [10.837]  Dummy shockt  –2.691***  –2.717***  –2.785***  –2.694***  –2.813***  –2.885***  –2.019***  –2.020***  –2.056***  [0.220]  [0.214]  [0.216]  [0.222]  [0.238]  [0.242]  [0.234]  [0.231]  [0.234]  Insolvencyt−1  –29.92  –30.101  –78.088***  –30.275  –30.206  –82.779***  –30.428  –30.505  7.059  [19.760]  [19.741]  [23.128]  [19.833]  [19.749]  [23.573]  [19.769]  [19.715]  [20.499]  Dummy shockt * insolvencyt−1        228.977***        248.198***        –247.375***        [65.552]        [68.088]        [51.922]  Sizet−1  –0.027  –0.025  –0.025  –0.120*  –0.123*  –0.124*  –0.067  –0.065  –0.066  [0.057]  [0.056]  [0.056]  [0.064]  [0.065]  [0.065]  [0.072]  [0.073]  [0.073]  Capitalt−1  12.387*  12.311*  12.317*  14.676**  14.843**  14.899**  12.860*  12.781*  12.751*  [7.331]  [7.323]  [7.325]  [7.033]  [7.004]  [7.007]  [7.176]  [7.103]  [7.095]  Liquidityt−1  1.827  1.799  1.808  1.604  1.679  1.685  2.239**  2.208**  2.199*  [1.148]  [1.150]  [1.148]  [1.166]  [1.172]  [1.170]  [1.133]  [1.125]  [1.122]  ROAt−1  –3.509  –3.961  –4.142  0.917  3.132  2.819  1.379  1.225  1.392  [25.002]  [24.841]  [24.808]  [25.343]  [24.141]  [24.223]  [24.668]  [24.644]  [24.555]  NPLt−1  –2.08  –2.151  –2.19  –2.228  –2.397  –2.462  –1.501  –1.452  –1.53  [3.030]  [3.039]  [3.037]  [3.087]  [3.025]  [3.024]  [3.018]  [2.999]  [2.991]  Depositst−1  2.760***  2.775***  2.752***  3.200***  3.122***  3.101***  2.491***  2.514***  2.501***  [0.581]  [0.577]  [0.578]  [0.657]  [0.665]  [0.667]  [0.706]  [0.682]  [0.683]  CB fundingt−1  7.717***  7.758***  7.746***  7.811***  7.801***  7.754***  5.543***  5.547***  5.514***  [1.650]  [1.636]  [1.638]  [1.656]  [1.671]  [1.677]  [2.058]  [2.050]  [2.050]  Constant  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Bank-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Firm-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  R-squared  0.002  0.002  0.002  0.002  0.002  0.002  0.002  0.002  0.002  Model  1  2  3  4  5  6  7  8  9  Log exposure =  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure     Dummy shock =  Dummy homeprices     Dummy homeprices     Dummy ABCP market     Log exposuret−1  –0.033*  –0.02  –0.019  0.012  0.047***  0.050***  –0.017  –0.016  –0.016  [0.018]  [0.019]  [0.020]  [0.008]  [0.014]  [0.015]  [0.012]  [0.011]  [0.011]  Log exposuret−1 * dummy shockt     –0.062*  –0.067*     –0.152***  –0.160***     –0.012  –0.014     [0.034]  [0.035]     [0.041]  [0.045]     [0.028]  [0.028]  Log exposuret−1 * insolvencyt−1        –6.006        –6.29        –6.442*        [4.809]        [6.527]        [3.763]  Log exposuret−1 * dummy shockt * insolvencyt−1        7.618        31.018        1.258        [15.955]        [21.985]        [10.837]  Dummy shockt  –2.691***  –2.717***  –2.785***  –2.694***  –2.813***  –2.885***  –2.019***  –2.020***  –2.056***  [0.220]  [0.214]  [0.216]  [0.222]  [0.238]  [0.242]  [0.234]  [0.231]  [0.234]  Insolvencyt−1  –29.92  –30.101  –78.088***  –30.275  –30.206  –82.779***  –30.428  –30.505  7.059  [19.760]  [19.741]  [23.128]  [19.833]  [19.749]  [23.573]  [19.769]  [19.715]  [20.499]  Dummy shockt * insolvencyt−1        228.977***        248.198***        –247.375***        [65.552]        [68.088]        [51.922]  Sizet−1  –0.027  –0.025  –0.025  –0.120*  –0.123*  –0.124*  –0.067  –0.065  –0.066  [0.057]  [0.056]  [0.056]  [0.064]  [0.065]  [0.065]  [0.072]  [0.073]  [0.073]  Capitalt−1  12.387*  12.311*  12.317*  14.676**  14.843**  14.899**  12.860*  12.781*  12.751*  [7.331]  [7.323]  [7.325]  [7.033]  [7.004]  [7.007]  [7.176]  [7.103]  [7.095]  Liquidityt−1  1.827  1.799  1.808  1.604  1.679  1.685  2.239**  2.208**  2.199*  [1.148]  [1.150]  [1.148]  [1.166]  [1.172]  [1.170]  [1.133]  [1.125]  [1.122]  ROAt−1  –3.509  –3.961  –4.142  0.917  3.132  2.819  1.379  1.225  1.392  [25.002]  [24.841]  [24.808]  [25.343]  [24.141]  [24.223]  [24.668]  [24.644]  [24.555]  NPLt−1  –2.08  –2.151  –2.19  –2.228  –2.397  –2.462  –1.501  –1.452  –1.53  [3.030]  [3.039]  [3.037]  [3.087]  [3.025]  [3.024]  [3.018]  [2.999]  [2.991]  Depositst−1  2.760***  2.775***  2.752***  3.200***  3.122***  3.101***  2.491***  2.514***  2.501***  [0.581]  [0.577]  [0.578]  [0.657]  [0.665]  [0.667]  [0.706]  [0.682]  [0.683]  CB fundingt−1  7.717***  7.758***  7.746***  7.811***  7.801***  7.754***  5.543***  5.547***  5.514***  [1.650]  [1.636]  [1.638]  [1.656]  [1.671]  [1.677]  [2.058]  [2.050]  [2.050]  Constant  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Bank-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Firm-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  R-squared  0.002  0.002  0.002  0.002  0.002  0.002  0.002  0.002  0.002  Table X. Explaining the change in domestic bank lending in Germany following shocks to exposures in the USA, for alternative shock measures The dependent variable is the quarter-on-quarter logarithmic change in domestic lending by banks to firms (Δlog domestic lending) and an ordinary least squares estimation is used. Dummy homeprices is equal to one after the second quarter of the year 2006, i.e., when home prices started to drop, and equal to zero otherwise. Dummy ABCP market is equal to one after the second quarter of 2007, and equal to zero otherwise, and captures the shock to the banks with conduit exposures. Table I contains all other variable definitions. In our estimations, the measurement for insolvency, capital, liquidity, ROA, NPL, deposits, and CB funding are in ratios. Coefficients are listed in the first row, robust standard errors clustered at bank level are reported in the row below, and the corresponding significance levels are adjacent to the coefficient. “Yes” indicates that the set of fixed effects is included. ***Significant at 1%. **Significant at 5%. *Significant at 10%. Model  1  2  3  4  5  6  7  8  9  Log exposure =  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure     Dummy shock =  Dummy homeprices     Dummy homeprices     Dummy ABCP market     Log exposuret−1  –0.033*  –0.02  –0.019  0.012  0.047***  0.050***  –0.017  –0.016  –0.016  [0.018]  [0.019]  [0.020]  [0.008]  [0.014]  [0.015]  [0.012]  [0.011]  [0.011]  Log exposuret−1 * dummy shockt     –0.062*  –0.067*     –0.152***  –0.160***     –0.012  –0.014     [0.034]  [0.035]     [0.041]  [0.045]     [0.028]  [0.028]  Log exposuret−1 * insolvencyt−1        –6.006        –6.29        –6.442*        [4.809]        [6.527]        [3.763]  Log exposuret−1 * dummy shockt * insolvencyt−1        7.618        31.018        1.258        [15.955]        [21.985]        [10.837]  Dummy shockt  –2.691***  –2.717***  –2.785***  –2.694***  –2.813***  –2.885***  –2.019***  –2.020***  –2.056***  [0.220]  [0.214]  [0.216]  [0.222]  [0.238]  [0.242]  [0.234]  [0.231]  [0.234]  Insolvencyt−1  –29.92  –30.101  –78.088***  –30.275  –30.206  –82.779***  –30.428  –30.505  7.059  [19.760]  [19.741]  [23.128]  [19.833]  [19.749]  [23.573]  [19.769]  [19.715]  [20.499]  Dummy shockt * insolvencyt−1        228.977***        248.198***        –247.375***        [65.552]        [68.088]        [51.922]  Sizet−1  –0.027  –0.025  –0.025  –0.120*  –0.123*  –0.124*  –0.067  –0.065  –0.066  [0.057]  [0.056]  [0.056]  [0.064]  [0.065]  [0.065]  [0.072]  [0.073]  [0.073]  Capitalt−1  12.387*  12.311*  12.317*  14.676**  14.843**  14.899**  12.860*  12.781*  12.751*  [7.331]  [7.323]  [7.325]  [7.033]  [7.004]  [7.007]  [7.176]  [7.103]  [7.095]  Liquidityt−1  1.827  1.799  1.808  1.604  1.679  1.685  2.239**  2.208**  2.199*  [1.148]  [1.150]  [1.148]  [1.166]  [1.172]  [1.170]  [1.133]  [1.125]  [1.122]  ROAt−1  –3.509  –3.961  –4.142  0.917  3.132  2.819  1.379  1.225  1.392  [25.002]  [24.841]  [24.808]  [25.343]  [24.141]  [24.223]  [24.668]  [24.644]  [24.555]  NPLt−1  –2.08  –2.151  –2.19  –2.228  –2.397  –2.462  –1.501  –1.452  –1.53  [3.030]  [3.039]  [3.037]  [3.087]  [3.025]  [3.024]  [3.018]  [2.999]  [2.991]  Depositst−1  2.760***  2.775***  2.752***  3.200***  3.122***  3.101***  2.491***  2.514***  2.501***  [0.581]  [0.577]  [0.578]  [0.657]  [0.665]  [0.667]  [0.706]  [0.682]  [0.683]  CB fundingt−1  7.717***  7.758***  7.746***  7.811***  7.801***  7.754***  5.543***  5.547***  5.514***  [1.650]  [1.636]  [1.638]  [1.656]  [1.671]  [1.677]  [2.058]  [2.050]  [2.050]  Constant  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Bank-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Firm-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  R-squared  0.002  0.002  0.002  0.002  0.002  0.002  0.002  0.002  0.002  Model  1  2  3  4  5  6  7  8  9  Log exposure =  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure     Dummy shock =  Dummy homeprices     Dummy homeprices     Dummy ABCP market     Log exposuret−1  –0.033*  –0.02  –0.019  0.012  0.047***  0.050***  –0.017  –0.016  –0.016  [0.018]  [0.019]  [0.020]  [0.008]  [0.014]  [0.015]  [0.012]  [0.011]  [0.011]  Log exposuret−1 * dummy shockt     –0.062*  –0.067*     –0.152***  –0.160***     –0.012  –0.014     [0.034]  [0.035]     [0.041]  [0.045]     [0.028]  [0.028]  Log exposuret−1 * insolvencyt−1        –6.006        –6.29        –6.442*        [4.809]        [6.527]        [3.763]  Log exposuret−1 * dummy shockt * insolvencyt−1        7.618        31.018        1.258        [15.955]        [21.985]        [10.837]  Dummy shockt  –2.691***  –2.717***  –2.785***  –2.694***  –2.813***  –2.885***  –2.019***  –2.020***  –2.056***  [0.220]  [0.214]  [0.216]  [0.222]  [0.238]  [0.242]  [0.234]  [0.231]  [0.234]  Insolvencyt−1  –29.92  –30.101  –78.088***  –30.275  –30.206  –82.779***  –30.428  –30.505  7.059  [19.760]  [19.741]  [23.128]  [19.833]  [19.749]  [23.573]  [19.769]  [19.715]  [20.499]  Dummy shockt * insolvencyt−1        228.977***        248.198***        –247.375***        [65.552]        [68.088]        [51.922]  Sizet−1  –0.027  –0.025  –0.025  –0.120*  –0.123*  –0.124*  –0.067  –0.065  –0.066  [0.057]  [0.056]  [0.056]  [0.064]  [0.065]  [0.065]  [0.072]  [0.073]  [0.073]  Capitalt−1  12.387*  12.311*  12.317*  14.676**  14.843**  14.899**  12.860*  12.781*  12.751*  [7.331]  [7.323]  [7.325]  [7.033]  [7.004]  [7.007]  [7.176]  [7.103]  [7.095]  Liquidityt−1  1.827  1.799  1.808  1.604  1.679  1.685  2.239**  2.208**  2.199*  [1.148]  [1.150]  [1.148]  [1.166]  [1.172]  [1.170]  [1.133]  [1.125]  [1.122]  ROAt−1  –3.509  –3.961  –4.142  0.917  3.132  2.819  1.379  1.225  1.392  [25.002]  [24.841]  [24.808]  [25.343]  [24.141]  [24.223]  [24.668]  [24.644]  [24.555]  NPLt−1  –2.08  –2.151  –2.19  –2.228  –2.397  –2.462  –1.501  –1.452  –1.53  [3.030]  [3.039]  [3.037]  [3.087]  [3.025]  [3.024]  [3.018]  [2.999]  [2.991]  Depositst−1  2.760***  2.775***  2.752***  3.200***  3.122***  3.101***  2.491***  2.514***  2.501***  [0.581]  [0.577]  [0.578]  [0.657]  [0.665]  [0.667]  [0.706]  [0.682]  [0.683]  CB fundingt−1  7.717***  7.758***  7.746***  7.811***  7.801***  7.754***  5.543***  5.547***  5.514***  [1.650]  [1.636]  [1.638]  [1.656]  [1.671]  [1.677]  [2.058]  [2.050]  [2.050]  Constant  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Bank-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Firm-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year-fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  1,664,262  R-squared  0.002  0.002  0.002  0.002  0.002  0.002  0.002  0.002  0.002  Our main finding remains unchanged: Banks with a higher exposure to the US real-estate sector cut their lending to German firms by more following a decrease in US home prices than banks that do not have such exposure. In this sample, also banks with a higher subprime exposure cut back domestic lending when US home prices go down. Moreover, we observe that increases in all types of exposures lead to an increased lending to less risky industry–region combinations. Finally, the triple interaction terms confirm the shift in lending for banks with higher total US exposure, real-estate exposure, and for banks that provided more liquidity to conduits. 4.5 Alternative Shock Measures One may argue that our three measures of exposures may be related to different shocks that occurred at different points in time. For instance, the third quarter of 2007 can be considered as a specific shock to the ABCP market when risk-averse investors started to avoid purchasing these commercial paper instruments (Acharya and Schnabl, 2010; Kacperczyk and Schnabl, 2010). However, we choose to employ a measure that flags problems the first and is also commonly considered the root cause of all ensuing problems, that is, the turning point in home prices. As a robustness test, we employ an alternative approach and replace our continuous variable of the change in home prices with two newly created dummies as presented in Table X. The first one, Dummy Homeprices, is equal to one after the second quarter of the year 2006, that is, when home prices started to drop, and equal to zero otherwise. This dummy variable is interacted with the US real-estate exposure and subprime exposure. The second dummy, Dummy ABCP Market, is equal to one after the second quarter of 2007, and equal to zero otherwise, and captures the shock to the banks with conduit exposures. As before we find that both measures negatively determine the change in lending. The interaction term of the Dummy Homeprices with the exposures is inversely related to domestic lending, and this time seems to matter for the subprime exposure too. In other words, banks with US real-estate exposure and subprime exposure cut back lending once US home prices start to decline. All remaining coefficients of interest have the expected sign but imprecisely estimated possibly due to lower variation in the newly introduced dummies. 4.6 Alternative Explanations 4.6.a. Strategic nature of German bank exposures in the USA Our identification strategy relies on the timing of the shock, that is, the exogenous changes in US home prices, and how these interact with bank exposures to the US real-estate market. But banks may have strategically chosen to expose themselves to the US real-estate sector, subprime lenders, and conduits. To deal with this issue we employ so far a strategy common in the literature by lagging bank exposures [à laKashyap and Stein (2000) and Jiménez et al. (2012), for example]. We also include bank-fixed effects to mop all observable and unobservable time-invariant bank heterogeneity and many bank-level variables to account for as much time-varying bank heterogeneity as possible, in this way accounting for their risk-taking incentives in the US real-estate market as well. In this section, we discuss what we do more. First, we start by controlling for a regional Herfindahl–Hirschmann Index (HHI) in Germany since regional bank competition can affect bank strategy in a time-varying manner. We present the mostly unaffected results in Table A1 in an Online Appendix. Next, we tighten pre-determination by lagging bank exposures by four quarters rather than by one quarter to further mitigate any impact of anticipation banks may have. The unaffected results are in Table A2 in the Online Appendix. Then, we instrument current exposures by past exposures and the one year lagged regional loan concentration (HHI). The unaffected results are in Table A3 in the Online Appendix. Finally, we saturate specifications with bank–year-fixed effects. Results are added to Table A3 in the Online Appendix as well.29 4.6.b. Bank type First, we account for spillovers to savings banks, which were indirectly exposed to subprime mortgages through their holdings in Landesbanks (Puri, Rocholl, and Steffen, 2011). This implies that a savings bank (without any exposure to the US real-estate market itself) would cut back lending because of the Landesbank’s risky portfolio. Our research design does not account for this type of transmission mechanism, as our exposure measures are bank specific. In other words, the control group may indeed include banks that are indirectly exposed to treatment. Although we aim to capture the initial shock, given the time period, spillovers to other banks may take place in our sample period, which may lead to an under-estimation of the impact. Therefore, we exclude savings banks from the sample and re-estimate our model. The results, presented in Table A4 in the Online Appendix, support our previous findings. There is still the overall contraction in domestic lending following the US home price shock for banks with higher total US exposure, US real-estate exposure, and conduit exposure. Moreover, these banks then also shift their lending to less riskier industry–region combinations. Next, we examine which exposed banks reduce lending and search for higher quality assets. We split our sample as “well capitalized” and “lowly capitalized banks” (with the split set at 50%) and re-estimate our model. The results imply that well capitalized and exposed banks cut back lending when US home prices start to decline. However, exposed banks with lower capital levels display the “flight to quality” behavior by reducing lending to riskier industry–regions following the decline in home prices. The results are presented in Table A5 in the Online Appendix. Finally, we consider the potential effects of a model-based regulation on credit risk as in Behn, Hasellmann, and Vig (2014). We would like to rule out that our results are driven by the change in lending behavior by German banks with the introduction of the internal ratings-based (IRB) regulation where banks could choose between the new approach and the standard approach and get better risk weights. Information on IRB loans in the credit register is available after 2008. We employ IRB Share defined as IRB loans to total bank loan portfolio lagged by one quarter. We first control for this variable in our model, and then interact it with the exposures, change in US home prices and the insolvency ratio to see if our initial results will be affected. While our main results with the double and triple interactions remain unchanged, we also observe that the IRB loans by banks with the US total exposure and US real-estate exposure experience a further reduction following a decrease in US home prices (Online Appendix Table A6). 4.7 Impact on Firm Borrowing So far, we have documented changes in the composition of domestic lending: Exposed banks cut their lending to domestic firms by more following a decrease in US home prices than banks that do not have such exposure. Moreover, these banks also shift their lending to safer industry–region combinations. We are also interested to see the consequences of the so-called “flight to quality” behavior; in other words, the indirect impact on the real activity. We investigate whether or not firms in riskier industry–region combinations are able to obtain funding from other banks when treated banks decrease lending to those firms. Similar to the approach in Schnabl (2012), we aggregate the data at the borrower firm level and explain total domestic borrowing with weighted bank exposures (weighted by the share of total borrowing from each lender). Table XI presents the estimation results that document a decrease in credit availability for borrowers of exposed banks when US home prices decline or when the borrowers belong to riskier industries and regions.30 This is indicated specifically by the negative and significant coefficient for all exposure types. Moreover, firms borrowing from banks that are exposed to the US real-estate sector experience a higher reduction in borrowing when US home prices go down. Also firms in riskier industry–region combinations that are engaged with banks with exposures to subprime lenders and conduits have an additional reduction in total borrowing. This finding is also in line with Popov and Rocholl (2018) who document that borrowers of affected banks experienced a significant decline in employment and in labor compensation after the crisis. Table XI. Impact on firm borrowing The dependent variable is the quarter-on-quarter logarithmic change in domestic lending by banks to firms (Δlog domestic lending) aggregated at the firm level and an ordinary least squares estimation is used. Table I contains all other variable definitions. In our estimations, the measurement for insolvency, capital, liquidity, ROA, NPL, deposits, and CB funding is in ratios. Coefficients are listed in the first row, robust standard errors clustered at bank level are reported in the row below, and the corresponding significance levels are adjacent to the coefficient. “Yes” indicates that the set of fixed effects is included. ***Significant at 1%. **Significant at 5%. *Significant at 10%. Model  1  2  3  4  Log exposure =  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  –0.342***  –0.446***  –0.156***  –0.480***  [0.013]  [0.016]  [0.011]  [0.015]  Log exposuret−1 * ΔUS homepricest  0.276*  0.444***  –0.042  –0.076  [0.158]  [0.166]  [0.209]  [0.170]  Log exposuret−1 * insolvencyt−1  1.303  3.486  –4.714*  –6.256***  [2.202]  [2.332]  [2.688]  [2.352]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1  75.956  74.564  14.659  4.695  [54.746]  [57.972]  [72.890]  [60.301]  ΔUS homepricest  –15.396***  –14.673***  –13.166***  –16.370***  [2.418]  [2.306]  [2.273]  [2.163]  Insolvencyt−1  –19.56  –32.532  21.626  48.980*  [30.548]  [29.074]  [26.916]  [26.694]  ΔUS homepricest * insolvencyt−1  –1,655.229**  –1,369.947**  –966.956  –703.299  [720.092]  [666.075]  [660.651]  [607.805]  Constant  Yes  Yes  Yes  Yes  Bank controls (aggregated)  Yes  Yes  Yes  Yes  Firm-fixed effects  Yes  Yes  Yes  Yes  Year-fixed effects  Yes  Yes  Yes  Yes  Observations  1,819,124  1,819,124  1,819,124  1,819,124  R-squared  0.140  0.141  0.140  0.141  Model  1  2  3  4  Log exposure =  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  –0.342***  –0.446***  –0.156***  –0.480***  [0.013]  [0.016]  [0.011]  [0.015]  Log exposuret−1 * ΔUS homepricest  0.276*  0.444***  –0.042  –0.076  [0.158]  [0.166]  [0.209]  [0.170]  Log exposuret−1 * insolvencyt−1  1.303  3.486  –4.714*  –6.256***  [2.202]  [2.332]  [2.688]  [2.352]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1  75.956  74.564  14.659  4.695  [54.746]  [57.972]  [72.890]  [60.301]  ΔUS homepricest  –15.396***  –14.673***  –13.166***  –16.370***  [2.418]  [2.306]  [2.273]  [2.163]  Insolvencyt−1  –19.56  –32.532  21.626  48.980*  [30.548]  [29.074]  [26.916]  [26.694]  ΔUS homepricest * insolvencyt−1  –1,655.229**  –1,369.947**  –966.956  –703.299  [720.092]  [666.075]  [660.651]  [607.805]  Constant  Yes  Yes  Yes  Yes  Bank controls (aggregated)  Yes  Yes  Yes  Yes  Firm-fixed effects  Yes  Yes  Yes  Yes  Year-fixed effects  Yes  Yes  Yes  Yes  Observations  1,819,124  1,819,124  1,819,124  1,819,124  R-squared  0.140  0.141  0.140  0.141  Table XI. Impact on firm borrowing The dependent variable is the quarter-on-quarter logarithmic change in domestic lending by banks to firms (Δlog domestic lending) aggregated at the firm level and an ordinary least squares estimation is used. Table I contains all other variable definitions. In our estimations, the measurement for insolvency, capital, liquidity, ROA, NPL, deposits, and CB funding is in ratios. Coefficients are listed in the first row, robust standard errors clustered at bank level are reported in the row below, and the corresponding significance levels are adjacent to the coefficient. “Yes” indicates that the set of fixed effects is included. ***Significant at 1%. **Significant at 5%. *Significant at 10%. Model  1  2  3  4  Log exposure =  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  –0.342***  –0.446***  –0.156***  –0.480***  [0.013]  [0.016]  [0.011]  [0.015]  Log exposuret−1 * ΔUS homepricest  0.276*  0.444***  –0.042  –0.076  [0.158]  [0.166]  [0.209]  [0.170]  Log exposuret−1 * insolvencyt−1  1.303  3.486  –4.714*  –6.256***  [2.202]  [2.332]  [2.688]  [2.352]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1  75.956  74.564  14.659  4.695  [54.746]  [57.972]  [72.890]  [60.301]  ΔUS homepricest  –15.396***  –14.673***  –13.166***  –16.370***  [2.418]  [2.306]  [2.273]  [2.163]  Insolvencyt−1  –19.56  –32.532  21.626  48.980*  [30.548]  [29.074]  [26.916]  [26.694]  ΔUS homepricest * insolvencyt−1  –1,655.229**  –1,369.947**  –966.956  –703.299  [720.092]  [666.075]  [660.651]  [607.805]  Constant  Yes  Yes  Yes  Yes  Bank controls (aggregated)  Yes  Yes  Yes  Yes  Firm-fixed effects  Yes  Yes  Yes  Yes  Year-fixed effects  Yes  Yes  Yes  Yes  Observations  1,819,124  1,819,124  1,819,124  1,819,124  R-squared  0.140  0.141  0.140  0.141  Model  1  2  3  4  Log exposure =  Log total US exposure  Log US real-estate exposure  Log US subprime exposure  Log conduit exposure  Log exposuret−1  –0.342***  –0.446***  –0.156***  –0.480***  [0.013]  [0.016]  [0.011]  [0.015]  Log exposuret−1 * ΔUS homepricest  0.276*  0.444***  –0.042  –0.076  [0.158]  [0.166]  [0.209]  [0.170]  Log exposuret−1 * insolvencyt−1  1.303  3.486  –4.714*  –6.256***  [2.202]  [2.332]  [2.688]  [2.352]  Log exposuret−1 * ΔUS homepricest * insolvencyt−1  75.956  74.564  14.659  4.695  [54.746]  [57.972]  [72.890]  [60.301]  ΔUS homepricest  –15.396***  –14.673***  –13.166***  –16.370***  [2.418]  [2.306]  [2.273]  [2.163]  Insolvencyt−1  –19.56  –32.532  21.626  48.980*  [30.548]  [29.074]  [26.916]  [26.694]  ΔUS homepricest * insolvencyt−1  –1,655.229**  –1,369.947**  –966.956  –703.299  [720.092]  [666.075]  [660.651]  [607.805]  Constant  Yes  Yes  Yes  Yes  Bank controls (aggregated)  Yes  Yes  Yes  Yes  Firm-fixed effects  Yes  Yes  Yes  Yes  Year-fixed effects  Yes  Yes  Yes  Yes  Observations  1,819,124  1,819,124  1,819,124  1,819,124  R-squared  0.140  0.141  0.140  0.141  5. Conclusion Motivated by the seminal works of Peek and Rosengren (1997) and Peek and Rosengren (2000), we study the international transmission of shocks through the German banking sector during the last financial turmoil triggered by the subprime mortgage crisis. In particular, using unique German bank exposure data, we investigate how exposures to the US real-estate market influenced domestic lending in Germany. We are interested in total bank exposure to the US real-estate market and its three salient components: Direct exposures to the US real-estate sector and to the subprime lenders in the USA, and indirect exposure as liquidity provided to ABCP conduits. Confirming previous studies on the transmission of shocks, we first document the overall contraction in lending in Germany following the home price shock. Our main aim is, however, to explore the heterogeneity in the contraction across banks and firms. In other words, we investigate whether differences in bank exposures to the US determine domestic lending in Germany when home prices started to decline in the USA, and whether there is a “flight to quality” in lending for those banks that were more exposed to the US real-estate market. We indeed find that banks with higher total exposure to the US real-estate market and, in particular, with higher exposure to the US real-estate sector and to conduits contract their lending to German firms more following a decrease in US home prices than banks with no such exposure. Moreover, these banks also prefer lending to industry–region combinations with lower insolvency ratios, especially following a decrease in US home prices. To sum up, we mainly document that possible losses abroad shift bank lending at home where the size of the effect depends on the type and the degree of exposure the bank has. Supplementary Material Supplementary data are available at Review of Finance online. Footnotes 1 Allen and Carletti (2013) show that risk shifting and asset substitution can result in a bubble in real-estate prices. 2 Our findings are unlikely to be driven by major regulatory changes as the sample period ends in 2009 whereas Basel III, for instance, developed by the Basel Committee to strengthen the banking sector, was not introduced before December 2010 (Basel Committee on Banking Supervision, 2010). 3 Recent research confirms that, during the global financial crisis, global banks transmitted shocks across borders through their local affiliates (see, among others, Cetorelli and Goldberg, 2011; Cetorelli and Goldberg, 2012; Albertazzi and Bottero, 2013; Claessens and van Horen, 2013; Cull and Martinez Peria, 2013; Allen et al., 2014; Bertay, 2014; de Haas and van Lelyveld, 2014; Ongena, Peydró, and van Horen, 2015; Acharya, Afonso, and Kovner, 2017). 4 Other papers studying the international transmission through the asset side of the banks include De Haas and Van Horen (2013) and Popov and Van Horen (2015). The former paper examines syndicated loans and finds that crisis-related write-downs negatively affected cross-border bank lending. The latter paper finds that large holdings of impaired sovereign debt negatively affected bank lending during the European sovereign debt crisis. Cuñat, Cvijanović, and Yuan (2013) study the domestic transmission of real-estate price shocks within the US using bank balance sheets. Ahrend and Goujard (2015) document that shocks to bank balance-sheets are able to predict systemic banking crises in debtor countries. 5 Not all savings banks were directly exposed to the US real-estate market; however, they were affected through their link with the Landesbanks during the crisis. We, on the other hand, focus on the initial shock and not on the spillovers to other banks. However, since we cover the crisis period as well, we re-estimate our model by excluding the savings banks from our sample. The results remain qualitatively unchanged. 6 On July 31, 2007, the news about the bankruptcy of the two Bear Stearns’ hedge funds invested in subprime mortgages reached the market. On August 7, 2007, the French Bank BNP Paribas pronounced its withdrawals from its three funds due to an inability to judge the “fair” value of their holdings. 7 Acharya, Schnabl, and Suarez (2013) document that it was banks and not investors that were negatively affected because banks had insured outside investors by providing explicit guarantees to conduits. 8 The Financial Times published the list of German banks and their conduits with the amounts of credit facilities provided. From the list, it is possible to observe that not only the big banks but also a lot of Landesbanks had been involved in this kind of risky activity. SachsenLB, LBBW (Landebank Baden–Württemberg), WestLB, and BayernLB were among the Landesbanks with especially large engagements. 9 Details on the credit register can be found in Schmieder (2006), and in published work by Schertler, Buch, and von Westernhagen (2006); Hayden, Porath, and von Westernhagen (2007); and Ongena, Tümer-Alkan, and von Westernhagen (2012), for example. The Bundesbank also maintains a website with papers based on its credit register. Recent prominent examples include, for example, Behn, Haselmann, and Wachtel (2016) and Haselmann, Schoenherr, and Vig (2018). 10 If exposures of €1.5 million or above existed during the reporting period but are partly or fully repaid, the remaining exposure is reported even if the amount is zero. We take the actual amounts of the exposures into consideration. 11 For a more detailed definition of the bank exposures, see Section 19 of the Banking Act (the versions applied before January 1, 2014). The following items are deemed not to be bank exposures: exposures to German central and local governments and communities, securities in the trading stock, undrawn loan commitments, shares in other enterprises, etc. For a more detailed definition on the exceptions of the bank exposures, see Section 20 of the Banking Act (the versions applied before January 1, 2014). 12 For example, lease receivables, mortgage loans, publicly guaranteed loans, and inter-bank loans (with a residual maturity of up to 1 year) were listed separately before 2008 under on-balance sheet activities. Off-balance sheet items included derivatives (other than written option positions), guarantees assumed to cover these and other off-balance sheet transactions (Deutsche Bundesbank, 1998). However, since 2008 the structure of items listed separately has changed to some extent. But the main structure due to which we are able to distinguish between on-balance sheet and off-balance sheet activities has remained unchanged. 13 Our approach is based on separating the pre-merger banks from the merged bank. In the end, we have three banks, which are treated independently from each other. We repeat this procedure as often as a merger takes place. Each time a newly merged bank receives a new identification number, we drop the target banks in that year (or quarter). 14 Ferreira and Gyourko (2015) document that the crisis was more due to prime borrowers, in contrast to the common belief that it was caused by subprime borrowers. We also consider this finding when classifying the exposures in the analysis. We start the analysis by investigating the direct exposure to the US real-estate market that includes prime borrowers as well. After that we assess the role of the exposures to subprime lenders in the USA. 15 During the Eurozone crisis, Greece, Ireland, and Portugal received bail-out funds of €148.6, €54.9, and €61.4 billion, respectively. Another Eurozone member, Spain, experienced problems in the real-estate sector reflected in decreasing home prices. Finally, exposure to offshore centers is an indication of SPV exposures. 16 Since the credit register does not contain exposures to ABCP conduits, we do not include the information on those exposures in Table III. The data on ABCP conduits come from Moody’s ABCP Query and are discussed in Table IV. 17 In robustness, we re-estimate our main models excluding savings banks (and then also cooperative banks). We will discuss the unaffected results later. 18 In this respect, we follow the recent literature (e.g., Jiménez et al., 2014; Ongena, Peydró, and van Horen, 2015). By controlling for bank size, we consider the relative importance of these exposures at the bank level. We also replace absolute exposures with relative exposures (exposure over capital) in unreported estimations. The results remain virtually unchanged. 19 The eighteen top subprime lenders have been identified in the credit register as borrower units. In total, 123 enterprises in the credit register belong to those eighteen top subprime lenders. 20 This is the number of observations of the estimated model that includes all controls. Due to the very large number of firms and banks in the sample we follow customary Stata practice by demeaning first at the firm level and then “absorbing” the fixed effects at the bank level, thereby deflating the R-squares. 21 We use the leverage ratio (total equity over total assets) as a measure of capital constraints. Using regulatory capital as an alternative measure does not affect the results. We also control for bank funding because banks that obtain financing from the wholesale market are found to have more problems during the crisis (de Haas and van Lelyveld, 2014). 22 Because we are mainly interested in the effect of bank-level exposures over time, including bank–time-fixed effects is problematic. Because few firms in Germany rely on multiple banks that are differentiated by their exposures in the USA, including firm–time-fixed effects equally robs the estimations of most if not all of their relevant variation. We discuss these issues further in Section 4.4. 23 ***Significant at 1%, **significant at 5%, and *significant at 10%. For convenience, we will also indicate the significance levels of the estimates that are mentioned further on in the text. 24 To make the reading of our results easier we multiple the estimated coefficients by 100. 25 To assess economic relevancy, we rely on the amounts of €1 billion for total US exposure, real-estate, and conduit exposures and €100 million for subprime exposure. This choice ensures ease and clarity of exposition, but it also broadly respects the absolute and relative order of magnitudes of the standard deviations and means of the exposure variables (Table VI). The standard deviation on real-estate exposure equals €1 billion and on subprime €116 million. On conduits, the standard deviation equals €4.5 billion, while its mean equals €1.5 billion. Finally, recall that five-index points in the S&P/Case–Shiller US National Home Price Index equals around two standard deviations for this index. 26 A one percentage point increase in default rates is often observed during bad times (Jacobson, Lindé, and Roszbach, 2013). We have to divide by 100 here as we had earlier rescaled the coefficients by 100 for easier reading. 27 Notice that for more than half the sample observations, the German banks involved have zero real-estate exposure in the USA, marking these banks to be at once a relevant and ideal control group. 28 Conduits may not have entirely been invested in real estate or at all. For example, “credit arbitrage” ABCP conduits invested heavily in securitized assets, such as asset-backed securities backed by residential mortgages and commercial mortgages, and were consequently more exposed to subprime US residential mortgage loans than other types of conduits. Other ABCP conduits, such as “multi-seller” or “single-seller” conduits, had primarily funded unsecured receivables by the time the financial crisis arrived. It is currently impossible for us to distinguish between the different types of assets present in the conduits. 29 We acknowledge that, in this setting, it is very difficult to find an instrument that would satisfy the exclusion restrictions while being correlated with the endogeneous regressor. However, we are confident that controlling for time-varying bank heterogeneity takes care of these concerns. 30 The number of observations is higher than for our bank–firm analysis because we take an expansive approach to missing observations in the original data by aggregating all available information at the firm level which is appropriate if data availability across firm–bank–quarter combinations is random (which we assess it to be). References Acharya V. V., Afonso G., Kovner A. ( 2017): How do global banks scramble for liquidity? Evidence from the asset-backed commercial paper freeze of 2007, Journal of Financial Intermediation  30, 1– 34. Google Scholar CrossRef Search ADS   Acharya V. V., Eisert T., Eufinger C., Hirsch C. W. ( 2016): Whatever It Takes: The Real Effects of Unconventional Monetary Policy , New York University, New York, NY, Mimeo. Acharya V. V., Schnabl P. ( 2010): Do global banks spread global imbalances? The case of asset-backed commercial paper during the financial crisis of 2007–09, IMF Economic Review  58, 37– 73. Google Scholar CrossRef Search ADS   Acharya V. V., Schnabl P., Suarez G. ( 2013): Securitization without risk transfer, Journal of Financial Economics  107, 515– 536. Google Scholar CrossRef Search ADS   Ahrend R., Goujard A. ( 2015): Global banking, global crises? The role of the bank balance-sheet channel for the transmission of financial crises, European Economic Review  80, 253– 279. Google Scholar CrossRef Search ADS   Albertazzi U., Bottero M. ( 2013): The procyclicality of foreign bank lending: evidence from the global financial crisis . Working paper 926, Bank of Italy, Rome. Allen F., Carletti E. ( 2013): What is systemic risk? Journal of Money, Credit and Banking  45, 121– 127. Google Scholar CrossRef Search ADS   Allen F., Hryckiewicz A., Kowalewski O., Tümer-Alkan G. ( 2014): Transmission of bank liquidity shocks in loan and deposit markets: the role of interbank borrowing and market monitoring, Journal of Financial Stability  15, 112– 126. Google Scholar CrossRef Search ADS   Basel Committee on Banking Supervision. ( 2010): Basel iii: A global regulatory framework for more resilient banks and banking systems. Report, Bank for International Settlements, Basel. Beber A., Brandt M. W., Kavajecz K. A. ( 2009): Flight-to-quality or flight-to-liquidity? Evidence from the euro-area bond market, Review of Financial Studies  22, 925– 957. Google Scholar CrossRef Search ADS   Becker B., Ivashina V. ( 2016): Financial Repression in the European Sovereign Debt Crisis , Swedish House of Finance, Stockholm, Mimeo. Behn M., Hasellmann R., Vig V. ( 2014): The Limits of Model-Based Regulation , SAFE, Frankfurt, Mimeo. Behn M., Haselmann R., Wachtel P. ( 2016): Procyclical capital regulation and lending, Journal of Finance  71, 919– 956. Google Scholar CrossRef Search ADS   Bernanke B. S., Gertler M., Gilchrist S. ( 1996): The financial accelerator and the flight to quality, Review of Economics and Statistics  78, 1– 15. Google Scholar CrossRef Search ADS   Bertay A. ( 2014): The Transmission of Real Estate Shocks through Multinational Banks , Tilburg University, Tilburg, Mimeo. Cetorelli N., Goldberg L. S. ( 2011): Global banks and international shock transmission: evidence from the crisis, IMF Economic Review  59, 41– 76. Google Scholar CrossRef Search ADS   Cetorelli N., Goldberg L. S. ( 2012): Follow the money: quantifying domestic effects of foreign bank shocks in the great recession, American Economic Review: Papers and Proceedings  102, 213– 218. Google Scholar CrossRef Search ADS   Claessens S., van Horen N. ( 2013): Impact of foreign banks, Journal of Financial Perspectives  1, 1– 18. Cull R., Martinez Peria M. S. ( 2013): Bank ownership and lending patterns during the 2008–2009 financial crisis: evidence from Latin America and Eastern Europe, Journal of Banking and Finance  37, 4861– 4878. Google Scholar CrossRef Search ADS   Cuñat V., Cvijanović D., Yuan K. ( 2013): How Did US Banks React to Capital Losses Induced by Real Estate Prices ?, London School of Economics, London, Mimeo. De Haas R., Van Horen N. ( 2013): Running for the exit: international bank lending during a financial crisis, Review of Financial Studies  26, 244– 285. Google Scholar CrossRef Search ADS   de Haas R., van Lelyveld I. ( 2014): Multinational banks and the global financial crisis: weathering the perfect storm?, Journal of Money, Credit and Banking  46, 333– 364. Google Scholar CrossRef Search ADS   De Jonghe O., Degryse H., Jakovljevic S., Mulier K., Schepens G. ( 2017): The Impact of Bank Shocks on Firm Level Outcomes and Bank Risk-Taking , National Bank of Belgium, Brussels, Mimeo. De Jonghe O., Dewachter H., Mulier K., Ongena S., Schepens G. ( 2016): Some Borrowers Are More Equal Than Others: Bank Funding Shocks and Credit Reallocation , National Bank of Belgium, Brussels, Mimeo. Dell’Ariccia G., Laeven L., Suarez G. A. ( 2017): Bank leverage and monetary policy’s risk-taking channel: evidence from the United States, Journal of Finance  72, 613– 654. Google Scholar CrossRef Search ADS   Ferreira F., Gyourko J. ( 2015): A new look at the U.S. foreclosure crisis: panel data evidence of prime and subprime borrowers from 1997 to 2012 . Working paper 21261, National Bureau of Economic Research, Cambridge, MA. Haselmann R. F. H., Schoenherr D., Vig V. ( 2018): Rent-seeking in elite networks, Journal of Political Economy , forthcoming. Hayden E., Porath D., von Westernhagen N. ( 2007): Does diversification improve the performance of German banks? Evidence from individual bank loan portfolios, Journal of Financial Services Research  32, 123– 140. Google Scholar CrossRef Search ADS   Hildebrand T., Rocholl J., Schulz A. ( 2012): Flight to Where? Evidence from Bank Investments during the Financial Crisis , ESMT, Berlin, Mimeo. Ioannidou V. P., Ongena S., Peydró J. L. ( 2015): Monetary policy, risk-taking and pricing: evidence from a quasi-natural experiment, Review of Finance  19, 95– 144. Google Scholar CrossRef Search ADS   Jacobson T., Lindé J., Roszbach K. ( 2013): Firm default and aggregate fluctuations, Journal of the European Economic Association  11, 945– 972. Google Scholar CrossRef Search ADS   Jiménez G., Ongena S., Peydró J. L., Saurina J. ( 2012): Credit supply and monetary policy: identifying the bank balance-sheet channel with loan applications, American Economic Review  102, 2301– 2326. Google Scholar CrossRef Search ADS   Jiménez G., Ongena S., Peydró J. L., Saurina J. ( 2014): Hazardous times for monetary policy: what do twenty-three million bank loans say about the effects of monetary policy on credit risk-taking?, Econometrica  82, 463– 505. Google Scholar CrossRef Search ADS   Kacperczyk M., Schnabl P. ( 2010): When safe proved risky: commercial paper during the financial crisis of 2007–2009, Journal of Economic Perspectives  24, 29– 50. 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   Lang W. W., Nakamura L. I. ( 1995): “ Flight to quality” in banking and economic activity, Journal of Monetary Economics  36, 145– 164. Google Scholar CrossRef Search ADS   Mian A. R., Sufi A. ( 2014): House of Debt: How They (and You) Caused the Great Recession, and How We Can Prevent It from Happening Again , University of Chicago Press, Chicago, IL. Moody's ( 2007): EMEA ABCP Market Exposure to US Residential Mortgages, International Structured Finance, Special Report. Murfin J. ( 2012): The supply-side determinants of loan contract strictness, Journal of Finance  67, 1565– 1601. Google Scholar CrossRef Search ADS   Ongena S., Peydró J. L., van Horen N. ( 2015): Shocks abroad, pain at home? Bank–firm level evidence on financial contagion during the recent financial crisis, IMF Economic Review  63, 698– 750. Google Scholar CrossRef Search ADS   Ongena S., Tümer-Alkan G., von Westernhagen N. ( 2012): Creditor concentration: an empirical investigation, European Economic Review  56, 830– 847. Google Scholar CrossRef Search ADS   Peek J., Rosengren E. S. ( 1997): The international transmission of financial shocks: the case of japan, American Economic Review  87, 495– 505. 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   Popov A., Rocholl J. ( 2018): Do credit shocks affect labor demand? Evidence from employment and wages during the financial crisis, Journal of Financial Intermediation , forthcoming. Popov A. A., Van Horen N. ( 2015): Exporting sovereign stress: evidence from syndicated bank lending during the euro area sovereign debt crisis, Review of Finance  19, 1825– 1866. 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   Schertler A., Buch C. M., von Westernhagen N. ( 2006): Heterogeneity in lending and sectoral growth: evidence from German bank-level data, International Economics and Economic Policy  3, 43– 72. Google Scholar CrossRef Search ADS   Schmieder C. ( 2006): The deutsche bundesbank’s large credit database (bakis–m and mimik), Schmollers Jahrbuch  126, 653– 663. 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   Segura A. ( 2018): Why did sponsor banks rescue their SIVs? A signaling model of rescues, Review of Finance 22, 661– 697. © 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 This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Review of Finance Oxford University Press

Do Exposures to Sagging Real Estate, Subprime, or Conduits Abroad Lead to Contraction and Flight to Quality in Bank Lending at Home?