How does the strength of monetary policy transmission depend on real economic activity?

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. How does the strength of monetary policy transmission depend on real economic activity? Horacio Sapriza and Judit Temesvary 2019-023 Please cite this paper as: Sapriza, Horacio, and Judit Temesvary (2019). “How does the strength of monetary policy transmission depend on real economic activity?,” Finance and Economics Discussion Series 2019-023. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2019.023. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

How does the strength of monetary policy transmission depend on real economic activity?1 Horacio Sapriza2 Judit Temesvary3 Federal Reserve Board Federal Reserve Board This version: February 2019 First version: November 2017 Abstract We study the relationship between the strength of the bank credit channel (BCC) of monetary policy and real GDP growth in the United States using quarterly commercial bank level data between 1986 and 2008. We find that the BCC was significantly stronger during periods of low economic growth. Monetary policy is more effective through this channel in spurring economic activity during periods of low growth, rather than in cooling the economy when growth is high. Furthermore, we find that the BCC operated through a broader range of loan categories and banks than previously documented, underscoring this channel’s economic relevance. JEL codes: E3, E5, G2 Keywords: Bank lending channel; Monetary policy transmission; Bank balance sheet; GDP growth 1 This paper previously circulated under the title “Monetary policy transmission and economic conditions: How does the strength of the bank credit channel depend on real economic activity?” The views expressed in this paper are solely those of the authors, and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System. We are thankful for the assistance of Edward Kim and Akber Khan in compiling the dataset. We also thank Johann Burgstaller, Matias Ossandon Busch, Ricardo Correa, Eric Fischer, Lars Norden, Ann Owen, Teodora Paligorova, Rebecca Zarutskie and seminar participants at the 2018 EFMA, the 2018 Summer Meetings of the IBEFA and the 2018 FMA Annual Meeting for valuable comments and suggestions. 2 Principal economist, Division of Monetary Affairs, Federal Reserve Board, Washington, DC. horacio.sapriza@frb.gov 3 Corresponding author. Principal economist, Division of Monetary Affairs, Federal Reserve Board, Washington, DC. judit.temesvary@frb.gov

1. Introduction In understanding the behavior of the macro economy, it is critical to understand the channels through which monetary policy operates. To what extent does the rise in interest rates impact the lending of U.S. banks? Does the strength of this lending effect depend on economic growth? In this paper, we address these questions in detail. Our findings may provide useful insights into the strength of the bank credit channel in the coming years. To address the above questions, we simultaneously investigate (1) if the bank credit channel was operational in U.S. banks’ domestic lending during the period from 1986 to 2008, and (2) whether economic activity affects the strength of the bank credit channel of monetary policy. Regarding (1), we document that the bank lending channel, identified by Kashyap and Stein (2000) in the total and C&I lending of smaller banks from 1976 to 1993, was also present in other major lending categories. In addition, we find that the bank balance sheet channel of monetary policy operated not only through small banks, but also through US banks with the largest market shares in lending, which highlights the economic importance of this transmission mechanism. Regarding (2), the main contribution of our paper is that we shed light on how economic activity affects the strength of the bank credit channel of monetary policy. What would such an effect look like? As the value of bank assets is more pro-cyclical than the value of liabilities, bank net worth is pro-cyclical. As a result, the external finance premium that “riskier” banks face during a monetary policy tightening is counter-cyclical. Therefore, we expect the bank credit channel to be stronger during times of weak economic growth. Indeed, our analysis finds that the underlying transmission mechanism of monetary policy is stronger when output growth is low, suggesting that the bank credit channel of monetary policy is a more potent tool when economic conditions 1

are weak. Moreover, our results suggest that for some of the major lending categories, the bank credit channel is operative almost exclusively in a low-growth environment. Interestingly, we do not find evidence of the relevance of the bank credit channel and the effect of economic activity on its strength in the post-2007 period, suggesting an important post-financial crisis change in the transmission mechanism of monetary policy through bank lending in the United States. The paper proceeds as follows. In the remainder of this section, we develop our hypotheses and results in the context of the related literature. In Section 2 we describe our bank-level data and controls. We present our estimation methodology in Section 3, and we discuss the empirical results in Section 4. In Section 5 we show additional specifications and conduct robustness tests, and we conclude in Section 6. 1.1 Hypothesis development and literature review One of our main findings is that the bank credit channel worked actively through U.S. banks’ domestic lending from 1986 to 2008 – i.e., that a tightening in U.S. monetary policy reduced the lending of U.S. commercial banks, and monetary policy easing expanded such lending. The result derives from two simultaneous mechanisms. First, a tightening in monetary policy reduces the availability of funding to banks (Kashyap and Stein, 2000; Kishan and Opiela, 2000, 2006), causing (smaller) banks with less balance sheet liquidity to cut their lending – i.e., the bank lending channel of monetary policy (Kashyap and Stein, 1995). Second, the monetary effects on lending act through the net worth of banks, the bank balance sheet channel – i.e., because less capitalized banks (which investors judge to be riskier) find it increasingly expensive to replenish their liabilities in a higher interest rate environment, and thus reduce their lending. 2

The extensive literature on the bank credit channel dates back to Bernanke and Gertler (1995), who argued that changes in monetary policy impact the real economy in part by affecting banks’ lending decisions. Kashyap and Stein (2000) established empirically the existence of the bank lending channel in the U.S. – one of the mechanisms through which the credit channel operates: a tightening in monetary policy makes reservable liabilities less available to banks, and hence reduces banks’ lending. Kashyap and Stein (2000) also showed that this bank lending channel is the strongest amongst those banks with the most limited access to non-reservable liabilities such as wholesale funding: smaller banks, and among those, banks with less balance sheet liquidity (proxied with low securities to assets ratios). In our analysis, we extend the period of the Kashyap and Stein (2000) study up to the global financial crisis and we show that the bank lending channel they identified continued to hold in the lending of smaller U.S. banks.4 This finding is consistent with other studies in the bank lending channel literature. For instance, Altunbas et al (2009) showed that access to securitization is negatively related to the strength of the bank lending channel. Motivated by den Haan et al (2007), who find differences in the strength of the bank lending channel across target sectors, we study C&I, residential real estate (RRE) and consumer lending in addition to total lending flows. We find that the lending channel operates through smaller U.S. banks’ RRE and consumer lending, even more strongly than via C&I lending.5 4 Smaller U.S. banks are those below the 95th percentile of the cross-sectional distribution of asset size in each period. Cetorelli and Goldberg (2012) also extended the analysis through the 2000s and found evidence of the continued existence of the bank lending channel, using an alternative measure of balance sheet liquidity: the liquid assets ratio. 5 In recent years, several new papers have studied the bank credit channel in a non-US context. Some papers focused on the strength of domestic monetary transmission in a foreign country (Jimenez et al, 2012; Ioannidou et al, 2015) while others examined the strength of the credit channel in an international context (Correa et al, 2015; Correa et al, 2017; Temesvary et al, 2018, among others). 3

The second mechanism underlying the bank credit channel is the bank balance sheet channel: following a tightening in monetary policy, investors judge financially “weaker”, less capitalized (lower net worth), banks to be riskier (Bernanke, Gertler, and Gilchrist, 1999; Halvorsen and Jacobsen, 2016). As a result, such banks face a relatively larger increase in funding costs in funding markets (the external finance premium), which limits their lending. Temesvary et al (2018) show evidence of the bank balance sheet channel for total lending in the U.S. domestic context and in U.S. banks’ cross-border lending, as do Correa et al (2017).6 We document the presence of the bank balance sheet channel of monetary policy for total, C&I, residential and consumer lending of smaller U.S. banks from 1986 to 2008. We find strong evidence that the lending of smaller U.S. banks with lower equity capital ratios (net worth) is significantly more affected by changes in U.S. monetary policy than the lending of better capitalized U.S. banks (Kishan and Opiela, 2000). In addition, we extend the study of the bank credit channel for lending by all U.S. banks, including larger banks, in contrast to most of the literature on monetary policy transmission via bank lending, which focuses on smaller banks. Weighing each observation by the lending bank’s market share in each market segment, we find strong evidence that the bank balance sheet channel indeed affects the lending of all U.S. banks, even when we account for banks’ market shares. This finding is policy relevant, as it provides evidence of the bank credit channel in the lending of the economically most significant banks with the highest market shares. 6 A key challenge in identifying both mechanisms of the bank credit channel is to disentangle the effect of monetary policy on credit supply from its effects on credit demand (Bernanke and Blinder, 1992). Kashyap and Stein (2000) address this issue by including controls for macroeconomic developments, while comparing the impact of monetary policy on the lending of smaller banks with lower vs. higher securities ratios. Temesvary et al (2018) use extensive sets of fixed effects in their analysis of the bank credit channel in U.S. banks’ domestic and cross-border lending to fully control for time-varying shocks to the demand for bank credit – while comparing the impact of monetary policy changes on the lending of more vs. less funding-constrained U.S. banks as Kashyap and Stein (2000) –. 4

Together with studying the credit channel in U.S. banks’ domestic lending, our main research question focuses on: to what extent does the strength of this credit channel depend on aggregate, non-bank specific, factors – namely, economic activity? While earlier papers with micro data have investigated the role of various bank-specific variables, only a few studies have looked at the cyclicality in the transmission of monetary policy.7 Aikman et al (2017) study how the “credit cycle” affects the impact of monetary policy on real economic activity, and find that monetary policy is more potent when credit conditions are tighter. We investigate (to our knowledge, for the first time in the literature) how cyclicality in economic activity affects the relationship between monetary policy and credit conditions, i.e., the strength of the bank credit channel. The works of Bernanke et al (1999) and Matsuyama (2007) suggest that since bank net worth is pro-cyclical, the external finance premium that “riskier” banks face during a period of monetary policy tightening is counter-cyclical. Therefore, we expect the bank credit channel to be more potent during times of weak economic growth. Indeed, we show that the strength of the bank credit channel is significantly stronger both statistically and economically when the growth rate of real GDP is lower, and that the difference in the effectiveness of the channel is present even in the (market share-weighted) lending of all U.S. commercial banks; that is, we find the effect also for large banks. Finally, we do not find strong evidence of the bank credit channel for the period after 2007, which is consistent with previous work suggesting that there has been a regime shift in the domestic 7 Kashyap and Stein (2000) examine the role of bank size, and find that the lending channel is the strongest for smaller U.S. banks, which are more limited in their ability to obtain external funding. Cetorelli and Goldberg (2012) study whether “globally active” U.S. banks respond differently to changes in U.S. monetary policy. They find that “globalness” insulates U.S. banks from monetary policy, as they draw on the liquidity of foreign affiliates as a buffer. 5

transmission mechanism of US monetary policy through bank lending since the financial crisis (Civelli et al, 2018). 2. Data description 2.1. Lending flows and balance sheet controls We use bank-level data on U.S. commercial banks from a merger-adjusted version of the publicly available Call Reports (Reports of Conditions and Income). Our dataset contains detailed balance sheet information for all U.S. commercial banks, with quarterly frequency between 1986 and 2008. Our dependent variables, i.e., our measures of lending, are total, C&I, residential, and consumer lending flows. As is standard in the related literature, for each market segment we define lending flows as the quarterly change in the natural logarithm of each bank’s stock of loans, multiplied by 100. Our main explanatory variables of interest are measures of how sensitive is a bank to changes in monetary policy. First, we use the securities to assets ratio (in accordance with the definition in Kashyap and Stein, 2000) as a proxy of “funding constraint” in our study of the bank lending channel. Second, we employ the equity capital ratio (as in Temesvary et al, 2018) to proxy banks’ “funding constraint” when we study the bank balance sheet channel. We also control for the effects of changes in banks’ balance sheet conditions on credit supply that are unrelated to changes in monetary policy. We do so by using a set of balance sheet controls that include real net income, return on assets, level and growth of real total assets, and the deposits to assets ratio (which proxies for a bank’s access to short-term liquid liabilities). We provide detailed definitions of the dependent and control variables, as well as summary statistics, in Table 1. We also include bank and, when possible, time (year:quarter)-specific fixed effects. 6

These fixed effects capture unobservable shocks to the demand and supply of credit that are specific to individual banks but invariant over time (such as banks’ business model), and which vary over time but affect banks similarly (such as macroeconomic developments). 2.2. Monetary policy and real economic activity measures The main macroeconomic variables of interest are our measure of changes in U.S. monetary policy, and a proxy for real economic activity. For the former, in line with the related literature (Kashyap and Stein; Correa et al, 2015; Temesvary et al, 2018; among others) we employ quarterly changes (expressed as percentage points) in the effective federal (fed) funds rate. Positive changes correspond to monetary tightening, while negative values indicate monetary easing. In robustness checks, we use monetary policy forecast errors as exogenous proxies of changes in monetary policy. In order to proxy real economic activity (economic growth) in the United States, we use quarterly changes in the real U.S. gross domestic product as our main measure. In additional specifications, we also use the GDP (output) gap and forecast values of GDP growth as exogenous proxies. 2.3. Controls for changes in credit demand As discussed above, an important task we face is to control for macroeconomic changes that may affect banking clients’ demand for credit in the United States. It is important to control for these effects, which are otherwise unobservable, to ensure our identification of the relationship between changes in U.S. monetary policy and banks’ credit supply decisions. We include several Macro Controls to address this issue: the unemployment rate, a recession dummy, quarterly inflation and real aggregate investment. In addition, as Demand Controls we include weighted aggregated 7

controls for credit demand and standards from banks’ responses to the quarterly Senior Loan Officer Opinion Surveys (SLOOS). In some specifications, we also replace such macro-level controls with time fixed effects to capture those time-varying macro-level shocks that are common across all commercial banks in our sample. These fixed effects also control for non-monetary policy related factors such as regulations, which might affect banks’ lending (Frame et al, 2018). 3. Estimation methodology We employ a continuous diff-in-diff estimation method to identify the bank lending and bank balance sheet channels of monetary policy, which has become standard in the related literature (see Kashyap and Stein, 2000; Cetorelli and Goldberg, 2012; among others). We take important steps beyond the estimation methodology of these earlier papers, however, by (1) examining the bank credit channel’s interaction with economic activity, and (2) including fixed effects and extensive sets of controls (as described in the previous section) to eliminate non-monetary policy related confounding effects on lending flows. Our main specification describes U.S. banks’ quarterly lending flows as follows. Let 𝑛𝑛 𝑌𝑌𝑗𝑗,𝑡𝑡 denote bank j’s holdings of loans of type n (where n is one of total, C&I, CRE, RRE, and consumer) at time t. Then captures the quarterly change (from time t-1 to time t) of the natural 𝑛𝑛 𝛥𝛥ln(𝑌𝑌)𝑗𝑗,𝑡𝑡 logarithm of bank j’s stock of loans of type n. Our specification is as follows: (1) 𝑛𝑛 𝛥𝛥ln(𝑌𝑌)𝑗𝑗,𝑡𝑡 = 8

4 𝑛𝑛 4 𝑢𝑢𝑢𝑢 4 𝑢𝑢𝑢𝑢 4 𝛼𝛼+∑𝑘𝑘=1𝜂𝜂𝛥𝛥ln(𝑌𝑌)𝑗𝑗,𝑡𝑡−𝑘𝑘 +∑𝑘𝑘=1𝛽𝛽𝑘𝑘𝑀𝑀𝑀𝑀𝑡𝑡−𝑘𝑘 +∑𝑘𝑘=1𝛾𝛾𝑘𝑘𝑀𝑀𝑀𝑀𝑡𝑡−𝑘𝑘 ×𝐶𝐶𝑗𝑗,𝑡𝑡−𝑘𝑘 +∑𝑘𝑘=1𝛿𝛿𝑘𝑘𝐶𝐶𝑗𝑗,𝑡𝑡−𝑘𝑘 4 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 4 𝑢𝑢𝑢𝑢 4 𝑢𝑢𝑢𝑢 +∑𝑘𝑘=1𝜁𝜁� � +(∑𝑘𝑘=1𝜃𝜃𝑘𝑘𝑀𝑀𝑀𝑀𝑡𝑡−𝑘𝑘 +∑𝑘𝑘=1𝜆𝜆𝑘𝑘𝑀𝑀𝑀𝑀𝑡𝑡−𝑘𝑘 ×𝐶𝐶𝑗𝑗,𝑡𝑡−𝑘𝑘 𝐶𝐶𝐶𝐶𝐵𝐵𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑗𝑗,𝑡𝑡−𝑘𝑘 𝑛𝑛 4 𝛥𝛥𝐷𝐷𝐷𝐷𝐵𝐵𝐵𝐵𝐷𝐷 𝑛𝑛 +∑𝑘𝑘=1𝜋𝜋𝑘𝑘𝐶𝐶𝑗𝑗,𝑡𝑡−𝑘𝑘) ×𝛥𝛥𝛥𝛥𝛥𝛥𝑀𝑀𝑡𝑡 +𝜂𝜂� � +𝜀𝜀𝑗𝑗,𝑡𝑡 𝐶𝐶𝐶𝐶𝐵𝐵𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑡𝑡−1 In Equation (1), denotes the quarter-to-quarter lending flow at the bank level as 𝑛𝑛 𝛥𝛥ln(𝑌𝑌)𝑗𝑗,𝑡𝑡 described above. The monetary policy variable MP is the quarterly change in the fed funds rate from time t-1 to t. Furthermore, C denotes the bank’s securities to assets ratio, later replaced by the equity capital to assets ratio.8 As in Kashyap and Stein (2000) and Cetorelli and Goldberg (2012), we focus on the cumulative impact of monetary policy changes over the four preceding quarters.9 Therefore, we include four lags of the monetary policy measure, the funding constraint proxy, and their interactions.10 Furthermore, is the quarterly change (from t-1 to t) in the U.S. real gross domestic product, as describ𝛥𝛥ed𝛥𝛥 𝛥𝛥ab𝑀𝑀o 𝑡𝑡 ve. The sets of Bank Controls and Demand Controls contains the variables described in the previous section. In addition, these vectors also include combinations of bank and time fixed effects to control for unobservable time-invariant bank-specific, and bank-invariant time-varying shocks, respectively, which impact lending flows. Our strategy for identifying the bank credit channel of U.S. monetary policy focuses on the sign of the cumulative coefficients on the interaction term of the bank’s funding ratio and the U.S. 8 A potential concern is the feedback effect that may exist from domestic economic conditions to monetary policy changes. Given our identification strategy we are less concerned about domestic macroeconomic feedback effects into monetary policy (Acharya et al, 2016). Indeed, we identify monetary transmission from the differential response of funding-constrained vs. funding-abundant banks. Even if macro shocks simultaneously affect all banks’ flows through monetary policy, the cross-bank differences in the strength of transmission should not be impacted. 9 Our use of lagged values of the bank funding ratios ensures that these ratios may at most reflect past strategic choices of banks. The inclusion of four lags of the quarterly interest rate changes (as is standard in the related literature) ensures that we capture the cumulative effect of monetary policy changes throughout the previous year. 10 Since a bank can choose its funding ratio strategically, the inclusion of lags of the funding ratio that are in time similar to those of the monetary policy changes (in their levels and interactions) reduces concerns that the funding ratio may be endogenous to the interest rate changes. 9

monetary policy change: .11 If U.S. banks that are flush with securities or equity capital 4 reduce their lending flow∑s 𝑘𝑘 le = s 1 s𝛾𝛾 𝑘𝑘 in response to a tightening in U.S. monetary policy than banks which have lower securities or equity capital ratios, we expect to find . Lastly, 4 expectations as to the sign of are mixed. In the international contex∑t, 𝑘𝑘 T = e 1 m𝛾𝛾𝑘𝑘 es>va0ry (2014) 4 finds that funding-constraine∑d 𝑘𝑘 b = a 1 n𝛿𝛿k 𝑘𝑘 s maintain higher (foreign) lending flows, suggesting a negative sum of coefficients. However, the results of Basset and Berrospide (2017) in the domestic context suggests a positive relationship between bank capitalization and lending for the CCAR banks over the past five years. The key focus of our paper is on the role of real economic activity on the strength of the bank credit channel. The bank credit channel of monetary policy rests on frictions in banks’ access to external funding, either due to inelastic supply of such funding (as in the bank lending channel) or agency costs (as in the bank balance sheet channel). Since bank net worth is pro-cyclical and thus the external finance premium is counter-cyclical (Bernanke et al 1999; Matsuyama, 2007), we expect both these frictions to be more pronounced when economic (and thus, market funding) conditions are tight, due to subdued economic activity. Therefore, we expect the bank credit channel to be stronger in times of low economic growth: and . In other 4 4 words, we expect that higher GDP growth attenuate the ∑st 𝑘𝑘 re = n 1 g𝜃𝜃t 𝑘𝑘 h >of0 the ba∑n 𝑘𝑘 k = 1 c𝜆𝜆re 𝑘𝑘 d<it c0hannel of monetary policy by reducing the market frictions that the bank credit channel rests on. 4. Estimation results 11 In Equation (1), the operation of the bank credit channel would suggest the direct effect of a tightening in U.S. monetary policy on bank flows to be negative: , but such a result is not needed for our identification scheme. In fact, for C&I lending, den Haan et al (42007) and Civelli et al (2018) document an increase in business loans in response to monetary tightening. ∑𝑘𝑘=1𝛽𝛽𝑘𝑘 <0 10

Between the two prongs of the bank credit channel, we first examine the bank lending channel and its interaction with economic activity, as laid out in the Introduction. In Table 2, we consider smaller US banks’ all-encompassing, total core lending.12 Columns 1 to 4 examine the role of banks’ securities holdings in mitigating the impact of monetary policy, as in Kasyhap and Stein (2000). In these estimations, we interact the bank lending channel strength measures with the U.S. real GDP growth rate. We find strong evidence of the existence of the bank lending channel: the cumulative interaction of the securities to assets ratio (SEC) with quarterly changes in the fed funds rate is consistently statistically significant at the 1 percent level. Furthermore, the results also strongly confirm that this monetary transmission effect is particularly potent when economic growth is low: the coefficients on the double interaction terms are consistently negative and highly statistically significant. These effects are robust to the inclusion of extensive sets of macroeconomic and credit demand controls (Columns 2 and 3) as well as to the most stringent specification with time fixed effects (Column 4). The calculations at the bottom of the table illustrate the role of economic activity. For instance in Column 1, a marginal increase in a bank’s SEC ratio mitigates the impact of a change in monetary policy on total lending by 0.05 percentage points (pp) when economic growth is low (at the 10th percentile of the distribution of real GDP growth over time, which corresponds to a quarterly growth rate of 0.6 percent). However, this impact is infinitesimal when growth is high (at the 90th percentile, or 1.85 percent per quarter). These results are also economically significant. For instance, for the specification in Column 1, our calculations (not shown in the table) reveal 12 Smaller U.S. banks are those below the 95th percentile of the cross-sectional distribution of asset size per period. 11

that in times of low economic growth, a funding constrained bank (at the 10th percentile of the SEC distribution, which in our sample is zero) would reduce its total lending flows by 1.94 pp more in response to a 100 bps tightening in U.S. monetary policy than would a funding abundant bank (at the 90th percentile, which corresponds to a near 38 percent SEC ratio). However, the relative magnitudes are much smaller, at 0.17 pp, during times of high economic growth. These findings, which are qualitatively consistent across specifications, suggest that in total lending the bank lending channel operates strongly when GDP growth is low, but does not operate when GDP growth is high. Note that in periods of low economic growth, we tend to think of monetary policy as typically easing, not tightening. Thus, the interpretation of our results through this lens is that monetary policy is likely to be more effective through the bank credit channel when it is trying to spur growth during periods of low growth, rather than in trying to cool down the economy in periods of high growth. We study the strength of the other mechanism of the bank credit channel, the bank balance sheet channel of monetary transmission, in Columns 5 to 8 of Table 2. To do so, we consider the interaction of changes in monetary policy with each bank’s equity capital to assets ratio (CAR). Therefore, this measure of funding constraint now replaces the SEC ratio, the funding constraint measure we used in Columns 1 to 4. These specifications, which examine the bank balance sheet channel of monetary policy in smaller banks’ lending over the 1986-2008 period, are related to the work of Kishan and Opiela (2000) – which focused on the 1980-1995 era. The results in Columns 5 through 8 confirm the existence of the bank balance sheet channel in U.S. banks’ total lending flows – especially in low growth periods. The coefficients on the (simple) the cumulative interaction of the fed funds rate changes and CAR are consistently positive and significant at the 1 percent level. In addition, the double interaction of this term with the GDP 12

growth rate is negative and statistically significant. Calculations based on the coefficients in Column 6 reveal that a 100 bps tightening in the fed funds rate would lower the total lending flows of a bank with median CAR by 0.47 pp during a period of low growth, but the comparable negative effect in a high growth environment would be a lower reduction by 0.34 pp. As before, the simple interaction term is significantly higher (at 0.09) when evaluated during a period of low growth, than in a high growth period (near zero). The economic effects are significant as well. A 100 bps increase in the fed funds rate would lower the total lending of a low capitalized bank (at the 10th percentile of the CAR distribution, which corresponds to 7.38 percent) by 0.51 pp more than the lending of a high capitalized bank (at the 90th percentile of CAR, which is 13.27 percent), in a low-growth environment. In contrast, when economic growth is high, this differential impact would be substantially smaller, at 0.03 pp. The results discussed so far have focused on smaller US banks, weighing each bank equally – that is, without taking into account each bank’s credit market share. However, in the highly concentrated US banking markets, the behavior of the largest banks can provide important insight about aggregate lending patterns. Thus, next we turn to studying the presence of the bank lending channel and the bank balance sheet channel in the weighted total lending flows of all U.S. commercial banks. The weighted estimations include all U.S. banks (the largest banks as well as the smaller banks we have studied thus far) and weigh each bank’s lending flows by that bank’s market share in the given market segment. As expected based on the hypotheses laid out in the Introduction, we find no evidence of the presence of the bank lending channel in the weighted results (not shown).13 In contrast, as 13 That is, the assertion from Kashyap and Stein (2000) that the bank lending channel operates only on the lending of smaller banks that have more limited access to outside funding markets. 13

shown in Columns 9 through 12 of Table 2, we document the presence of the bank balance sheet channel in the weighted total lending flows regressions. The coefficients on the relevant simple interaction term show that in response to a 100 bps increase in the fed funds rate over four quarters, a 1 pp higher capital ratio would correspond to a 0.73 to 0.78 pp weaker lending decline. The double interaction terms reveal that even a 1 pp increase in the GDP growth rate would attenuate this differential impact by 0.79 to 0.88 pp. These relative magnitudes imply that in the weighted estimations, the bank balance sheet channel of monetary policy transmission is present only when economic growth is very low – notably, below the 25th percentile of the distribution of economic growth rates over our sample period. For instance, in terms of economic significance, the monetary transmission-mitigating impact of a higher CAR is 0.24 to 0.28 pp when economic growth is very low (at the 10th percentile, bottom rows of Table 2). Our strong evidence of the bank credit channel for banks’ total lending in Table 2 leads us to hone in on each of the economically more relevant core lending categories. We first evaluate the presence of the bank lending channel both in business financing, where we consider C&I and commercial real estate (CRE) lending, and in household financing, where we consider residential real estate (RRE) and consumer lending. Then, we perform a similar set of econometric exercises by lending category to assess the presence of the bank balance sheet channel. Table 3 shows the results for C&I lending flows, estimated from the same specifications that we used for total lending. Columns 1 through 4 evaluate the bank lending channel, and we see that most of the relevant interaction coefficient estimates are larger during periods of low economic activity. For instance, in Column 1 the interaction of cumulative monetary policy changes with the SEC ratio (the simple interaction term) is 0.06 in a low growth environment, but is only 0.03 when economic growth is high. Furthermore, calculations also reveal that in response to a 100 bps 14

tightening in the fed funds rate (over four quarters), the lending of a funding-constrained bank would respond by 2.32 pp more than the lending of a funding-abundant bank – when economic growth is low. In a high-growth environment, this differential is only 1.18 pp. The results of the fourth column, which is our most complete fixed effects specification, suggest that the bank lending channel was operative during our sample period. However, this channel was not significantly stronger in periods of weak output growth, which highlights the importance of other lending categories in driving the differential effectiveness of monetary policy documented for total lending. Columns 5 to 8 provide the results for CRE lending. However, the lack of significance and change in signs of the coefficients of interest across specifications indicate that the bank lending channel did not operate through CRE lending during our sample period. The lack of significance for CRE lending, the weak result obtained for the fixed effects specification in C&I lending, and the sole focus of earlier work (Kashyap and Stein, 2000, Cetorelli and Goldberg, 2012) on total and C&I lending flows, suggest that our disaggregated analysis (evaluating all the major loan type categories separately) can be quite helpful to better explain how the bank lending channel of monetary policy operates in the United States. Therefore, in Table 4 we run the same specifications as in Table 3, now for lending to households: residential and consumer lending flows. Importantly, we find evidence of the bank lending channel of monetary policy both in residential (Columns 1 through 4) and consumer lending (Columns 5 through 8). Looking at residential flows first, the simple and double interaction terms are consistently positive and negative, respectively, and significant at the 1 percent level throughout. For instance, in Column 1, the value of the simple interaction term (the mitigating effect of a higher SEC ratio on the direct impact of monetary policy) is 0.06 when evaluated at low economic growth, but is much smaller near zero, when evaluated in a high growth environment (as show in the calculations 15

at the bottom of the table). Furthermore, this same column also reveals that a 100 bps tightening in the fed funds rate over four quarters would reduce the residential lending of a funding constrained bank by 1.47 pp more than the lending of a more funding abundant bank, in a low economic growth environment. However, this differential effect is substantially lower, at 0.06 pp, in a high growth environment. Turning to consumer lending in Columns 5 through 8, the magnitudes of the coefficients of interest are somewhat smaller, but the simple and double interaction coefficients are consistently and significantly positive and negative, respectively, as before. As shown for example in Column 5, the simple interaction term is 0.04 in a low growth environment, but infinitesimal in a highgrowth environment. Accordingly, a 100 bps tightening in U.S. monetary policy over four quarters would reduce the consumer lending of a funding-constrained U.S. bank by 1.59 pp more than the consumer lending flows of a funding-abundant bank, in a low-growth environment. This differential effect is smaller, at near 0.41 pp, in a high-growth environment. The results (also in Tables 2 and 3) reveal that banks with higher SEC ratios lend substantially more, all else equal, than their more funding-constrained counterparts. For each major lending category, we now turn to analyze the strength of the second mechanism of the bank credit channel, the bank balance sheet channel of monetary transmission. In Table 5 we repeat the specifications in Table 3, now interacting changes in monetary policy with each bank’s capital ratio. As expected, Columns 1 through 4 show that C&I lending by low capitalized banks responds more strongly to changes in U.S. monetary policy, and the differential is larger when economic growth is low. For instance, Column 2 reveals that a 100 bps tightening in the fed funds rate would lower the C&I lending growth of a low capitalized bank 0.63 pp more than that of a high capitalized bank when economic growth is low – but by much less (0.03 pp) 16

when growth is high. Importantly, the specifications for C&I lending throughout Table 5 consistently show that well-capitalized banks maintain significantly higher lending growth than low capitalized ones.14 In contrast, as shown in Columns 5 to 8, there is neither evidence of higher CRE lending growth for higher capitalized banks, nor evidence supporting the presence of a bank balance sheet channel operating through this lending category during the period. In Table 6 we repeat the Table 5 specifications, studying the bank balance sheet channel of monetary policy by using the equity capital to assets ratio as a measure of a bank’s funding constraint. However, instead of focusing on C&I and CRE, we run the analysis for the major household financing categories, i.e., RRE and consumer lending. We find some evidence that lower capitalized banks’ residential lending flows respond stronger to changes in U.S. monetary policy, especially in lower growth periods. The coefficients on the simple and double interaction terms are always positive and negative, respectively, and generally statistically significant, including in the most stringent specification that includes bank and time fixed effects. In Column 2, for instance, the differential response of a low capitalized bank to a 100 bps tightening in the fed funds rate is 0.36 pp greater than that of a low capitalized bank when growth is low, but this differential response is near zero when economic growth is high. There is no evidence of the bank balance sheet channel in consumer lending (Columns 5 through 8 of Table 6). However, the result that better capitalized banks lend significantly more, prevails throughout all specifications. In Tables 7 and 8, we use the weighted estimations to analyze the presence of the bank balance sheet channel in each of the major lending categories. To this end, as we did for total lending, we include all U.S. banks and weigh each bank’s lending flows by that bank’s market 14 Basset and Berrospide (2017) find similar effects for the CCAR banks for the 2012-2016 period. 17

share in the given market segment. In Table 7 we study the bank balance sheet channel for business financing, looking at the C&I and CRE lending flows of all U.S. commercial banks. We find significant and large coefficient estimates in our analysis of weighted C&I lending flows (Columns 1 through 4). The simple interaction coefficient estimates range from 1.10 to 1.25 in magnitude, and the double interaction coefficients are in the 1.08 to 1.13 range – all significant at the 1 percent level. The relative magnitudes of these coefficients suggest that, as in the case of total lending flows, the bank balance sheet channel of monetary policy transmission is present in C&I lending flows only in times of low economic growth. Additionally, the strong positive relationship between capitalization and lending growth that we highlighted when discussing non-weighted results, no longer holds in our weighted estimations for either C&I or CRE lending. The CRE lending results, shown in Columns 5 to 8, also suggest that the bank balance sheet channel was not present during our sample period. In Table 8, we examine the strength of the bank balance sheet channel in all U.S. banks’ weighted residential and consumer lending flows. There is no evidence of the existence of the bank balance sheet channel of monetary policy transmission in residential lending flows (Columns 1 through 4). However, we do find strong and consistent evidence in consumer lending flows (Columns 5 through 8) – where the coefficients on the interaction terms are large and significant at the 1 percent level throughout. The economic significance of these results is notable: for instance, calculations based on the coefficient estimates in Column 7 imply that for a bank with the median capital to assets ratio, a 100 bps increase in the fed funds rate would reduce consumer lending flows by 1.14 pp, at the median economic growth rate. The negative and significant coefficients on the double interaction terms suggest that these negative lending effects are significantly larger in periods of low economic growth. Indeed, a 100bps increase in the fed funds 18

rate lowers the lending growth of a low capitalized bank substantially more than that of a high capitalized bank when economic growth is low – but this differential reverses sign when growth is high, suggesting the lack of a bank balance sheet channel. 4.2. Alternative specifications To alleviate concerns about potential endogeneity of GDP growth to monetary policy actions driving our results, we also repeat our estimations instrumenting actual concurrent GDP growth with one-quarter lagged forecast values. This instrumental variable is yesterday’s prediction of today’s GDP growth rate, and thus it is not correlated with today’s macro indicators. Table 9 repeats the most stringent specification, with fixed effects, for smaller US banks and each major loan category, using the IV formulation.15 Our IV results confirm the presence of the bank lending and bank balance sheet channels for total lending and all the core lending categories except CRE. The findings are consistent in magnitude and statistical significance with the main outcomes we described earlier, suggesting that our benchmark findings are robust to correcting for potential endogeneity bias. The last five columns of Table 9 repeats the IV regressions testing the bank balance sheet channel, now including all banks and weighing each observation by the bank’s market share in each lending segment. The relevant coefficients have the expected signs but are not statistically significant, except for RRE lending. The bank balance sheet channel has only been present in this lending category, and has operated largely during periods of low output growth. 15 The first column in Table 9 corresponds to the Column 4 specification of Table 2; Column 2 is the Column 4 specification of Table 3, and so forth. In this way, for each loan type we repeat the most complete specification, which uses bank and time fixed effects. 19

We also extend our analysis to the crisis/post-crisis period of 2008-2016. Earlier papers established the substantial reduction in banks’ loan supply following the large crisis-induced reduction in the availability of liquid funding to banks (Altunbas et al, 2011; Cetorelli and Goldberg, 2012). Importantly, in this exercise our goal is to examine the crisis effect on the relationship between monetary policy and bank lending, rather than studying the direct lending effect of the crisis which these earlier papers focused on. In Table 10, we repeat the most complete specifications for each loan type from Tables 2 through 8 for the 2008-2016 period using Krippner (2016)’s shadow short-term interest rate as our measure of the stance of the post-2007 “nonconventional” monetary policy actions (that is, the period during which the effective lower bound was binding). During this period, we do not find consistently significant evidence of the existence of either the bank lending or the bank balance sheet channel of monetary policy transmission. Banks’ lending incentives may be different during this period due to the large quantities of reserves they hold. Still, it is an interesting finding in the domestic context, seeing as Morais et al (2017) and Temesvary et al (2018) found some evidence of the bank credit channel in U.S. banks’ lending in foreign countries. 5. Robustness 5.1. Using monetary policy forecast errors In Table 1A, we repeat our benchmark specifications, now using the Romer and Romer monetary policy forecast residuals as our monetary policy measures (as updated by Wieland and Yang, 2015) in place of the quarterly change in the fed funds rate. In these estimations (which are select specifications from Tables 2 through 8) we continue to find evidence of the bank lending channel in smaller banks’ lending: the simple and double interactions in specifications using the SEC ratio 20

are significantly positive and negative, respectively. However, we do not find consistent evidence of the balance sheet channel using the Romer residuals as our measures of monetary policy (Columns 6 through 15). 5.2. Bank credit channel in an underperforming economy – When GDP growth is below trend In Table 2A, we examine the bank credit channel when the US economy is operating “below trend” – that is, when the output gap (as defined by the Congressional Budget Office, or CBO) is below zero.16 Regarding the bank lending channel (Columns 1 through 5), results suggests that this channel operates stronger when the economy is under-performing. This finding is in line with the literature which shows that the funding cost-alleviating role of banks’ net worth is particularly expressed when economic activity is low (Bernanke, Gertler and Gilchrist, 1999). Regarding the bank balance sheet channel, this channel also operates stronger through smaller banks’ lending when economic activity is weak (Columns 6 through 10) – however, in weighted estimations such ranking is not consistently observable (Columns 11 through 15). This finding is in line with the implications from our benchmark results: the bank balance sheet channel is more cyclical in the lending of smaller US banks. 5.3. Asymmetries: Bank credit channel during monetary policy tightening vs. easing Another interesting question to examine is the extent to which there may be asymmetries in the operation of the bank credit channel, depending on whether the Federal Reserve is tightening or easing monetary policy. Kishan and Opiela (2000) has established evidence of such asymmetries 16 The CBO calculates GDP growth trend, and therefore GDP gaps, over the 1949-2017 period. As result of using a different “reference period”, there are notably higher number of “below trend” observations during our 1986-2008 period than “above trend”. However, we believe that using the CBO’s longer reference period is more representative than calculating the GDP growth trend over our sample. 21

during the first part of our sample, for both the bank lending and bank balance sheet channels. In addition to expanding the study of such asymmetries through 2008, more importantly we also examine how asymmetries may be related to real economic conditions. In Table 3A, we repeat the most complete specifications (including bank and time fixed effects) from Tables 2 through 4. We now separately examine our results during quarters of monetary policy easing (that is, when the cumulative change in the fed funds rate over the preceding four quarters is negative, in Columns 1 through 5) and monetary policy tightening (when the cumulative change in the fed funds rate is positive, in Columns 6 through 10). Similarly, in Tables 4A and 5A we repeat the most complete specifications from Tables 2, 5 and 6 (the bank balance sheet channel in smaller banks’ lending) and Tables 2, 7 and 8 (the bank balance sheet channel in weighted estimations), respectively. These results show that the cyclicality of the bank lending channel (Table 3A, second row) is somewhat more detectable during periods of monetary policy easing. Similarly, Table 4A shows that the bank balance sheet channel is substantially stronger during periods of monetary policy easing (Columns 1 through 5). Such ranking is less clear when looking at the bank balance sheet channel in weighted estimations (Table 5A). The general policy implication of these asymmetries is that during periods of low economic growth, expansionary monetary policy’s operation through the bank credit channel through smaller banks’ lending is especially potent. 5.4. Bank credit channel and the business cycle In additional estimations (available from the authors upon request), we also examine the strength of the bank credit channel at various points along the business cycle. We do so by replacing real GDP growth with the CBO output gap in Equation (1) – thus, now studying both “above trend” 22

and “below trend” periods. These results confirm our earlier evidence that the bank credit channel is counter-cyclical – however, there are notable differences across sectors in this cyclicality. 5.5. Bank credit channel and the business cycle – Using monetary policy forecast errors We also re-examine the relationship between the bank credit channel and the business cycle, now using monetary policy forecast errors as our monetary policy measure. Specifically, we repeat our benchmark specifications using the CBO’s output gap as measures of the business cycle, and we also replace the quarterly change in the fed funds rate with the Romer and Romer monetary policy forecast residuals. Doing so, we strongly confirm our benchmark results that (1) the bank credit channel operates strongly, and (2) the bank credit channel is counter-cyclical (that is, particularly strong when economic activity is weak). 5.6. Bank balance sheet channel of monetary policy in “smaller” banks’ lending In our analysis, we focused on the subset of smaller banks to examine the bank lending channel and bank balance sheet channels of monetary policy. We then proceeded to conduct weighted estimations to study the bank balance sheet channel in all banks’ lending. In our benchmark hypotheses and results we did not focus on the bank lending channel in all bank’s lending, because the work of Kashyap and Stein (2000) suggests such channel operates only in the lending of the (more funding access-constrained) smaller banks. However, for completeness in additional specifications we also carry out these estimations, and as expected, do not find evidence of the bank lending channel in weighted estimations on all banks’ lending. 6. Conclusions 23

We study the bank credit channel of monetary transmission in U.S. commercial banks’ lending flows. We find that monetary policy is more effective through the bank credit channel when it is trying to spur growth during periods of low growth, rather than in trying to moderate economic activity in periods of high growth. Our results are fourfold: first, we provide evidence of the bank lending channel in various types of lending by smaller U.S. banks from 1986 to 2008, and we show that this channel is significantly stronger in times of low economic growth. Second, we show that the bank balance sheet channel is also present in these banks’ lending flows, and it is also stronger in a low growth environment. Third, when we consider all U.S. commercial banks’ weighted lending flows, we find strong evidence of the bank balance sheet channel. Fourth, we do not detect a bank credit channel post-2007. In understanding the behavior of the macro economy, it is critical to understand the channels through which monetary policy operates. Our estimates, based on historical analysis, may provide insight into the strength of the bank credit channel in the coming years. 24

References Aikman, David, Andreas Lehnert, Nellie Liang and Michele Modugno (2017). “Credit, financial conditions, and monetary policy transmission,” Brookings Institution Working Paper 39. Acharya, V.V., T. Eisert, C. Eufinger and C.W. Hirsch (2016). Whatever it Takes: The Real Effects of Unconventional Monetary Policy. New York University, New York NY. Altunbas, Yener, Leonardo Gambacorta and David Marques-Ibanez (2009). “Securitization and the bank lending channel,” European Economic Review 53, pp. 996-1029. ___ (2011). “The bank lending channel: Lessons from the crisis,” Economic Policy 26(66), pp. 135-182. Basset, William and Jose Berrospide (2017). “The Impact of Stress Tests on Bank Lending,” Working paper. Bernanke, B.S. and A.S. Blinder (1992). “The Federal Funds Rate and the Channels of Monetary Transmission, American Economic Review 82, pp. 901 921. Bernanke, B.S. and M. Gertler (1995). “Inside the Black Box: The Credit Channel of Monetary ‐ Policy Transmission,” Journal of Economic Perspectives 9, pp. 27 48. Bernanke, B. S., Mark Gertler and Simon Gilchrist (1996). “The Financial Accelerator and the ‐ Flight to Quality,” Review of Economics and Statistics 78(1), pp. 1–15. Cetorelli, Nicola and Linda S. Goldberg (2012a). “Banking Globalization and Monetary Transmission,” Journal of Finance 67, pp. 1811 1843. Civelli, A., S. Barraza and N. Zaniboni (forthcoming). “Business loans and the transmission of ‐ monetary policy,” Journal of Financial & Quantitative Analysis. Correa, Ricardo, Linda S. Goldberg and Tara Rice (2015). “International Banking and Liquidity Risk Transmission: Evidence from the United States,” IMF Economic Review 63, pp. 626– 643. den Haan, Wouter J., Steven W. Sumner and Guy M. Yamashiro (2007). “Bank loan portfolios and the monetary transmission mechanism,” Journal of Monetary Economics 54(3), pp. 904- 924. Frame, W. Scott, Atanas Mihov and Leandro Sanz (forthcoming). “Foreign Investment, Regulatory Arbitrage, and the Risk of U.S. Banking Organizations,” Journal of Financial & Quantitative analysis. Halvorsen, Jorn and Dag Henning Jacobsen (2016). “The bank lending channel empirically revisited,” Journal of Financial Stability 27, 95-105. Ioannidou, V.P., Steven Ongena and Jose-Luis Peydró (2015). “Monetary Policy, Risk Taking and Pricing: Evidence from a Quasi Natural Experiment,” Review of Finance 19, pp. 95 144. ‐ Jiménez, G., Steven Ongena, S., Jose-Luis Peydró and J. Saurina (2012). “Credit Supply and ‐ ‐ Monetary Policy: Identifying the Bank Balance Sheet Channel with Loan Applications,” American Economic Review 102, pp. 2301 2326. ‐ ‐ 25

Kashyap, Anil K. and Jeremy C. Stein, Jeremy (1995). “The impact of monetary policy on bank balance sheets,” Carnegie-Rochester Conference Series on Public Policy 42, pp. 151-195. ___ (2000). “What Do A Million Observations on Banks Say About the Transmission of Monetary Policy?” American Economic Review 90, 407 428. Kishan, Ruby P. and Timothy P. Opiela (2000). “Bank size, bank capital and the bank lending ‐ channel,” Journal of Money, Credit and Banking 32(1), 121-141. ___ (2006). “Bank capital and loan asymmetry in the transmission of monetary policy,” Journal of Banking & Finance 30, 259-285. Krippner, Leo (2016). “Documentation for United States Measures of Monetary Policy,” Reserve Bank of New Zealand. Morais, Bernardo, Jose-Luis Peydró and Claudia Ruiz (2017). “The International Bank Lending Channel of Monetary Policy Rates and QE: Credit Supply, Reach for Yield, and Real Effects,” International Finance Discussion Papers 1137. ‐ ‐ Temesvary, Judit (2014). “The Determinants of U.S. Banks’ International Activities,” Journal of Banking and Finance 44, pp. 233 247. Temesvary, Judit, Steven Ongena and Ann L. Owen (2018). “A Global Lending Channel ‐ Unplugged? Does U.S. Monetary Policy Affect Cross border and Affiliate Lending by Global U.S. Banks?” Journal of International Economics 112, pp. 50-69. ‐ Wieland, J.F. and M.-J. Yang (2017). “Financial Dampening,” Working paper. 26

Table 1: Summary statistics and definitions of model variables [1] [2] [3] [4] [5] [6] [7] [8] Variables Units Definition mean sd min p25 p50 p75 max N Lending flows Quarterlychangeinthenaturallogarithmof Total % each bank'sstockoftotal loans,multiplied by 100 2.088 5.637 -49.75 -0.73 1.845 4.621 49.96 523,517 Commercial & Industrial (C&I) % Defined as above, for C&I loans 1.708 12.47 -49.99 -4.638 1.388 8.008 50 512,551 Residential % Defined as above, for residential loans 1.869 8.312 -50 -1.841 1.372 5.066 49.98 519,370 Other categories % Defined as above, for "other" loans 0.888 9.789 -49.98 -3.629 0.74 5.271 49.98 517,041 Measures of funding constraint Totalequitycapitaldividedbytotalassets, Leverage ratio % multiplied by 100 9.835 2.344 6.366 8.092 9.244 11.02 18.21 519,323 Total securities holdings divided by total Securities to assets ratio % assets, multiplied by 100 12.51 16.21 0 0 4.39 21.36 65.47 523,516 Bank characteristics Natural logarithm of each bank's net Log of real net income Integers income, deflated using the PCE of 2016 Q3 5.872 1.673 -1.184 5.123 6.04 6.918 11.57 523,517 Total income divided by total assets, Return on assets % multiplied by 100 0.686 0.539 -3.434 0.336 0.629 0.992 2.58 523,517 Natural logarithm of each bank's total Log of real total assets Integers assets, deflated using the PCE of 2016 Q3 11.11 1.009 7.275 10.4 11.08 11.78 14.12 523,517 Quarterlychangeinthenaturallogarithmof Growth of real total assets % each bank's real total assets, times 100 1.775 4.713 -11.99 -0.808 1.352 3.789 30.78 523,517 Total deposits divided by total assets, Deposits to assets ratio % multiplied by 100 0.857 0.063 0.0391 0.834 0.872 0.897 0.947 523,517 Macro controls Number of unemployed divided by the Unemployment rate % labor force 5.537 1.011 3.9 4.7 5.5 6.2 7.6 523,517 Indicator variable which takes a value of 1 in quarters where the U.S. economy is in a Recession dummy 0/1 recessionascharacterizedbytheNBER,and 0 otherwise 0.0939 0.292 0 0 0 0 1 523,517 Quarterly change in the effective federal Quarterly change in the fed funds rate % funds rate -0.0951 0.503 -1.424 -0.242 -0.013 0.226 0.701 523,517 Quarterlychangeinthenaturallogarithmof Quarterly real GDP growth % total U.S. GDP, deflated using the PCE of 2016 Q3 1.247 0.587 -1.994 1.017 1.232 1.612 2.436 523,517 Quarterlychangeinthenaturallogarithmof Quarterly inflation % the U.S. PCE 0.425 0.297 -1.451 0.332 0.431 0.532 1.037 523,517 Natural logarithm oftotal U.S.investment, Log of real investment Integers deflated by the PCE of 2016 Q3 7.309 0.328 6.726 7.039 7.389 7.533 7.797 523,517 Demand and risk controls Excessbondpremium(monthlyaveragedto Excess bond premium % quarterly) -0.00618 0.56 -0.832 -0.408 -0.092 0.159 2.528 523,517 Macro weighted SLOOS variable - demand SLOOS demand index Integers across all loan categories -0.0168 0.201 -0.568 -0.139 -0.0371 0.132 0.389 497,328 MacroweightedSLOOSvariable-standards SLOOS standards index Integers across all loan categories 0.0806 0.219 -0.196 -0.0814 0.0384 0.178 0.873 514,728

Table 2: The impact of changes in U.S. monetary policy on the total lending of U.S. banks with different securities to assets or leverage ratios: 1986-2008 [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] Lending flows: Total Total Total Total Total Total Total Total Total Total Total Total Securities to Securities to Securities to Securities to Ratio Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Assets Assets Assets Assets Variables Quarterly Change in the Fed 0.108 0.324 0.912 -0.798 -0.878 -1.838 -5.385 -2.99 -2.991 Funds Rate {t-1 to t-4} [0.0745] [0.0948]*** [0.115]*** [0.269]*** [0.28]*** [0.29]*** [3.046]* [2.937] [3.261] -0.070 -0.473 -0.668 0.525 0.147 1.174 6.246 3.872 3.798 Quarterly Change in the Fed [0.0546] [0.0641]*** [0.0707]*** [0.19]*** [0.195] [0.201]*** [2.502]** [2.397] [2.787] Funds Rate * GDP Growth {t-1 to t-4} Quarterly Change in the Fed 0.074 0.027 0.034 0.023 0.090 0.131 0.125 0.158 0.750 0.735 0.718 0.729 Funds Rate * Ratio {t-1 to t-4} [0.00378]*** [0.00393]*** [0.00398]*** [0.00464]*** [0.0268]*** [0.0268]*** [0.0268]*** [0.0268]*** [0.339]** [0.332]** [0.329]** [0.329]** Quarterly Change in the Fed -0.038 -0.019 -0.024 -0.020 -0.047 -0.074 -0.073 -0.095 -0.789 -0.809 -0.793 -0.815 Funds Rate * Ratio * GDP [0.0025]*** [0.00261]*** [0.00265]*** [0.00323]*** [0.019]** [0.0191]*** [0.0191]*** [0.0191]*** [0.303]***[0.292]*** [0.291]*** [0.295]*** Growth {t-1 to t-4} 0.024 0.020 0.027 0.029 0.188 0.152 0.141 0.142 -0.013 -0.188 -0.186 -0.171 Ratio {t-1 to t-4} [0.00269]*** [0.00303]*** [0.0031]*** [0.00358]*** [0.0233]*** [0.0235]*** [0.0235]*** [0.0235]*** [0.391] [0.394] [0.393] [0.388] -0.010 0.012 0.007 0.008 -0.033 -0.034 -0.027 -0.030 0.143 0.178 0.174 0.166 Ratio * GDP Growth {t-1 to t-4} [0.00181]*** [0.00201]*** [0.00206]*** [0.00243]*** [0.0165]** [0.0165]** [0.0165] [0.0165]* [0.329] [0.323] [0.321] [0.316] Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls No Yes Yes n/p No Yes Yes n/p No Yes Yes n/p Demand Controls No No Yes n/p No No Yes n/p No No Yes n/p Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects No No No Yes No No No Yes No No No Yes R-squared 0.20 0.21 0.20 0.22 0.20 0.21 0.21 0.21 0.14 0.16 0.16 0.17 Number of Observations 821,146 821,146 787,422 821,146 523,517 523,517 523,167 523,517 544,999 544,999 544,636 544,999 Interaction of changes in the federal funds rate with the securities to assets or leverage ratio, cumulative over four quarters, at the 10th percentile of GDP growth 0.051 0.015 0.020 0.011 0.062 0.087 0.081 0.101 0.277 0.250 0.242 0.240 Interaction of changes in the federal funds rate with the securities to assets or leverage ratio, cumulative over four quarters, at the 90th percentile of GDP growth 0.004 -0.009 -0.010 -0.015 0.004 -0.005 -0.009 -0.017 -0.702 -0.754 -0.741 -0.771 Note:ThetableaboveshowstheimpactofU.S.monetarypolicyonU.S.banks'totallending,between1986and2008.Thedependentvariableisthequarterlychangeinthenaturallogarithmofeachbank'sstockof totalloans.Columns1through8showsresultsforthelendingof"small"banks,where"small"indicatesabankwhosetotalassetsizeisbelowthe95thpercentileofthecross-sectionalsizedistributionofallU.S.banks inagivenquarter."Ratio"indicatesthesecuritiestoassetsratioinColumns1through4,andtheLeverageRatioinColumns5through12.Thecoefficientsshowthecumulativeeffect(overtheprecedingfourquarters) ofaoneunitchangeintheexplanatoryvariable(aslistedintheleftmostcolumn)onpercentagepointchangesinlendingflows(asindicatedatthetopofeachcolumn).Variabledefinitionsandsummarystatisticsare shown in Table 1. Standard errors, clustered at the bank level, are shown below the coefficients. Stars indicate the significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1. 28

Table 3: The impact of changes in U.S. monetary policy on the business lending of "small" U.S. banks with different securities to assets ratios: 1986-2008 [1] [2] [3] [4] [5] [6] [7] [8] Lending flows: Commercial Commercial Commercial Commercial Commercial Commercial Commercial Commercial Variables & Industrial & Industrial & Industrial & Industrial Real Estate Real Estate Real Estate Real Estate 1.637 1.014 0.684 -0.812 -1.103 -0.064 Quarterly Change in the Fed Funds Rate {t-1 to t-4} [0.163]*** [0.213]*** [0.265]*** [0.0178]*** [0.0174]*** [0.0221]*** Quarterly Change in the Fed Funds Rate * GDP Growth -0.778 -0.491 -0.425 0.581 0.909 1.318 {t-1 to t-4} [0.118]*** [0.141]*** [0.161]*** [0.0143]*** [0.0121]*** [0.0133]*** Quarterly Change in the Fed Funds Rate * Securities to 0.076 0.043 0.055 0.022 0.045 -0.001 -0.027 0.000 Assets Ratio {t-1 to t-4} [0.0083]*** [0.00883]*** [0.00904]*** [0.0102]** [0.000848]*** [0.000843] [0.00125]*** [0.000518] Quarterly Change in the Fed Funds Rate * Securities to -0.024 -0.014 -0.021 -0.009 -0.019 0.004 0.013 0.000 Assets Ratio * GDP Growth {t-1 to t-4} [0.00551]*** [0.00583]** [0.00596]*** [0.00701] [0.000604]***[0.000528]***[0.000754]*** [0.000367] 0.026 0.031 0.027 0.032 -0.035 0.015 0.033 -0.001 Securities to Assets Ratio {t-1 to t-4} [0.00603]*** [0.00667]*** [0.00685]*** [0.00782]*** [0.000603]*** [0.000558]*** [0.000601]*** [0.000457] -0.024 -0.005 -0.003 -0.005 0.003 -0.022 -0.032 0.001 Securities to Assets Ratio * GDP Growth {t-1 to t-4} [0.00406]*** [0.00445] [0.00457] [0.00533] [0.000382]*** [0.000374]*** [0.000418]*** [0.000314]*** Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls No Yes Yes n/p No Yes Yes n/p Demand Controls No No Yes n/p No No Yes n/p Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects No No No Yes No No No Yes R-squared 0.06 0.06 0.05 0.06 0.31 0.53 0.59 0.78 Number of Observations 743,255 743,255 714,507 743,255 823,290 823,290 789,422 823,290 Interaction of changes in the federal funds rate with the securities to assets ratio, cumulative over four quarters, at the 10th percentile of GDP growth 0.061 0.035 0.043 0.017 0.034 0.001 -0.019 0.000 Interaction of changes in the federal funds rate with the securities to assets ratio, cumulative over four quarters, at the 90th percentile of GDP growth 0.031 0.017 0.017 0.005 0.010 0.006 -0.003 0.000 Note: The table above showsthe impactof U.S. monetarypolicyonU.S. banks' business(C&I andCRE)lending,between1986and2008.The dependentvariable isthe quarterlychange inthe naturallogarithmofeach"small"bank'sstockofC&I(Columns1through4)andCRE(Columns5through8)loans,where"small"indicatesabankwhosetotalassetsizeisbelowthe95thpercentile ofthecross-sectionalsizedistributionofallU.S.banksinagivenquarter.The coefficientsshow thecumulative effect(overtheprecedingfourquarters)ofa oneunitchangeintheexplanatory variable(aslistedintheleftmostcolumn)onpercentagepointchangesinlendingflows(asindicatedatthetopofeachcolumn).VariabledefinitionsandsummarystatisticsareshowninTable1. Standard errors, clustered at the bank level, are shown below the coefficients. Stars indicate the significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1. 29

Table 4: The impact of changes in U.S. monetary policy on the household lending of "small" U.S. banks with different securities to assets ratios: 1986-2008 [1] [2] [3] [4] [5] [6] [7] [8] Lending flows: Residential Residential Residential Residential Consumer Consumer Consumer Consumer Variables Real Estate Real Estate Real Estate Real Estate -1.293 -0.706 -0.337 1.675 1.476 1.261 Quarterly Change in the Fed Funds Rate {t-1 to t-4} [0.114]*** [0.143]*** [0.179]* [0.122]*** [0.158]*** [0.193]*** Quarterly Change in the Fed Funds Rate * GDP Growth 0.548 -0.229 0.037 -0.450 -0.380 -0.067 {t-1 to t-4} [0.0837]*** [0.0987]** [0.112] [0.0895]*** [0.107]*** [0.12] Quarterly Change in the Fed Funds Rate * Securities to 0.096 0.065 0.055 0.052 0.057 0.015 0.024 0.015 Assets Ratio {t-1 to t-4} [0.00521]*** [0.00553]*** [0.00562]*** [0.00639]*** [0.0056]*** [0.00596]** [0.00608]*** [0.00684]** Quarterly Change in the Fed Funds Rate * Securities to -0.053 -0.041 -0.036 -0.035 -0.025 -0.017 -0.022 -0.012 Assets Ratio * GDP Growth {t-1 to t-4} [0.00348]*** [0.00369]*** [0.00375]*** [0.0045]*** [0.00373]*** [0.00395]*** [0.00402]*** [0.00471]** 0.046 0.024 0.044 0.040 0.019 -0.008 -0.004 0.027 Securities to Assets Ratio {t-1 to t-4} [0.00395]*** [0.0044]*** [0.00449]*** [0.00526]*** [0.00405]*** [0.00455]* [0.00464] [0.00527]*** -0.025 -0.006 -0.017 -0.013 -0.002 0.032 0.029 0.005 Securities to Assets Ratio * GDP Growth {t-1 to t-4} [0.00265]*** [0.00293]** [0.00299]*** [0.00358]*** [0.00272] [0.00306]*** [0.00311]*** [0.00362] Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls No Yes Yes n/p No Yes Yes n/p Demand Controls No No Yes n/p No No Yes n/p Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects No No No Yes No No No Yes R-squared 0.08 0.08 0.08 0.09 0.08 0.09 0.08 0.10 Number of Observations 781,190 781,190 752,037 781,190 780,257 780,257 749,836 780,257 Interaction of changes in the federal funds rate with the securities to assets ratio, cumulative over four quarters, at the 10th percentile of GDP growth 0.064 0.040 0.033 0.031 0.042 0.005 0.011 0.008 Interaction of changes in the federal funds rate with the securities to assets ratio, cumulative over four quarters, at the 90th percentile of GDP growth -0.002 -0.012 -0.012 -0.013 0.011 -0.016 -0.016 -0.006 Note:ThetableaboveshowstheimpactofU.S.monetarypolicyonU.S.banks'household(residentialandconsumer)lending,between1986and2008.Thedependentvariableisthequarterlychangein thenaturallogarithmofeach"small"bank'sstockofresidentialrealestate(Columns1through4)andconsumer(Columns5through8)loans,where"small"indicatesa bankwhosetotalassetsizeis belowthe95thpercentileofthecross-sectionalsizedistributionofallU.S.banksinagivenquarter.Thecoefficientsshowthecumulativeeffect(overtheprecedingfourquarters)ofaoneunitchangein theexplanatoryvariable(aslistedintheleftmostcolumn)onpercentagepointchangesinlendingflows(asindicatedatthetopofeachcolumn).Variabledefinitionsandsummarystatisticsareshownin Table 1. Standard errors, clustered at the bank level, are shown below the coefficients. Stars indicate the significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1. 30

Table 5: The impact of changes in U.S. monetary policy on the business lending of "small" U.S. banks with different leverage ratios: 1986-2008 [1] [2] [3] [4] [5] [6] [7] [8] Lending flows: Commercial & Commercial Commercial Commercial Commercial Commercial Commercial Commercial Variables Industrial & Industrial & Industrial & Industrial Real Estate Real Estate Real Estate Real Estate 0.211 0.241 -0.060 0.424 0.775 -0.140 Quarterly Change in the Fed Funds Rate {t-1 to t-4} [0.633] [0.664] [0.703] [0.0755]*** [0.0649]*** [0.0651]** Quarterly Change in the Fed Funds Rate * GDP Growth 0.752 0.121 0.440 0.372 0.185 1.223 {t-1 to t-4} [0.457]* [0.471] [0.494] [0.0546]*** [0.0443]*** [0.0455]*** Quarterly Change in the Fed Funds Rate * Leverage 0.132 0.162 0.150 0.170 -0.097 -0.022 -0.005 0.008 Ratio {t-1 to t-4} [0.0638]** [0.0639]** [0.0639]** [0.0641]*** [0.00745]*** [0.0062]*** [0.00609] [0.00599] Quarterly Change in the Fed Funds Rate * Leverage -0.085 -0.091 -0.086 -0.095 0.080 0.020 0.009 0.004 Ratio * GDP Growth {t-1 to t-4} [0.0463]* [0.0464]* [0.0465]* [0.0466]** [0.0054]*** [0.00431]*** [0.00427]** [0.00422] 0.282 0.243 0.220 0.208 0.070 -0.040 -0.012 -0.002 Leverage Ratio {t-1 to t-4} [0.0531]*** [0.0536]*** [0.0536]*** [0.0538]*** [0.00587]*** [0.00399]*** [0.00358]*** [0.00337] -0.097 -0.105 -0.091 -0.087 0.024 0.025 0.009 0.000 Leverage Ratio * GDP Growth {t-1 to t-4} [0.0379]** [0.0381]*** [0.0381]** [0.0383]** [0.00406]*** [0.00298]*** [0.00264]*** [0.00249] Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls No Yes Yes n/p No Yes Yes n/p Demand Controls No No Yes n/p No No Yes n/p Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects No No No Yes No No No Yes R-squared 0.06 0.06 0.06 0.06 0.48 0.71 0.74 0.77 Number of Observations 482,104 482,104 481,815 482,104 524,617 524,617 524,264 524,617 Interaction of changes in the federal funds rate with the leverage ratio, cumulative over four quarters, at the 10th percentile of GDP growth 0.081 0.108 0.099 0.113 -0.050 -0.010 0.001 0.011 Interaction of changes in the federal funds rate with the leverage ratio, cumulative over four quarters, at the 90th percentile of GDP growth -0.024 -0.005 -0.008 -0.005 0.049 0.015 0.012 0.015 Note:ThetableaboveshowstheimpactofU.S.monetarypolicyonU.S.banks'business(C&IandCRE)lending,between1986and2008.Thedependentvariableisthequarterlychangeinthenatural logarithmofeach"small"bank'sstockofC&I(Columns1through4)andCRE(Columns5through8)loans,where"small"indicatesabankwhosetotalassetsizeisbelowthe95thpercentileofthe cross-sectionalsizedistributionofallU.S.banksinagivenquarte.Thecoefficientsshowthecumulativeeffect(overtheprecedingfourquarters)ofaoneunitchangeintheexplanatoryvariable(as listedintheleftmostcolumn)onpercentagepointchangesinlendingflows(asindicatedatthetopofeachcolumn).VariabledefinitionsandsummarystatisticsareshowninTable1.Standarderrors, clustered at the bank level, are shown below the coefficients. Stars indicate the significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1. 31

Table 6: The impact of changes in U.S. monetary policy on the household lending of "small" U.S. banks with different leverage ratios: 1986-2008 [1] [2] [3] [4] [5] [6] [7] [8] Lending flows: Residential Residential Residential Residential Consumer Consumer Consumer Consumer Variables Real Estate Real Estate Real Estate Real Estate -1.106 -0.081 -1.022 1.471 1.022 -0.416 Quarterly Change in the Fed Funds Rate {t-1 to t-4} [0.399]*** [0.415] [0.441]** [0.457]*** [0.482]** [0.509] Quarterly Change in the Fed Funds Rate * GDP 0.653 -0.431 0.715 -0.204 -0.572 0.986 Growth {t-1 to t-4} [0.293]** [0.301] [0.315]** [0.318] [0.33]* [0.348]*** Quarterly Change in the Fed Funds Rate * Leverage 0.061 0.091 0.073 0.099 0.024 0.057 0.036 0.052 Ratio {t-1 to t-4} [0.0396] [0.0397]** [0.0397]* [0.0397]** [0.045] [0.0451] [0.0452] [0.0452] Quarterly Change in the Fed Funds Rate * Leverage -0.045 -0.050 -0.043 -0.056 -0.002 -0.003 0.008 -0.005 Ratio * GDP Growth {t-1 to t-4} [0.0292] [0.0293]* [0.0293] [0.0294]* [0.0315] [0.0317] [0.0317] [0.0317] 0.128 0.169 0.132 0.133 0.163 0.181 0.145 0.130 Leverage Ratio {t-1 to t-4} [0.0348]*** [0.035]*** [0.0351]*** [0.0352]*** [0.038]*** [0.0383]*** [0.0383]*** [0.0383]*** -0.016 -0.049 -0.026 -0.030 -0.032 -0.058 -0.037 -0.031 Leverage Ratio * GDP Growth {t-1 to t-4} [0.0247] [0.0248]** [0.0248] [0.0248] [0.0273] [0.0274]** [0.0274] [0.0274] Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls No Yes Yes n/p No Yes Yes n/p Demand Controls No No Yes n/p No No Yes n/p Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects No No No Yes No No No Yes R-squared 0.08 0.09 0.09 0.09 0.09 0.10 0.10 0.11 Number of Observations 507,201 507,201 506,876 507,201 503,045 503,045 502,713 503,045 Interaction of changes in the federal funds rate with the leverage ratio, cumulative over four quarters, at the 10th percentile of GDP growth 0.034 0.061 0.047 0.066 0.023 0.055 0.041 0.049 Interaction of changes in the federal funds rate with the leverage ratio, cumulative over four quarters, at the 90th percentile of GDP growth -0.021 -0.001 -0.005 -0.003 0.020 0.051 0.051 0.043 p yp y ( ) g, p the quarterlychange inthe naturallogarithmof each"small"bank'sstockofresidential(Columns1through4)andconsumer(Columns5through8)loans,where "small"indicates abank whosetotalassetsizeisbelowthe95thpercentileofthecross-sectionalsizedistributionofallU.S.banksinagivenquarter.Thecoefficientsshowthecumulativeeffect(overthepreceding four quarters) of a one unit change in the explanatory variable (as listed in the leftmost column) on percentage point changes in lending flows (as indicated at the top of each column). Variable definitionsandsummarystatisticsareshowninTable1.Standarderrors,clusteredatthebanklevel,areshownbelowthecoefficients.Starsindicatethesignificanceofthecoefficients:*** p<0.01, ** p<0.05, * p<0.1. 32

Table 7: Weighted regressions: The impact of changes in U.S. monetary policy on the business lending of U.S. banks with different leverage ratios: 1986-2008 [1] [2] [3] [4] [5] [6] [7] [8] Lending flows: Commercial & Commercial Commercial Commercial Commercial Commercial Commercial Commercial Variables Industrial & Industrial Real Estate & Industrial Real Estate Real Estate Real Estate Real Estate -8.346 -7.541 -6.962 0.774 2.786 2.124 Quarterly Change in the Fed Funds Rate {t-1 to t-4} [3.015]*** [2.893]*** [3.405]** [2.098] [2.278] [2.123] Quarterly Change in the Fed Funds Rate * GDP Growth {t- 9.784 7.460 7.070 0.082 -0.557 -0.120 1 to t-4} [2.975]*** [2.948]** [3.278]** [1.23] [1.194] [1.148] Quarterly Change in the Fed Funds Rate * Leverage Ratio 1.129 1.100 1.083 1.107 -0.151 -0.158 -0.158 -0.118 {t-1 to t-4} [0.325]*** [0.284]*** [0.282]*** [0.288]*** [0.201] [0.201] [0.209] [0.22] Quarterly Change in the Fed Funds Rate * Leverage Ratio -1.130 -1.134 -1.114 -1.121 0.109 0.059 0.059 0.039 * GDP Growth {t-1 to t-4} [0.304]*** [0.287]*** [0.285]*** [0.274]*** [0.127] [0.127] [0.129] [0.131] 0.078 -0.139 -0.146 -0.136 -0.116 -0.287 -0.241 -0.205 Leverage Ratio {t-1 to t-4} [0.474] [0.395] [0.395] [0.371] [0.19] [0.21] [0.211] [0.219] -0.261 -0.184 -0.184 -0.207 0.0417 0.105 0.0845 0.059 Leverage Ratio * GDP Growth {t-1 to t-4} [0.353] [0.295] [0.294] [0.278] [0.138] [0.151] [0.152] [0.159] Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls No Yes Yes n/p No Yes Yes n/p Demand Controls No No Yes n/p No No Yes n/p Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects No No No Yes No No No Yes R-squared 0.15 0.17 0.17 0.18 0.13 0.15 0.15 0.16 Number of Observations 499,935 499,935 499,637 499,935 543,903 543,903 543,537 543,903 Interaction of changes in the federal funds rate with the leverage ratio, cumulative over four quarters, at the 10th percentile of GDP growth 0.451 0.420 0.415 0.434 -0.086 -0.123 -0.123 -0.095 Interaction of changes in the federal funds rate with the leverage ratio, cumulative over four quarters, at the 90th percentile of GDP growth -0.950 -0.987 -0.967 -0.956 0.050 -0.050 -0.050 -0.047 Note:ThetableaboveshowstheimpactofU.S.monetarypolicyonU.S.banks'business(C&IandCRE)lending,between1986and2008.Thedependentvariableisthequarterlychangein thenaturallogarithmofeachbank'sstockofC&I(Columns1through4)andCRE(Columns5through8)loans.Eachobservationisweightedbythegivenbank'sshare inthe totalmarket segment.Thecoefficientsshowthecumulativeeffect(overtheprecedingfourquarters)ofaoneunitchangeintheexplanatoryvariable(aslistedinthe leftmostcolumn)onpercentage pointchangesinlendingflows(asindicatedatthetopofeachcolumn).VariabledefinitionsandsummarystatisticsareshowninTable1. Standarderrors,clusteredat thebanklevel,are shown below the coefficients. Stars indicate the significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1. 33

Table 8: Weighted regressions: The impact of changes in U.S. monetary policy on the household lending of U.S. banks with different leverage ratios: 1986-2008 [1] [2] [3] [4] [5] [6] [7] [8] Lending flows: Residential Residential Residential Residential Consumer Consumer Consumer Consumer Variables Real Estate Real Estate Real Estate Real Estate 7.409 13.950 15.180 -17.090 -15.780 -19.380 Quarterly Change in the Fed Funds Rate {t-1 to t-4} [4.198]* [6.356]** [7.703]** [4.788]*** [4.686]*** [4.819]*** Quarterly Change in the Fed Funds Rate * GDP Growth -2.770 -6.674 -8.447 15.230 13.060 15.870 {t-1 to t-4} [3.159] [4.838] [6.402] [3.939]*** [3.591]*** [3.456]*** Quarterly Change in the Fed Funds Rate * Leverage -0.885 -1.015 -0.981 -0.740 2.329 2.216 2.142 1.958 Ratio {t-1 to t-4} [0.407]** [0.554]* [0.529]* [0.456] [0.525]*** [0.515]*** [0.515]*** [0.533]*** Quarterly Change in the Fed Funds Rate * Leverage 0.345 0.329 0.315 0.146 -1.980 -1.901 -1.854 -1.733 Ratio * GDP Growth {t-1 to t-4} [0.315] [0.385] [0.378] [0.325] [0.454]*** [0.459]*** [0.45]*** [0.467]*** -1.047 -1.451 -1.420 -1.273 -0.027 -0.119 -0.163 -0.200 Leverage Ratio {t-1 to t-4} [0.552]* [0.735]** [0.713]** [0.6]** [0.528] [0.527] [0.54] [0.544] 1.140 1.331 1.298 1.174 0.151 0.166 0.201 0.263 Leverage Ratio * GDP Growth {t-1 to t-4} [0.417]*** [0.535]** [0.514]** [0.42]*** [0.414] [0.402] [0.408] [0.414] Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls No Yes Yes n/p No Yes Yes n/p Demand Controls No No Yes n/p No No Yes n/p Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects No No No Yes No No No Yes R-squared 0.18 0.20 0.20 0.22 0.20 0.21 0.21 0.23 Number of Observations 525,036 525,036 524,702 525,036 522,153 522,153 521,808 522,153 Interaction of changes in the federal funds rate with the leverage ratio, cumulative over four quarters, at the 10th percentile of GDP growth -0.678 -0.818 -0.792 -0.652 1.141 1.075 1.030 0.918 Interaction of changes in the federal funds rate with the leverage ratio, cumulative over four quarters, at the 90th percentile of GDP growth -0.250 -0.410 -0.401 -0.471 -1.314 -1.282 -1.269 -1.231 Note: The table above shows the impact of U.S. monetarypolicyonU.S. banks'household(residentialrealestateandconsumer)lending,between1986and2008. Thedependent variableisthequarterlychangeinthenaturallogarithmofeachbank'sstockofRRE(Columns1through4)andconsumer(Columns5through8)loans.Eachobservationisweightedby thegivenbank'sshareinthetotalmarketsegment.Thecoefficientsshowthecumulativeeffect(overtheprecedingfourquarters)ofaoneunitchangeintheexplanatoryvariable(as listedintheleftmostcolumn)onpercentagepointchangesinlendingflows(asindicatedatthetopofeachcolumn).VariabledefinitionsandsummarystatisticsareshowninTable1. Standard errors, clustered at the bank level, are shown below the coefficients. Stars indicate the significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1. 34

Table 9: IV estimations: The impact of changes in U.S. monetary policy on the lending of U.S. banks with different securities to assets or leverage ratios: 1986-2008 [1] [3] [2] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] Lending flows: Commercial Commercial Residential Commercial Commercial Residential Commercial Commercia Residential Total Consumer Total Consumer Total Consumer & Industrial Real Estate Real Estate & Industrial Real Estate Real Estate & Industrial l Real Real Estate Sample of banks: Small Small Small Small Small Small Small Small Small Small All All All All All Securities to Securities to Securities to Securities to Securities to Ratio Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Assets Assets Assets Assets Assets Variables Quarterly Change in the Fed Funds Rate * 0.134 0.124 -0.023 0.058 0.202 0.003 0.138 -0.006 -0.011 0.118 0.040 0.027 0.075 -0.508 0.597 Ratio {t-1 to t-4} [0.00613]*** [0.0144]*** [0.00154]*** [0.00936]*** [0.0102]*** [0.00989] [0.0219]*** [0.00311]* [0.0146] [0.0175]*** [0.129] [0.0885] [0.187] [0.463] [0.322]* Quarterly Change in -0.099 -0.090 0.003 -0.046 -0.140 -0.030 -0.045 0.007 -0.034 -0.087 -0.137 -0.009 -0.118 -0.008 -0.523 the Fed Funds Rate * [0.00424]*** [0.01]*** [0.000916]***[0.00652]*** [0.00715]*** [0.00643]*** [0.0143]*** [0.00204]*** [0.00941]***[0.0115]*** [0.0875] [0.0649] [0.158] [0.164] [0.22]** Ratio * GDP Growth {t- 1 to t-4} 0.088 0.098 -0.098 -0.027 0.106 0.088 0.610 -0.054 0.220 0.969 5.068 0.346 1.103 23.130 -4.489 Ratio {t-1 to t-4} [0.0105]*** [0.0239]*** [0.00292]*** [0.0166] [0.0174]*** [0.086] [0.192]*** [0.0216]** [0.131]* [0.153]*** [2.508]** [1.511] [2.048] [13.65]* [3.458] Ratio * GDP Growth {t- -0.021 0.020 0.036 -0.013 -0.015 0.022 0.004 0.034 0.017 0.017 0.037 0.007 -0.017 -0.137 0.196 1 to t-4} [0.00123]*** [0.00272]***[0.000357]***[0.00181]*** [0.00197]*** [0.00236]*** [0.00521] [0.000732]*** [0.00354]***[0.00414]***[0.0358] [0.0334] [0.0385] [0.0852] [0.118]* Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p Demand Controls n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.09 0.01 0.46 -0.01 0.02 -0.06 0.00 0.19 -0.05 -0.20 -0.64 0.00 -0.05 -3.85 -0.38 Number of Observations 683,236 623,385 697,873 657,835 652,807 523,297 481,841 528,907 506,970 502,802 544,780 543,687 499,678 524,808 521,908 Interaction of changes in the federal funds rate with the securities to assets or leverage ratio, cumulative over four quarters, at the 10th percentile of GDP growth 0.075 0.070 -0.021 0.030 0.118 -0.015 0.111 -0.002 -0.031 0.066 -0.042 0.022 0.004 -0.513 0.283 Interaction of changes in the federal funds rate with the securities to assets or leverage ratio, cumulative over four quarters, at the 90th percentile of GDP growth -0.048 -0.042 -0.017 -0.027 -0.057 -0.051 0.056 0.006 -0.074 -0.042 -0.212 0.010 -0.143 -0.523 -0.365 Note:ThetableaboveshowstheresultsofIVestimationsoftheimpactofU.S.monetarypolicyonU.S.banks' lendingtovariousmarketsegments,between1986and2008.TheinstrumentsforGDPgrowtharetheone-quarterlaggedGDP growthforecastvalues,andyearandquarterfixedeffects.Eachcolumnrepresentsthemostcompletespecificationforeachloantype(thatis,includingbankandtimefixedeffectsaswellasbankbalancesheetcontrols)fromTables2through 8.Thecoefficientsshowthecumulativeeffect(overtheprecedingfourquarters)ofaoneunitchangeintheexplanatoryvariable(aslistedintheleftmostcolumn)onpercentagepointchangesinlendingflows(asindicatedatthetopofeach column). Standard errors, clustered at the bank level, are shown below the coefficients. Stars indicate the significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1. 35

Table 10: Post-2007 period: The impact of changes in U.S. monetary policy on the lending of U.S. banks with different securities to assets or leverage ratios: 2008-2016 [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] Commercial Commercial Residential Commercial Commercial Residential Commercial Commercial Residential Total Consumer Total Consumer Total Consumer & Industrial Real Estate Real Estate & Industrial Real Estate Real Estate & Industrial Real Estate Real Estate Lending flows: Sample of banks: Small Small Small Small Small Small Small Small Small Small All All All All All Securities to Securities to Securities to Securities to Securities Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Ratio Assets Assets Assets Assets to Assets Variables Quarterly Change in the Short-term Shadow Rate * Ratio {t-1 to t-4} 0.003 -0.014 0.002 -0.003 -0.013 0.031 0.033 0.008 0.017 -0.043 -0.075 -0.062 0.088 -0.193 0.000 [0.00837] [0.0205] [0.00065]*** [0.00885] [0.0145] [0.0304] [0.0749] [0.0039]** [0.0433] [0.0563] [0.205] [0.211] [0.146] [0.512] [0.382] Quarterly Change in the Short-term Shadow Rate * Ratio * GDP Growth {t- 1 to t-4} -0.003 -0.003 -0.001 -0.012 -0.009 -0.027 -0.091 -0.001 -0.059 -0.053 0.223 0.084 -0.016 0.505 0.126 [0.00901] [0.0202] [0.000613]** [0.00912] [0.0136] [0.0297] [0.0712] [0.00366] [0.0428] [0.0551] [0.197] [0.192] [0.129] [0.346] [0.329] Ratio {t-1 to t-4} 0.038 0.031 0.001 0.023 0.029 0.268 0.400 -0.004 0.215 0.091 0.116 0.218 0.004 0.013 -0.290 [0.00765]*** [0.0154]** [0.000676] [0.00997]** [0.0118]** [0.0264]*** [0.0552]*** [0.00424] [0.037]*** [0.0456]** [0.157] [0.247] [0.132] [0.225] [0.229] Ratio * GDP Growth {t-1 -0.022 -0.028 -0.063 0.327 to t-4} -0.004 0.004 0.000 0.001 -0.001 0.000 0.011 -0.018 -0.001 -0.174 0.041 [0.00429] [0.00967] [0.000318] [0.0054] [0.00682] [0.0146] [0.0355] [0.00182] [0.0207] [0.0259] [0.129] [0.148] [0.0861] [0.166] [0.233] Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p Demand Controls n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.27 0.09 0.86 0.14 0.13 0.26 0.09 0.87 0.13 0.13 0.26 0.23 0.35 0.27 0.36 Number of Observations 177,257 167,032 179,235 173,327 164,957 160,753 152,098 161,842 157,447 150,546 168,637 159,278 168,856 163,921 156,933 Interaction of changes in the short-term shadow interest rate with the securities to assets or leverage ratio, cumulative over four quarters, at the 10th percentile of GDP growth 0.001 -0.016 0.002 -0.010 -0.018 0.015 -0.021 0.008 -0.019 -0.074 0.058 -0.012 0.079 0.110 0.076 Interaction of changes in the short-term shadow interest rate with the securities to assets or leverage ratio, cumulative over four quarters, at the 90th percentile of GDP growth -0.002 -0.020 0.000 -0.024 -0.029 -0.019 -0.134 0.007 -0.093 -0.140 0.335 0.092 0.058 0.736 0.232 Note:ThetableaboveshowstheimpactofU.S.monetarypolicyonU.S.banks' lendingtovariousmarketsegments,between2008and2016.Eachcolumnrepresentsthemostcompletespecificationforeachloantype (thatis,includingbankandtimefixedeffectsaswellasbankbalancesheetcontrols)fromTables2through8.Thecoefficientshowthecumulativeeffect(overtheprecedingfourquarters)ofaoneunitchangeinthe explanatoryvariable(aslistedintheleftmostcolumn)onpercentagepointchangesinlendingflows(asindicatedatthetopofeachcolumn).Standarderrors,clusteredatthebanklevel,areshownbelowthecoefficients. Stars indicate the significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1. 36

Table 1A: The impact of changes in U.S. monetary policy on the lending of U.S. banks with different securities to assets or leverage ratios: 1986-2008 [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] Commercial Commercial Residential Commercial Commercial Residential Commercial Commercial Residential Lending flows: Total Consumer Total Consumer Total Consumer & Industrial Real Estate Real Estate & Industrial Real Estate Real Estate & Industrial Real Estate Real Estate Sample of Banks: Small Small Small Small Small Small Small Small Small Small All All All All All Securities to Securities to Securities to Securities to Securities to Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Ratio Assets Assets Assets Assets Assets Variables Romer and Romer Residual {t-1 to t-4} -0.436 1.982 -0.447 -1.050 2.442 -0.625 1.763 -0.335 0.172 0.623 -13.860 -14.840 2.300 -7.610 -34.220 [0.144]*** [0.321]*** [0.0262]*** [0.222]*** [0.237]*** [0.542] [1.305] [0.0904]*** [0.819] [0.869] [7.018]** [6.847]** [4.096] [12.17] [12.39]*** Romer and Romer Residual * GDP Growth {t-1 to t-4} -0.126 -1.533 -0.086 0.251 -1.164 -0.433 -1.556 -0.335 -1.079 0.393 11.620 16.380 0.189 6.296 25.740 [0.108] [0.238]*** [0.0199]*** [0.165] [0.175]*** [0.414] [0.995] [0.0689]*** [0.625]* [0.659] [6.528]* [5.002]*** [2.888] [5.944] [9.779]*** Romer and Romer Residual * Ratio {t-1 to t- 4} 0.093 0.078 -0.008 0.083 0.076 -0.003 0.084 0.006 -0.014 0.141 1.945 2.427 -0.395 0.409 4.704 [0.00758]*** [0.0175]*** [0.00095]*** [0.0105]*** [0.0117]*** [0.0546] [0.132] [0.00814] [0.081] [0.0853]* [0.833]** [0.736]*** [0.407] [1.321] [1.332]*** Romer and Romer Residual * Ratio * GDP Growth {t-1 to t-4} -0.039 -0.016 0.012 -0.041 -0.033 0.083 0.001 0.009 0.098 -0.084 -1.593 -2.260 0.151 -0.410 -3.439 [0.00564]*** [0.013] [0.000682]***[0.00779]*** [0.0087]*** [0.0416]** [0.1] [0.00616] [0.0616] [0.0648] [0.792]** [0.552]*** [0.305] [0.681] [1.037]*** Ratio {t-1 to t-4} 0.017 -0.009 -0.013 0.038 -0.003 0.166 0.277 0.056 0.110 0.188 0.332 0.392 -0.066 -0.658 0.371 [0.00201]*** [0.00453]** [0.000271]***[0.00299]*** [0.00308] [0.0196]*** [0.0456]*** [0.00349]*** [0.0297]*** [0.0317]*** [0.241] [0.365] [0.153] [0.426] [0.439] Ratio * GDP Growth {t-1 to t-4} -0.008 -0.007 0.001 -0.022 0.008 -0.012 -0.072 -0.018 -0.012 -0.030 -0.093 -0.483 0.016 0.878 -0.148 [0.00136]*** [0.00313]** [0.000168]***[0.00203]***[0.00211]*** [0.0133] [0.0319]** [0.00215]*** [0.0205] [0.0221] [0.172] [0.278]* [0.113] [0.349]** [0.326] Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls No No No No No No No No No No No No No No No Demand Controls No No No No No No No No No No No No No No No Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects No No No No No No No No No No No No No No No R-squared 0.20 0.06 0.63 0.08 0.09 0.21 0.06 0.66 0.08 0.10 0.15 0.16 0.13 0.17 0.22 Number of Observations 801,708 725,451 818,486 762,592 762,359 506,014 465,926 511,473 490,306 486,711 526,638 482,979 525,484 507,414 505,096 Interaction of changes in the Romer and Romer residual with the securities to assets or leverage ratio at the 10th percentile of GDP growth 0.069 0.068 -0.002 0.058 0.056 0.046 0.085 0.012 0.045 0.091 0.989 1.071 -0.304 0.163 2.641 Interaction of changes in the Romer and Romer residual with the securities to assets or leverage ratio at the 90th percentile of GDP growth 0.020 0.049 0.013 0.007 0.015 0.150 0.087 0.023 0.167 -0.013 -0.986 -1.731 -0.117 -0.345 -1.624 Note:ThetableaboveshowstheimpactofU.S.monetarypolicyonU.S.banks' lendingtovariousmarketsegments,between1986and2008.Eachcolumnrepresentsthebaselinespecificationforeachloantype(thatis, includingbankfixedeffectsaswellasbankbalancesheetcontrols)fromTables2through8.FourlagsofeachindependentvariablesareincludedintheIVregressions.Thecoefficientsshowthe effect(overthepreceding quarter)ofaoneunitchangeintheexplanatoryvariable(aslistedintheleftmostcolumn)onpercentagepointchangesinlendingflows(asindicatedatthetopofeachcolumn).Standarderrors,clusteredatthebanklevel,are shown below the coefficients. Stars indicate the significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1. 37

Table 2A: The impact of changes in U.S. monetary policy on the lending of U.S. banks with different securities to assets or leverage ratios when GDP is "below trend": 1986-2008 [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] Commercial Commercial Residential Commercial Commercial Residential Commercial Commercial Residential Total Consumer Total Consumer Total Consumer & Industrial Real Estate Real Estate & Industrial Real Estate Real Estate & Industrial Real Estate Real Estate Lending flows: Sample of banks: Small Small Small Small Small Small Small Small Small Small All All All All All Securities to Securities to Securities to Securities to Securities to Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Leverage Ratio Assets Assets Assets Assets Assets Variables Quarterly Change in the Fed Funds Rate {t-1 to t-4} -0.0731 1.445 -0.163 -1.707 0.825 -0.957 -0.15 0.593 -1.449 0.599 -3.375 -9.6 -0.00437 4.677 -7.245 [0.0856] [0.186]*** [0.0162]*** [0.131]*** [0.141]*** [0.3]*** [0.709] [0.072]*** [0.443]*** [0.513] [3.507] [3.533]*** [2.25] [3.399] [5.842] Quarterly Change in the Fed Funds Rate * GDP Growth {t-1 to t- 4} 0.153 -0.381 -0.15 0.764 0.329 0.639 1.096 -0.24 0.802 0.424 3.969 8.996 0.217 -2.056 6.228 [0.0644]** [0.138]*** [0.0119]*** [0.0979]*** [0.105]*** [0.211]*** [0.511]** [0.0481]*** [0.329]** [0.357] [1.922]** [2.61]*** [1.416] [2.54] [4.82] Quarterly Change in the Fed Funds Rate * Ratio {t-1 to t-4} 0.076 0.083 0.040 0.106 0.070 0.132 0.171 -0.040 0.088 0.084 0.399 1.137 -0.063 -0.736 0.986 [0.00417]***[0.00907]***[0.00064]*** [0.00563]*** [0.00615]*** [0.03]*** [0.0713]** [0.007]*** [0.044]** [0.0507]* [0.388] [0.398]*** [0.229] [0.317]** [0.645] Quarterly Change in the Fed Funds Rate * Ratio * GDP Growth {t-1 to t-4} -0.035 -0.029 -0.024 -0.055 -0.032 -0.068 -0.097 0.032 -0.052 -0.031 -0.476 -0.962 0.078 0.368 -0.791 [0.0029]*** [0.0062]***[0.000422]***[0.00384]*** [0.00423]***[0.0213]*** [0.0518]* [0.00485]*** [0.033] [0.0354] [0.244]* [0.285]*** [0.139] [0.244] [0.53] Ratio {t-1 to t-4} 0.015 0.013 -0.007 0.051 0.010 0.211 0.286 0.036 0.108 0.175 -0.104 0.086 -0.158 -0.753 -0.165 [0.00299]*** [0.00668]**[0.000391]*** [0.0044]*** [0.0045]** [0.0276]*** [0.0629]*** [0.00405]*** [0.0412]***[0.0445]*** [0.5] [0.369] [0.23] [0.45]* [0.601] Ratio * GDP Growth -0.049 -0.100 0.958 0.298 {t-1 to t-4} 0.004 -0.008 -0.001 -0.024 0.013 -0.002 -0.005 -0.049 0.228 -0.307 0.072 [0.00208]* [0.0046]* [0.000273]***[0.00304]*** [0.00311]*** [0.0202]** [0.0463]** [0.00328] [0.0299] [0.0327] [0.438] [0.275] [0.163] [0.378]** [0.472] Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls No No No No No No No No No No No No No No No Demand Controls No No No No No No No No No No No No No No No Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects No No No No No No No No No No No No No No No R-squared 0.20 0.07 0.67 0.09 0.09 0.21 0.07 0.67 0.09 0.11 0.17 0.18 0.16 0.24 0.24 Number of Observations 627,260 564,790 639,942 593,918 594,950 380,295 348,236 384,402 367,727 365,061 395,600 360,763 394,384 380,413 378,779 Interaction of changes in the federal funds rate with the securities to assets or leverage ratio, cumulative over four quarters, at the 10th percentile of GDP growth 0.055 0.066 0.026 0.073 0.050 0.091 0.113 -0.021 0.057 0.065 0.113 0.560 -0.016 -0.515 0.511 Interaction of changes in the federal funds rate with the securities to assets or leverage ratio, cumulative over four quarters, at the 90th percentile of GDP growth 0.011 0.030 -0.004 0.005 0.010 0.007 -0.008 0.018 -0.008 0.027 -0.477 -0.633 0.080 -0.059 -0.469 Note:ThetableaboveshowstheimpactofU.S.monetarypolicyonU.S.banks' lendingtovariousmarketsegments,between1986and2008.Eachcolumnrepresentsthefirstspecificationforeachloantype(t3ha8t is, includingtimefixedeffectsaswellasbankbalancesheetcontrols)fromTables2through8.Thecoefficientshowthecumulativeeffect(overtheprecedingfourquarters)ofaoneunitchangeintheexplanatoryvariable (aslistedintheleftmostcolumn)onpercentagepointchangesinlendingflows(asindicatedatthetopofeachcolumn).Standarderrors,clusteredatthebanklevel,areshownbelowthecoefficients.Starsindicatethe significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1.

Table 3A: The impact of changes in U.S. monetary policy on the lending of "small" U.S. banks with different securities to assets ratios: 1986-2008 Monetary easing vs. tightening [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] Monetary Policy Environment: Monetary Policy Easing Monetary Policy Tightening Commercial Commercial Residential Commercial Commercial Residential Lending flows: Total Consumer Total Consumer & Industrial Real Estate Real Estate & Industrial Real Estate Real Estate Variables Quarterly Change in the Fed Funds Rate 0.042 0.016 -0.001 0.036 0.006 0.047 0.024 0.002 0.056 0.037 * Securities to Assets Ratio {t-1 to t-4} [0.00744]*** [0.016] [0.000816] [0.0109]*** [0.0109] [0.00904]*** [0.0214] [0.00104]* [0.0128]*** [0.0142]*** Quarterly Change in the Fed Funds Rate * Securities to Assets Ratio * GDP -0.035 -0.010 0.000 -0.028 -0.016 -0.026 -0.007 0.000 -0.041 -0.005 Growth {t-1 to t-4} [0.00431]*** [0.0094] [0.000496] [0.00633]*** [0.00647]** [0.00604]*** [0.0146] [0.000684] [0.00884]*** [0.00975] Securities to Assets Ratio {t-1 to t-4} 0.042 0.045 0.000 0.029 0.029 0.019 0.023 0.000 0.059 0.008 [0.00517]*** [0.0112]*** [0.000627] [0.00782]*** [0.00751]*** [0.00708]*** [0.0162] [0.000912] [0.0105]*** [0.0113] Securities to Assets Ratio * GDP Growth -0.001 -0.013 0.001 -0.005 0.002 0.006 -0.006 0.000 -0.028 0.006 {t-1 to t-4} [0.00334] [0.00728]* [0.000402]** [0.00509] [0.00494] [0.00455] [0.0105] [0.000592] [0.00675]*** [0.00729] Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p Demand Controls n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.24 0.08 0.83 0.11 0.12 0.24 0.09 0.57 0.11 0.12 Number of Observations 473,987 428,398 483,852 449,522 450,607 347,159 314,857 354,575 331,668 329,650 Interaction of changes in the federal funds rate with the securities to assets ratio, cumulative over four quarters, at the 10th percentile of GDP growth 0.022 0.010 -0.001 0.019 -0.004 0.031 0.019 0.002 0.031 0.034 Interaction of changes in the federal funds rate with the securities to assets ratio, cumulative over four quarters, at the 90th percentile of GDP growth -0.022 -0.002 0.000 -0.016 -0.024 -0.001 0.010 0.002 -0.021 0.028 Note:ThetableaboveshowstheimpactofU.S.monetarypolicyonU.S.banks' lendingtovariousmarketsegments,between1986and2008,duringperiodsofmonetarypolicyeasing(Columns1through5)and monetarytightening(Columns6throug10).Eachcolumnrepresentsthemostcompletespecificationforeachloantype(thatis,includingbankandtimefixedeffectsaswellasbankbalancesheetcontrols)from Tables2through4.Thecoefficientshowthecumulativeeffect(overtheprecedingfourquarters)ofaoneunitchangeintheexplanatoryvariable(aslistedintheleftmostcolumn)onpercentagepointchangesin lending flows (as indicated at the top of each column). Standard errors, clustered at the bank level, are shown below the coefficients. Stars indicate the significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1. 39

Table 4A: The impact of changes in U.S. monetary policy on the lending of "small" U.S. banks with different leverage ratios: 1986-2008 Monetary easing vs. tightening [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] Monetary Policy Environment: Monetary Policy Easing Monetary Policy Tightening Commercial Commercial Residential Commercial Commercial Residential Lending flows: Total Consumer Total Consumer & Industrial Real Estate Real Estate & Industrial Real Estate Real Estate Variables Quarterly Change in the Fed Funds Rate * 0.212 0.074 0.013 0.004 -0.025 0.159 0.069 0.002 0.016 0.179 leverage Ratio {t-1 to t-4} [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** Quarterly Change in the Fed Funds Rate * -0.151 -0.109 -0.002 -0.046 0.011 -0.062 -0.038 0.007 0.045 -0.084 Leverage Ratio * GDP Growth {t-1 to t-4} [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** 0.141 0.075 0.003 0.053 0.084 0.159 0.204 -0.009 0.193 0.018 Leverage Ratio {t-1 to t-4} [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** Leverage Ratio * GDP Growth {t-1 to t-4} -0.037 -0.028 -0.003 0.020 -0.010 -0.057 -0.067 0.005 -0.086 0.032 [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** [0.00744]*** Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p Demand Controls n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.23 0.08 0.79 0.11 0.13 0.24 0.09 0.65 0.12 0.14 Number of Observations 316,477 289,886 319,898 305,580 304,175 207,040 192,218 209,235 201,621 198,870 Interaction of changes in the federal funds rate with the leverage ratio, cumulative over four quarters, at the 10th percentile of GDP growth 0.121 0.008 0.011 -0.024 -0.018 0.122 0.046 0.006 0.043 0.129 Interaction of changes in the federal funds rate with the leverage ratio, cumulative over four quarters, at the 90th percentile of GDP growth -0.067 -0.128 0.009 -0.082 -0.004 0.044 -0.002 0.014 0.100 0.024 Note:ThetableaboveshowstheimpactofU.S.monetarypolicyonU.S.banks' lendingtovariousmarketsegments,between1986and2008,duringperiodsofmonetarypolicyeasing(Columns1through5)and monetarytightening(Columns6throug10).Eachcolumnrepresentsthemostcompletespecificationforeachloantype(thatis,includingbankandtimefixedeffectsaswellasbankbalancesheetcontrols)fromTables 2,5and6.Thecoefficientshowthecumulativeeffect(overtheprecedingfourquarters)ofaoneunitchangeintheexplanatoryvariable(aslistedintheleftmostcolumn)onpercentagepointchangesinlendingflows (as indicated at the top of each column). Standard errors, clustered at the bank level, are shown below the coefficients. Stars indicate the significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1. 40

Table 5A: The impact of changes in U.S. monetary policy on the lending of U.S. banks with different leverage ratios: 1986-2008 Monetary easing vs. tightening [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] Monetary Policy Environment: Monetary Policy Easing Monetary Policy Tightening Commercial Commercial Residential Commercial Commercial Residential Lending flows: Total Consumer Total Consumer & Industrial Real Estate Real Estate & Industrial Real Estate Real Estate Variables Quarterly Change in the Fed Funds Rate * 0.382 1.304 0.069 -2.457 2.526 0.995 2.006 -1.418 0.770 1.475 Leverage Ratio {t-1 to t-4} [0.577] [0.613]** [0.257] [1.065]** [0.815]*** [0.655] [0.884]** [0.721]** [1.249] [1.354] Quarterly Change in the Fed Funds Rate * -0.651 -1.011 -0.146 0.862 -1.695 -1.089 -1.372 0.709 -1.004 -1.136 Leverage Ratio * GDP Growth {t-1 to t-4} [0.279]** [0.303]*** [0.148] [0.496]* [0.529]*** [0.617]* [0.657]** [0.318]** [0.862] [0.988] -0.352 -0.121 -0.174 -1.607 0.366 -0.386 -0.893 -0.637 -2.007 -2.411 Leverage Ratio {t-1 to t-4} [0.527] [0.529] [0.212] [0.713]** [0.711] [0.518] [0.938] [0.838] [1.458] [1.129]** 0.258 -0.154 0.024 1.350 -0.033 0.306 0.230 0.454 1.383 1.731 Leverage Ratio * GDP Growth {t-1 to t-4} [0.373] [0.349] [0.14] [0.466]*** [0.484] [0.34] [0.64] [0.595] [0.983] [0.831]** Bank Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Macro Controls n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p Demand Controls n/p n/p n/p n/p n/p n/p n/p n/p n/p n/p Bank Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Time Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.22 0.23 0.22 0.32 0.28 0.22 0.19 0.21 0.26 0.30 Number of Observations 328,961 300,050 327,843 315,855 315,320 216,038 199,885 216,060 209,181 206,833 Interaction of changes in the federal funds rate with the leverage ratio, cumulative over four quarters, at the 10th percentile of GDP growth -0.009 0.697 -0.019 -1.940 1.509 0.342 1.183 -0.993 0.168 0.793 Interaction of changes in the federal funds rate with the leverage ratio, cumulative over four quarters, at the 90th percentile of GDP growth -0.822 -0.566 -0.201 -0.862 -0.610 -1.020 -0.532 -0.106 -1.087 -0.627 Note:ThetableaboveshowstheimpactofU.S.monetarypolicyonU.S.banks' lendingtovariousmarketsegments,between1986and2008,duringperiodsofmonetarypolicyeasing(Columns1through5)and monetarytightening(Columns6throug10).Eachcolumnrepresentsthemostcompletespecificationforeachloantype(thatis,includingbankandtimefixedeffectsaswellasbankbalancesheetcontrols)fromTables 2,7and8.Thecoefficientshowthecumulativeeffect(overtheprecedingfourquarters)ofaoneunitchangeintheexplanatoryvariable(aslistedintheleftmostcolumn)onpercentagepointchangesinlendingflows (as indicated at the top of each column). Standard errors, clustered at the bank level, are shown below the coefficients. Stars indicate the significance of the coefficients: *** p<0.01, ** p<0.05, * p<0.1. 41