Got Rejected? Real Effects of Not Getting a Loan

Got Rejected? Real Effects of Not Getting a Loan Abstract Using a lender cutoff rule that generates plausibly exogenous variation in credit supply, I investigate a new channel through which funding shocks are transmitted to the real economy. Based on a sample of more than 15,000 loan applications from small- and medium-sized enterprises, I find that precautionary savings motives can aggravate real effects: low-liquidity firms whose loan applications were rejected increase cash holdings and cut noncash assets in excess of the requested loan amount. These results point to the amplifying effect of precautionary savings motives in the transmission of credit supply shocks. Received December 11, 2016; editorial decision February 22, 2018 by Editor Philip Strahan. Considerable attention has been devoted to the role of firms’ cash holdings. One theory postulates that firms hold cash for precautionary motives, because cash protects them against adverse funding shocks. However, little is known about the role of cash holdings in the transmission of funding shocks. Do firms draw down their cash holdings after a funding shock, thereby cushioning any real effects on asset growth, investment, and employment? Or do precautionary savings motives lead firms to increase cash holdings after a funding shock, thus amplifying real effects? This paper tries to fill this gap by analyzing plausibly exogenous variation in credit supply and tracing the subsequent development of cash holdings and real effects at affected firms over time. As is the case with many banks, the bank I look at (a major European lending institution) uses a cutoff rule when lending to small and medium-sized enterprises (SMEs). The data set consists of almost 17,000 loan applications from SMEs between 2009 and 2012. Each firm is assigned a continuous hard information rating. Loan applications with a rating better than the cutoff are accepted, while loan applications with a rating worse than the cutoff are subject to an additional review, leading to a sharp drop in the loan acceptance rate at the cutoff. This setup provides three unique features. First, the setup provides a variation in credit supply that is highly discrete in nature: firms just below and just above the cutoff are very similar in terms of credit quality, yet one group of firms has access to credit while the other group of firms does not. Second, the sample consists of firms that have all applied for a loan; that is, the setup allows me to clearly distinguish between credit supply and credit demand. Third, the lender cutoff rule imposes a credit quantity constraint on firms whose loan applications have been rejected: instead of demanding a higher interest rate, the bank at hand rejects near-quality loan applications. The setup therefore provides a fundamental example of credit rationing which, in contrast to bank-health-induced credit supply shocks, is prevalent during good and bad economic times. While financial constraints can translate into either higher cost of funds or credit quantity constraints, credit quantity constraints are more prevalent in practice, pointing to the importance of credit rationing in practice (Almeida and Campello 2001). Using a regression discontinuity design, I document the following effects: First, while larger firms (total assets above EUR 3 mil) are able to substitute the loss in funding from the sample bank, small firms (total assets below or equal to EUR 3 mil) are not. Consequently, small firms below the cutoff lose approximately 10% of their debt funding and need to cut their assets by 8% relative to small firms above the cutoff. Second, subsequent real effects crucially depend on the firms’ liquidity: firms with high liquidity—measured as the ratio of current assets to current liabilities—decrease their cash holdings after a credit supply shock. As a result, these firms are able to absorb the credit supply shock without a significant effect on asset growth, investments, and employment. In contrast, firms with low liquidity increase their cash holdings after a loan rejection. As a consequence, these firms need to cut noncash assets in excess of the quantity implied by the credit supply shock and thus, investment and employment decline significantly at these firms. These effects are also economically meaningful: a EUR 1 decrease in loan supply for a low-liquidity firm leads to a EUR 0.36 increase in cash holdings, a drop in investments of EUR 0.64 and a drop in current assets excluding cash of EUR 0.50. These results therefore point to the amplifying role of precautionary savings motives in the transmission of credit supply shocks. The underlying mechanism can plausibly be explained by firms updating their beliefs about future financing availability. A loan rejection decreases a firm’s belief about future financing availability and therefore increases the firm’s targeted level of precautionary savings. The regression discontinuity design relies on the assumption that firms just below and just above the cutoff are similar in all respects apart from the loan acceptance decision. Future growth opportunities are plausibly better for firms above the cutoff, resulting in an omitted variable bias if growth opportunities are not properly controlled for. The RDD aims to mitigate this concern by focusing on a narrow bandwidth around the cutoff. However, the typical bias-versus-precision tradeoff in an RDD might lead to a bandwidth that is too large to entirely rule out this concern. I therefore control, among other variables, for pre-application growth, industry x time fixed effects, and region x time fixed effects, all of which proxy for future growth opportunities. Furthermore, using a holdout sample of loan applications larger than EUR 1 million, in which the bank does not use the same cutoff, I do not find any significant effects. This supports the evidence from the main sample, suggesting that differences in post-application growth rates across the cutoff can indeed be attributed to the discrete change in credit constraints at the cutoff.1 The paper lies at the intersection of the literature on the corporate demand for liquidity and the literature on credit constraints. The work by Duchin, Ozbas, and Sensoy (2010) has highlighted the importance of cash holdings for the transmission of bank-health-induced credit supply shocks. This paper complements this work in two important dimensions: first, I use cash holdings as a left-hand-side variable and find that low-liquidity firms increase cash holdings following a credit supply shock. Second, whereas Duchin et al. (2010) show that cash accumulated in the past can help to dampen credit supply shocks, I show that precautionary savings motives can have exactly the opposite effect: firms update their beliefs about the optimal level of cash holdings after being subject to a credit supply shock, leading them to cut noncash assets in excess of the original credit supply shock. Keynes (1936) was the first to argue that liquidity helps financially constrained firms to pursue profitable investment opportunities when they occur. This precautionary savings motive for holding cash is formally modeled in Almeida et al. (2004) who show that financially constrained firms save a positive fraction of their cash flows, while unconstrained firms do not. Extensions of this idea include Han and Qiu (2007) who show that cash flow volatility is positively related to firms’ precautionary savings demands; and Acharya, Almeida and Campello (2007) who model the trade-off between saving and reducing short-term debt to show that constrained firms save cash instead of reducing short-term debt whenever their hedging needs are high. Lin and Paravisini (2011) empirically document that financially constrained firms hold more cash and exhibit higher operating cash flow risk and higher stock market betas. Using a large sample from 1980 to 2006, Bates et al. (2009) confirm that precautionary savings motives play an important role in explaining cash ratios at U.S. industrial firms. Riddick and Whited (2009) caution against using a simple correlation between savings and cash flow to gauge precautionary savings motives. These simple correlations might be misleading if productivity shocks are serially correlated and firms thus tend to invest more and save less after a positive productivity and cash flow shock. This paper adds to the literature by identifying a plausibly exogenous shock to credit supply and identifying the subsequent change in firms’ cash holdings. I find that firms with low liquidity increase their cash holdings after the credit supply shock, thus pointing to the crucial role of precautionary savings motives for financially constrained firms. The literature on credit rationing (Stiglitz and Weiss 1981; Sofianos, Wachtel, and Melnik 1990; Berger and Udell 1992; Banerjee and Duflo 2014), or more generally, on real effects of financial constraints (Fazzari, Hubbard, and Petersen 1988; Lamont 1997; Rauh 2006; Campello, Graham, and Harvey 2010; Faulklender and Petersen 2012; Banerjee and Duflo 2014) has seen an increasing awareness since the financial crisis. Supply of credit via banks can have significant real effects (Bernanke 1983). Prior literature has analyzed real effects of credit constraints either due to changes in monetary policy (Gertler and Gilchrist 1994; Kashyap and Stein 2000; Jiménez et al. 2012), due to dispersion in lender health (Gan 2007; Duchin et al. 2010; Chodorow-Reich 2014; Acharya et al. Forthcoming; Balduzzi, Brancati, and Schiantarelli 2017; Cingano, Manaresi, and Sette 2013; Bentolila et al. 2017; Popov and Rocholl Forthcoming), or due to debt maturity effects (Almeida et al. 2012). This paper adds to the literature by highlighting the importance of liquidity holdings in the transmission of credit supply shocks. In particular, my results point to the amplifying effect of precautionary savings motives in the transmission of credit supply shocks. 1. Institutional Setup and Data 1.1 Loan granting process I access data on 16,855 SME loan applications from 13,484 firms between 2009 and 2012 from a major German bank. The size of the loan applications ranges from EUR 10,000 to EUR 1 million. For loan applications up to EUR 1 million, loan-granting decisions are governed by a cutoff regime that creates plausibly exogenous variation in the likelihood of receiving a loan.2 All loan applications are from limited liability firms outside the financial sector.3 I apply two filters to the original data: first, subsidiaries of larger firms are excluded from the sample because the existence of a parent company is likely to impair the effect of any credit supply shock. Second, I exclude firms with total assets of less than EUR 350,000 as these are only subject to very rudimentary disclosure requirements (this filter will be described in more detail below). Both filters together exclude less than 5% of the original sample. In the first step, the bank aggregates hard information from various sources (account activity, balance sheet and profit and loss data, firm type/age/location, and information from a private credit registry) into a continuous internal rating. This continuous internal rating ranges from 0.5 (best) to 11.5 (worst) and is mapped into rating grades ranging from 1 (best) to 11 (worst). Figure 1 depicts a distribution of rating grades for all loan applications. Figure 1 View largeDownload slide Distribution of ratings This figure provides a distribution of rating grades for the sample of all loan applications between January 2009 and December 2012. For variable definitions, see Table 1. Figure 1 View largeDownload slide Distribution of ratings This figure provides a distribution of rating grades for the sample of all loan applications between January 2009 and December 2012. For variable definitions, see Table 1. In the second step, loan applications are grouped into three distinct buckets. The loan officer can grant loan applications with a rating grade between 1 and 7 without consent from the risk management department. Loan applications with a rating grade of 8 or 9 are subject to further review by the risk management department, which then takes the final accept/reject decision.4 The risk management department can also reduce the loan volume granted, implying that the results from this paper should be interpreted as reflecting both the intensive and the extensive margin of loan supply.5 The risk management department bases their decisions on an analysis of the available data sources and can also request further details or clarification on some of the inputs. Such cutoff rules are widely used when granting loans; in particular because a more precise signal about an applicant’s credit quality is most valuable for applicants in the middle of the creditworthiness spectrum.6 This setup induces a discontinuity in the likelihood of loan application acceptance. The setup therefore provides a fundamental example of credit rationing: instead of demanding a higher interest rate, the bank at hand rejects near-quality loan applications. As can be seen from Figure 2, the likelihood of an accept-decision is over 80% for rating grades between 1 and 7, and it precipitously drops to 50% for rating grades 8 and 9.7 Figure 2 View largeDownload slide Loan acceptance rates by rating This figure depicts the likelihood of loan application acceptance as a function of the continuous rating for the sample of all loan applications between January 2009 and December 2012. For variable definitions, see Table 1. Figure 2 View largeDownload slide Loan acceptance rates by rating This figure depicts the likelihood of loan application acceptance as a function of the continuous rating for the sample of all loan applications between January 2009 and December 2012. For variable definitions, see Table 1. Finally, loan applications with a rating grade of 10–11 are subject to a separate “red-light-process” and lending criteria are akin to debtor-in-possession financing rules. Thus, there is another discontinuity in the likelihood of acceptance between rating grades 9 and 10. However, as the number of loan applications with a rating of 10–11 is very low (see Figure 1), the following analysis focuses on the discontinuity between rating grades 7 and 8. 1.2 Measuring real effects after the accept/reject decision Measuring real effects after the accept/reject decision requires company information in the year(s) after the loan application has been made. This information is not entirely available at the bank, in particular for firms whose loan applications have been rejected. I thus rely on annual reports that need to be filed according to mandatory disclosure requirements. Bureau van Dijk’s DAFNE database provides access to these data in a computer-accessible form. Matching of bank data to Bureau van Dijk’s DAFNE data base is straightforward, as both share a common identifier.8 1.2.1 Mandatory disclosure requirements In Germany, all limited-liability firms are required to disclose their financial statements within 12 months after the end of the fiscal year. These disclosure requirements are mandated by commercial law and are akin to Regulation S-X (“Form and content of and requirements for financial statements”) by the SEC in the United States. However, the scope of firms covered by the disclosure requirements is significantly broader compared to the United States: all firms with limited liability need to disclose financial statements, independent of whether or not they are publicly listed and independent of the number of owners of the firm. There are three exemptions from these disclosure requirements: First, as implied above by the term “limited liability,” the rule does not apply to firm types where owners have full personal liability for all obligations of the firm (e.g., sole proprietorships). Second, subsidiaries do not have to separately disclose their annual reports. The disclosure of the parent company’s financial statements has an exempting effect for subsidiaries. Third, different disclosure requirements apply to financial firms (banks and insurance companies). The sample at hand only includes nonfinancial firms with limited liability, and I exclude subsidiaries as per the discussion above. 1.2.2 Granularity of disclosure requirements The disclosure requirements explicitly specify the items that need to be disclosed. These rules are akin to §210.5 of Regulation S-X in the United States that lists and defines balance sheet items to be disclosed to the SEC. The granularity of the disclosure requirement varies by size of the corporation with size being measured via total assets, revenues, and the number of employees. I summarize the disclosure requirements in Table A1 (see the appendix). Table 1 provides variable definitions. Table 1 Explanation of variables Name  Source  Description  Ratings, cutoff, and loan acceptance  Rating  Bank  Internal continuous rating ranging from 0.5 (best) to 11.5 (worst)  Rating grade  Bank  Mapping of the continuous rating to rating grades, ranging from 1 (continuous rating from 0.5 to 1.5) to 11 (continuous rating from 10.5 to 11.5)  Cutoff (0/1)  Bank  Dummy variable equal to one if a loan application has a rating grade of 8 or worse, that is, cannot be directly accepted by the loan officer  Accepted (0/1)  Bank  Dummy equal to one if a loan offer is made to the client  Loan characteristics  Loan amount     Notional amount of the loan application in EUR ‘000  Collateralized (0/1)  Bank  Dummy equal to one if a loan is collateralized (either by a physical collateral or a third party guarantee)  Firm characteristics at the time of the loan application  Firm age  Bank  Age of the firm in years since incorporation  Relationship age  Bank  Number of years that the firm has had an account at the bank without interruption  Revenues  Bank  Revenues of the firm in EUR million according to its financial statement in the fiscal year prior to the loan application (based on German accounting standards)  Number of employees  Bank  Number of employees of the firm in the fiscal year prior to the loan application  Total assets  Bank  Total assets of the firm EUR million according to its financial statement in the fiscal year prior to the loan application (based on German accounting standards)  Cash and cash equivalents (CCE)/Total assets  Bank  Cash and cash equivalents from the fiscal year preceding the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. Cash and cash equivalents are defined as the sum of cash and marketable securities  Current assets excl. CCE/Total sssets  Bank  Current assets (i.e., short-term assets) excluding cash and cash equivalents from the fiscal year preceding the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. For the constituents of current assets, see Table A1  Investments/Total assets  Bank  Investment assets from the fiscal year preceding the loan application, scaled by total asset of the firm in the fiscal year preceding the loan application. For the constituents of investment assets, see Table A1  Equity-to-asset ratio  Bank  Equity-to-asset ratio of the firm according to its financial statement in the fiscal year prior to the loan application  EBIT margin  Bank  Ratio of EBIT (earnings before interest and taxes) to revenues of the firm according to its financial statement in the fiscal year prior to the loan application  Liquidity  Bank  Ratio of current assets to current liabilities of the firm according to its financial statement in the fiscal year prior to the loan application  Changes in firm characteristics after the time of the loan application  Change in loan volume with the bank  Bank  Percentage change in loan volume with the bank from 1 month prior to 3, 12, and 24 months after the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application  Change in assets  DAFNE  Percentage change in total assets from the fiscal year preceding the loan application to fiscal year following the loan application  Change in current assets  DAFNE  Percentage change in current assets (i.e., short-term assets) from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. For the constituents of current assets, see Table A1  Change in cash and cash equivalents  DAFNE  Percentage change in cash and cash equivalents from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. Cash and cash equivalents are defined as the sum of cash and marketable securities  Change in investments  DAFNE  Percentage change in investment assets (i.e., long-term assets) from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. For the constituents of investment assets, see Table A1  Change in debt  DAFNE  Percentage change in debt from the fiscal year preceding the loan application to fiscal year following the loan application (e.g., year-end 2009 and year-end 2011 for a loan application in 2010), scaled by total assets of the firm in the fiscal year preceding the loan application. Debt includes bonds, bank debt, and trade payable, see Table A1  Change in equity  DAFNE  Percentage change in equity from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application  Change in employment  DAFNE  Percentage change in employment from the fiscal year preceding the loan application to fiscal year following the loan application  Name  Source  Description  Ratings, cutoff, and loan acceptance  Rating  Bank  Internal continuous rating ranging from 0.5 (best) to 11.5 (worst)  Rating grade  Bank  Mapping of the continuous rating to rating grades, ranging from 1 (continuous rating from 0.5 to 1.5) to 11 (continuous rating from 10.5 to 11.5)  Cutoff (0/1)  Bank  Dummy variable equal to one if a loan application has a rating grade of 8 or worse, that is, cannot be directly accepted by the loan officer  Accepted (0/1)  Bank  Dummy equal to one if a loan offer is made to the client  Loan characteristics  Loan amount     Notional amount of the loan application in EUR ‘000  Collateralized (0/1)  Bank  Dummy equal to one if a loan is collateralized (either by a physical collateral or a third party guarantee)  Firm characteristics at the time of the loan application  Firm age  Bank  Age of the firm in years since incorporation  Relationship age  Bank  Number of years that the firm has had an account at the bank without interruption  Revenues  Bank  Revenues of the firm in EUR million according to its financial statement in the fiscal year prior to the loan application (based on German accounting standards)  Number of employees  Bank  Number of employees of the firm in the fiscal year prior to the loan application  Total assets  Bank  Total assets of the firm EUR million according to its financial statement in the fiscal year prior to the loan application (based on German accounting standards)  Cash and cash equivalents (CCE)/Total assets  Bank  Cash and cash equivalents from the fiscal year preceding the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. Cash and cash equivalents are defined as the sum of cash and marketable securities  Current assets excl. CCE/Total sssets  Bank  Current assets (i.e., short-term assets) excluding cash and cash equivalents from the fiscal year preceding the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. For the constituents of current assets, see Table A1  Investments/Total assets  Bank  Investment assets from the fiscal year preceding the loan application, scaled by total asset of the firm in the fiscal year preceding the loan application. For the constituents of investment assets, see Table A1  Equity-to-asset ratio  Bank  Equity-to-asset ratio of the firm according to its financial statement in the fiscal year prior to the loan application  EBIT margin  Bank  Ratio of EBIT (earnings before interest and taxes) to revenues of the firm according to its financial statement in the fiscal year prior to the loan application  Liquidity  Bank  Ratio of current assets to current liabilities of the firm according to its financial statement in the fiscal year prior to the loan application  Changes in firm characteristics after the time of the loan application  Change in loan volume with the bank  Bank  Percentage change in loan volume with the bank from 1 month prior to 3, 12, and 24 months after the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application  Change in assets  DAFNE  Percentage change in total assets from the fiscal year preceding the loan application to fiscal year following the loan application  Change in current assets  DAFNE  Percentage change in current assets (i.e., short-term assets) from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. For the constituents of current assets, see Table A1  Change in cash and cash equivalents  DAFNE  Percentage change in cash and cash equivalents from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. Cash and cash equivalents are defined as the sum of cash and marketable securities  Change in investments  DAFNE  Percentage change in investment assets (i.e., long-term assets) from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. For the constituents of investment assets, see Table A1  Change in debt  DAFNE  Percentage change in debt from the fiscal year preceding the loan application to fiscal year following the loan application (e.g., year-end 2009 and year-end 2011 for a loan application in 2010), scaled by total assets of the firm in the fiscal year preceding the loan application. Debt includes bonds, bank debt, and trade payable, see Table A1  Change in equity  DAFNE  Percentage change in equity from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application  Change in employment  DAFNE  Percentage change in employment from the fiscal year preceding the loan application to fiscal year following the loan application  “Bank” denotes that the variables comes from bank-internal data, and “Dafne” denotes that the variable comes from Bureau van Dijk’s Dafne database. All variables, except for Rating, Rating grade, and the dummy variables, are winsorized at the 1% and 99% level. Table 1 Explanation of variables Name  Source  Description  Ratings, cutoff, and loan acceptance  Rating  Bank  Internal continuous rating ranging from 0.5 (best) to 11.5 (worst)  Rating grade  Bank  Mapping of the continuous rating to rating grades, ranging from 1 (continuous rating from 0.5 to 1.5) to 11 (continuous rating from 10.5 to 11.5)  Cutoff (0/1)  Bank  Dummy variable equal to one if a loan application has a rating grade of 8 or worse, that is, cannot be directly accepted by the loan officer  Accepted (0/1)  Bank  Dummy equal to one if a loan offer is made to the client  Loan characteristics  Loan amount     Notional amount of the loan application in EUR ‘000  Collateralized (0/1)  Bank  Dummy equal to one if a loan is collateralized (either by a physical collateral or a third party guarantee)  Firm characteristics at the time of the loan application  Firm age  Bank  Age of the firm in years since incorporation  Relationship age  Bank  Number of years that the firm has had an account at the bank without interruption  Revenues  Bank  Revenues of the firm in EUR million according to its financial statement in the fiscal year prior to the loan application (based on German accounting standards)  Number of employees  Bank  Number of employees of the firm in the fiscal year prior to the loan application  Total assets  Bank  Total assets of the firm EUR million according to its financial statement in the fiscal year prior to the loan application (based on German accounting standards)  Cash and cash equivalents (CCE)/Total assets  Bank  Cash and cash equivalents from the fiscal year preceding the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. Cash and cash equivalents are defined as the sum of cash and marketable securities  Current assets excl. CCE/Total sssets  Bank  Current assets (i.e., short-term assets) excluding cash and cash equivalents from the fiscal year preceding the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. For the constituents of current assets, see Table A1  Investments/Total assets  Bank  Investment assets from the fiscal year preceding the loan application, scaled by total asset of the firm in the fiscal year preceding the loan application. For the constituents of investment assets, see Table A1  Equity-to-asset ratio  Bank  Equity-to-asset ratio of the firm according to its financial statement in the fiscal year prior to the loan application  EBIT margin  Bank  Ratio of EBIT (earnings before interest and taxes) to revenues of the firm according to its financial statement in the fiscal year prior to the loan application  Liquidity  Bank  Ratio of current assets to current liabilities of the firm according to its financial statement in the fiscal year prior to the loan application  Changes in firm characteristics after the time of the loan application  Change in loan volume with the bank  Bank  Percentage change in loan volume with the bank from 1 month prior to 3, 12, and 24 months after the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application  Change in assets  DAFNE  Percentage change in total assets from the fiscal year preceding the loan application to fiscal year following the loan application  Change in current assets  DAFNE  Percentage change in current assets (i.e., short-term assets) from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. For the constituents of current assets, see Table A1  Change in cash and cash equivalents  DAFNE  Percentage change in cash and cash equivalents from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. Cash and cash equivalents are defined as the sum of cash and marketable securities  Change in investments  DAFNE  Percentage change in investment assets (i.e., long-term assets) from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. For the constituents of investment assets, see Table A1  Change in debt  DAFNE  Percentage change in debt from the fiscal year preceding the loan application to fiscal year following the loan application (e.g., year-end 2009 and year-end 2011 for a loan application in 2010), scaled by total assets of the firm in the fiscal year preceding the loan application. Debt includes bonds, bank debt, and trade payable, see Table A1  Change in equity  DAFNE  Percentage change in equity from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application  Change in employment  DAFNE  Percentage change in employment from the fiscal year preceding the loan application to fiscal year following the loan application  Name  Source  Description  Ratings, cutoff, and loan acceptance  Rating  Bank  Internal continuous rating ranging from 0.5 (best) to 11.5 (worst)  Rating grade  Bank  Mapping of the continuous rating to rating grades, ranging from 1 (continuous rating from 0.5 to 1.5) to 11 (continuous rating from 10.5 to 11.5)  Cutoff (0/1)  Bank  Dummy variable equal to one if a loan application has a rating grade of 8 or worse, that is, cannot be directly accepted by the loan officer  Accepted (0/1)  Bank  Dummy equal to one if a loan offer is made to the client  Loan characteristics  Loan amount     Notional amount of the loan application in EUR ‘000  Collateralized (0/1)  Bank  Dummy equal to one if a loan is collateralized (either by a physical collateral or a third party guarantee)  Firm characteristics at the time of the loan application  Firm age  Bank  Age of the firm in years since incorporation  Relationship age  Bank  Number of years that the firm has had an account at the bank without interruption  Revenues  Bank  Revenues of the firm in EUR million according to its financial statement in the fiscal year prior to the loan application (based on German accounting standards)  Number of employees  Bank  Number of employees of the firm in the fiscal year prior to the loan application  Total assets  Bank  Total assets of the firm EUR million according to its financial statement in the fiscal year prior to the loan application (based on German accounting standards)  Cash and cash equivalents (CCE)/Total assets  Bank  Cash and cash equivalents from the fiscal year preceding the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. Cash and cash equivalents are defined as the sum of cash and marketable securities  Current assets excl. CCE/Total sssets  Bank  Current assets (i.e., short-term assets) excluding cash and cash equivalents from the fiscal year preceding the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. For the constituents of current assets, see Table A1  Investments/Total assets  Bank  Investment assets from the fiscal year preceding the loan application, scaled by total asset of the firm in the fiscal year preceding the loan application. For the constituents of investment assets, see Table A1  Equity-to-asset ratio  Bank  Equity-to-asset ratio of the firm according to its financial statement in the fiscal year prior to the loan application  EBIT margin  Bank  Ratio of EBIT (earnings before interest and taxes) to revenues of the firm according to its financial statement in the fiscal year prior to the loan application  Liquidity  Bank  Ratio of current assets to current liabilities of the firm according to its financial statement in the fiscal year prior to the loan application  Changes in firm characteristics after the time of the loan application  Change in loan volume with the bank  Bank  Percentage change in loan volume with the bank from 1 month prior to 3, 12, and 24 months after the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application  Change in assets  DAFNE  Percentage change in total assets from the fiscal year preceding the loan application to fiscal year following the loan application  Change in current assets  DAFNE  Percentage change in current assets (i.e., short-term assets) from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. For the constituents of current assets, see Table A1  Change in cash and cash equivalents  DAFNE  Percentage change in cash and cash equivalents from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. Cash and cash equivalents are defined as the sum of cash and marketable securities  Change in investments  DAFNE  Percentage change in investment assets (i.e., long-term assets) from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application. For the constituents of investment assets, see Table A1  Change in debt  DAFNE  Percentage change in debt from the fiscal year preceding the loan application to fiscal year following the loan application (e.g., year-end 2009 and year-end 2011 for a loan application in 2010), scaled by total assets of the firm in the fiscal year preceding the loan application. Debt includes bonds, bank debt, and trade payable, see Table A1  Change in equity  DAFNE  Percentage change in equity from the fiscal year preceding the loan application to fiscal year following the loan application, scaled by total assets of the firm in the fiscal year preceding the loan application  Change in employment  DAFNE  Percentage change in employment from the fiscal year preceding the loan application to fiscal year following the loan application  “Bank” denotes that the variables comes from bank-internal data, and “Dafne” denotes that the variable comes from Bureau van Dijk’s Dafne database. All variables, except for Rating, Rating grade, and the dummy variables, are winsorized at the 1% and 99% level. All firms that are subject to the disclosure requirements—independent of their size—need to disclose basic balance sheet items, consisting of two main items on the asset side and two main items on the liability side.9 The two main items on the asset side are current assets (i.e., short-term assets) and investment assets. The two main items on the liability side are equity and debt. The debt item combines both bank debt and trade payables. Firms that exceed two of three size criteria (1, EUR 350,000 in assets; 2, EUR 700,000 in revenues; 3, more than 10 employees) are subject to further disclosure requirements. These firms are required to further decompose the balance sheet items discussed above. In particular, current assets have to be decomposed into inventory, trade receivables, securities, and cash holdings; and investment assets have to be decomposed into intangible assets; property, plant, and equipment; and financial investments. As some of the following analyses require these items to be available, I exclude firms that are too small to be required to file these items (less than 5% of the original sample). Larger firms—those exceeding two of the following three criteria: (1) EUR 4.84 million in assets; (2) EUR 9.68 million in revenues; and (3) more than 50 employees—need to provide a further breakdown of asset and liability positions and disclose a profit and loss statement. These firms constitute only 25% of the firms in the sample and I thus do not use these items in the following analyses. 1.2.3 Time line for collection of data items I collect the data items for the year preceding the loan application, the year of the loan application and the year following the loan application. For example, for a loan application from May 2010 I collect data from the annual reports 2009, 2010, and 2011. In some cases, data are not available in the DAFNE database. This can be due to one of the following reasons: first, the firm is not active any more, either due to insolvency or because it was discontinued for different reasons. These firms can be clearly identified as any discontinuation and the respective cause has to be reported to the public register of corporations. Second, in a few cases, data are not available even though companies are legally required to file the data. I thoroughly check that any of these instances of missing data are not systematically related to a reject/accept decision in Table A2. 2. Descriptive Statistics Table 1 explains all variables in detail, and Table 2 presents descriptive statistics. All variables, except for the rating and the dummy variables, are winsorized at the 1% and 99% levels.10 The average rating is 5.78 (median: 6.00), that is, below the cutoff rating of 7.5 that defines risk management involvement. Figure 1 provides a rating distribution. The proportion of loan applications with a rating above the cutoff rating of is 81%, with 19% being below the cutoff rating. The average loan volume is EUR 527,000 (median: EUR 500,000), with 56% of the loans being collateralized. The mean loan volume corresponds to about 10% of the mean balance sheet size (EUR 5.2 million). Table 2 Descriptive statistics    Unit  N  Mean  Median  SD  Ratings and cutoff                 Rating grade  Number (1$$=$$best, 11$$=$$worst)  16,855  5.78  6.00  2.00  Rating (continuous)  Number (0.5$$=$$best, 11.5$$=$$worst)  16,855  5.80  5.65  1.98  Cutoff  Dummy (0/1)  16,855  0.81  1.00  0.39  Accepted  Dummy (0/1)  16,855  0.72  1.00  0.45  Loan characteristics                 Loan amount  EUR ’000  16,855  526.80  500.00  345.2  Collateralized  Dummy (0/1)  16,855  0.56  1.00  0.50  Other firm characteristics                 Firm age  Years  16,855  20.98  17.00  17.79  Relationship age  Years  16,855  9.05  5.00  10.86  Revenues  EUR mil  16,855  9.70  5.37  13.70  Number of employees  Number  16,855  54.73  30.00  81.59  Total assets  EUR mil  16,855  5.18  2.58  8.46  Asset growth  Number  16,855  0.13  0.10  0.30  Cash and cash equivalents (CCE)/Total assets  Number  16,855  0.12  0.06  0.14  Current assets excl. CCE/Total assets  Number  16,855  0.59  0.62  0.24  Investments/Total assets  Number  16,855  0.26  0.19  0.22  Equity-to-asset ratio  Number  16,855  0.29  0.26  0.22  EBIT margin  Number  16,855  0.06  0.05  0.08  Liquidity  Number  16,855  2.10  1.46  2.04     Unit  N  Mean  Median  SD  Ratings and cutoff                 Rating grade  Number (1$$=$$best, 11$$=$$worst)  16,855  5.78  6.00  2.00  Rating (continuous)  Number (0.5$$=$$best, 11.5$$=$$worst)  16,855  5.80  5.65  1.98  Cutoff  Dummy (0/1)  16,855  0.81  1.00  0.39  Accepted  Dummy (0/1)  16,855  0.72  1.00  0.45  Loan characteristics                 Loan amount  EUR ’000  16,855  526.80  500.00  345.2  Collateralized  Dummy (0/1)  16,855  0.56  1.00  0.50  Other firm characteristics                 Firm age  Years  16,855  20.98  17.00  17.79  Relationship age  Years  16,855  9.05  5.00  10.86  Revenues  EUR mil  16,855  9.70  5.37  13.70  Number of employees  Number  16,855  54.73  30.00  81.59  Total assets  EUR mil  16,855  5.18  2.58  8.46  Asset growth  Number  16,855  0.13  0.10  0.30  Cash and cash equivalents (CCE)/Total assets  Number  16,855  0.12  0.06  0.14  Current assets excl. CCE/Total assets  Number  16,855  0.59  0.62  0.24  Investments/Total assets  Number  16,855  0.26  0.19  0.22  Equity-to-asset ratio  Number  16,855  0.29  0.26  0.22  EBIT margin  Number  16,855  0.06  0.05  0.08  Liquidity  Number  16,855  2.10  1.46  2.04  This table presents summary statistics for the sample of all loan applications between January 2009 and December 2012. For variable definitions, see Table 1. Table 2 Descriptive statistics    Unit  N  Mean  Median  SD  Ratings and cutoff                 Rating grade  Number (1$$=$$best, 11$$=$$worst)  16,855  5.78  6.00  2.00  Rating (continuous)  Number (0.5$$=$$best, 11.5$$=$$worst)  16,855  5.80  5.65  1.98  Cutoff  Dummy (0/1)  16,855  0.81  1.00  0.39  Accepted  Dummy (0/1)  16,855  0.72  1.00  0.45  Loan characteristics                 Loan amount  EUR ’000  16,855  526.80  500.00  345.2  Collateralized  Dummy (0/1)  16,855  0.56  1.00  0.50  Other firm characteristics                 Firm age  Years  16,855  20.98  17.00  17.79  Relationship age  Years  16,855  9.05  5.00  10.86  Revenues  EUR mil  16,855  9.70  5.37  13.70  Number of employees  Number  16,855  54.73  30.00  81.59  Total assets  EUR mil  16,855  5.18  2.58  8.46  Asset growth  Number  16,855  0.13  0.10  0.30  Cash and cash equivalents (CCE)/Total assets  Number  16,855  0.12  0.06  0.14  Current assets excl. CCE/Total assets  Number  16,855  0.59  0.62  0.24  Investments/Total assets  Number  16,855  0.26  0.19  0.22  Equity-to-asset ratio  Number  16,855  0.29  0.26  0.22  EBIT margin  Number  16,855  0.06  0.05  0.08  Liquidity  Number  16,855  2.10  1.46  2.04     Unit  N  Mean  Median  SD  Ratings and cutoff                 Rating grade  Number (1$$=$$best, 11$$=$$worst)  16,855  5.78  6.00  2.00  Rating (continuous)  Number (0.5$$=$$best, 11.5$$=$$worst)  16,855  5.80  5.65  1.98  Cutoff  Dummy (0/1)  16,855  0.81  1.00  0.39  Accepted  Dummy (0/1)  16,855  0.72  1.00  0.45  Loan characteristics                 Loan amount  EUR ’000  16,855  526.80  500.00  345.2  Collateralized  Dummy (0/1)  16,855  0.56  1.00  0.50  Other firm characteristics                 Firm age  Years  16,855  20.98  17.00  17.79  Relationship age  Years  16,855  9.05  5.00  10.86  Revenues  EUR mil  16,855  9.70  5.37  13.70  Number of employees  Number  16,855  54.73  30.00  81.59  Total assets  EUR mil  16,855  5.18  2.58  8.46  Asset growth  Number  16,855  0.13  0.10  0.30  Cash and cash equivalents (CCE)/Total assets  Number  16,855  0.12  0.06  0.14  Current assets excl. CCE/Total assets  Number  16,855  0.59  0.62  0.24  Investments/Total assets  Number  16,855  0.26  0.19  0.22  Equity-to-asset ratio  Number  16,855  0.29  0.26  0.22  EBIT margin  Number  16,855  0.06  0.05  0.08  Liquidity  Number  16,855  2.10  1.46  2.04  This table presents summary statistics for the sample of all loan applications between January 2009 and December 2012. For variable definitions, see Table 1. The bank collects firm characteristics during the application process so that firm characteristics in the year prior to the loan application are available on a more granular level than mandated by the disclosure requirements discussed above. I thus make use of the firm characteristics collected by the bank for the following descriptive statistics.11 The average firm is 21 years old (median: 17 years) and has a relationship with the bank for 9.1 years (median: 5 years). It has EUR 9.7 million in revenues (median: EUR 5.4 million) and 55 employees (median: 30 employees). According to the German Federal Statistical Office, the median revenue of all German firms in 2012 (excluding self-employed workers) was EUR 5.0 million. Thus, the average firm size is largely representative of the average German firm and significantly smaller than samples of listed firms or firms active in the syndicated loan market. Total assets amount to an average of 5.2 million (median: 2.6 million), with an average yearly nominal asset growth of 13% (median: 10%). Cash and cash equivalents account for 12% of total assets, other current assets account for 59% of total assets, and investments account for 26% of total assets. The average equity-to-asset ratio is 29% (median: 26%), the average liquidity ratio (current assets divided by current liabilities) is 2.10 (median: 1.46). Current assets are defined as the sum of inventory, trade receivables, marketable securities, and cash, while current liabilities are equal to the sum of trade payables, debt with a remaining maturity of less than 1 year, and other current liabilities. The average profitability, measured as the EBIT-margin (EBIT divided by revenues), is 6% (median: 5%). 3. Empirical Strategy and Results 3.1 Empirical strategy The lender cutoff rule provides a plausibly exogenous variation in loan supply. Thus, the cutoff rating can be used in a regression discontinuity design (Thistlewaite and Campbell (1960); Lee and Lemieux 2009):   \begin{align} y_{i,t}& = \beta\,{\cdot}\,\textit{BelowCutOff(0/1) }+ g_{1}\textit{(DifferenceToCutOff)}\notag \\ &\quad + g_{2}\textit{(DifferenceToCutOff)}\,{\cdot}\,\textit{BelowCutOff(0/1) } + \gamma\,{\cdot}\,\textit{Controls }+ \varepsilon , \end{align} (1) where $$y_{i}$$ is the variable of interest (e.g., change in loan volume, cash holdings, or investments from the year prior to the year after the loan application) for firm $$i$$ applying for a loan application at time $$t$$, BelowCutOff(0/1) is a dummy equal to one if the rating is below the cutoff rating (i.e., a rating of 7.5 or worse), DifferenceToCutOff is the difference between the continuous internal rating (not the binned rating) and the cutoff rating and g1 and g2 are polynomials fitted to the right and left-hand side of the cutoff rating.$$^{\thinspace }$$12 Regression (1) is a reduced-form model, and the coefficient of interest, $$\beta $$, provides the intent-to-treat effect, that is, the difference between firms below and firms above the cutoff. Since the likelihood of receiving a loan jumps by less than one at the cutoff, we need to use a 2SLS to measure the impact of an exogenous change in loan supply on the outcome variable of interest. The first stage measures the magnitude of the credit supply shock at the cutoff, and I discuss various first stages in Section 3.2. Throughout the paper, I use a local linear regression; that is, the functions $$g_{1}$$ and $$g_{2}$$ are linear functions, and I restrict the sample to a local bandwidth of $$\pm 2$$ notches around the threshold. The bandwidth has been determined using the rule-of-thumb bandwidth selector by Fan and Gijbels (1996). The same bandwidth is chosen consistently across all tables to allow for a meaningful comparison.13 Identification in the RDD comes from a cross-sectional analysis. Controls is a set of loan and firm characteristics and fixed effects. Loan controls are taken from the initial loan application and include the requested loan amount and a collateral dummy, which is equal to one if the initial loan application is for a collateralized loan. Firm characteristics include the logarithm of firm age (in years), the logarithm of 1 plus the length of the lending relationship (number of years that the firm has had an account at the bank without interruption), the logarithm of firm revenues (in EUR million), the logarithm of the number of employees, the equity-to-asset ratio, the EBIT margin (earnings before interest and taxes, depreciation and amortization divided by firm revenues), and the liquidity ratio (current liabilities divided by current assets).14 All firm characteristics are determined as of the fiscal year prior to the date of the loan application. Fixed effects include industry x year fixed effects and one-digit ZIP code x year fixed effects. Equation (1) is estimated using a linear model and all standard errors are clustered at the branch level.15$$^{\mathrm{,}}$$16 The regression discontinuity design relies on the continuity in the conditional expectation function of the outcome variable around the cutoff in absence of treatment (Hahn, Todd, and van der Klaauw 2001). Researchers typically rely on two tests to rationalize this assumption. First, a no-manipulation test to rule out sorting around the threshold. Economically, manipulation is not an issue here, as the rating is purely based on hard information.17 A formal McCrary density test (McCrary 2008) does not reject the no-manipulation assumption (see Figure 1; Table A3). A second common test is to verify that there is no discontinuity in any of the control variables around the cutoff. Panel A of Figure A1 shows levels of control variables at the end of the fiscal year prior to the loan application as a function of the running variable, and panel B shows pre-application changes for assets, investment, employment, and revenues.18Table A4 provides the corresponding econometric tests. There is no evidence of a consistent jump in either pre-application levels or pre-application growth rates at the cutoff. 3.2 The impact of the lender cutoff rule on firms’ financing 3.2.1 Loan acceptance rates In the first step, I estimate Equation (1) using the acceptance dummy as the dependent variable. The acceptance dummy is equal to 1 if the bank accepts a loan application. The test thus fulfills a simple purpose, that is, to confirm that the cutoff rule described in Section 1.1 is indeed reflected in the data. Column 1 of Table 3 presents the results. Table 3 The impact of the lender cutoff rule on firms’ financing    Loan acceptance  Change in loan volume with the bank  Change in total debt (all banks and nonbanks)  Change in equity     (1)  (2)  (3)  (4)  (5)  (6)  Dependent variable Model  Acceptance dummy (0/1) Linear  Time horizon: 3 months Linear  Time horizon: 12 months Linear  Time horizon: 24 months Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Methodology  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Parameter  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Inference                                      BelowCutOff (0/1)  –0.280***  (–12.27)  –0.062***  (–6.93)  –0.060***  (–5.55)  –0.070***  (–4.70)  –0.071***  (–3.29)  –0.015**  (–2.50)  Trends below/Above cutoff                                      (Rating-CutOff) x  –0.003  (–0.36)  0.028***  (4.33)  0.008  (1.27)  0.000  (0.00)  –0.006  (–0.37)  –0.007*  (–1.84)      BelowCutOff (0/1)                                      (Rating-CutOff) x  –0.028*  (–1.79)  –0.018**  (–2.80)  –0.020**  (–2.24)  –0.026**  (–2.76)  0.019  (1.29)  –0.009*  (–1.83)  (1- BelowCutOff (0/1))                                      Firm controls  Yes  Yes  Yes  Yes  Yes  Yes  Loan controls  Yes  Yes  Yes  Yes  Yes  Yes  Industry x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Region x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Diagnostics                    Adj. R$$^{\mathrm{2}}$$(%)  19.99  13.41  10.06  13.15  11.97  5.61  N  8,807  8,807  8,807  8,807  8,807  8,807     Loan acceptance  Change in loan volume with the bank  Change in total debt (all banks and nonbanks)  Change in equity     (1)  (2)  (3)  (4)  (5)  (6)  Dependent variable Model  Acceptance dummy (0/1) Linear  Time horizon: 3 months Linear  Time horizon: 12 months Linear  Time horizon: 24 months Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Methodology  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Parameter  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Inference                                      BelowCutOff (0/1)  –0.280***  (–12.27)  –0.062***  (–6.93)  –0.060***  (–5.55)  –0.070***  (–4.70)  –0.071***  (–3.29)  –0.015**  (–2.50)  Trends below/Above cutoff                                      (Rating-CutOff) x  –0.003  (–0.36)  0.028***  (4.33)  0.008  (1.27)  0.000  (0.00)  –0.006  (–0.37)  –0.007*  (–1.84)      BelowCutOff (0/1)                                      (Rating-CutOff) x  –0.028*  (–1.79)  –0.018**  (–2.80)  –0.020**  (–2.24)  –0.026**  (–2.76)  0.019  (1.29)  –0.009*  (–1.83)  (1- BelowCutOff (0/1))                                      Firm controls  Yes  Yes  Yes  Yes  Yes  Yes  Loan controls  Yes  Yes  Yes  Yes  Yes  Yes  Industry x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Region x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Diagnostics                    Adj. R$$^{\mathrm{2}}$$(%)  19.99  13.41  10.06  13.15  11.97  5.61  N  8,807  8,807  8,807  8,807  8,807  8,807  This table estimates the effect of the lender cutoff rule on credit supply using a regression discontinuity design. Column 1 uses the acceptance dummy as the dependent variable to test whether the lender cutoff rule is confirmed in the data. The acceptance dummy is equal to 1 if the bank makes a loan offer to the firm and equals 0 if the bank does not make a loan offer to the firm. Columns 2–4 provide results using the subsequent change in loan volume with the bank as the dependent variable. The subsequent change in the loan volume is measured as the logarithm of the ratio of the loan volume of the firm at the bank 3, 12, and 24 months after the loan application date divided by the loan volume of the firm at the bank 1 month prior to the loan application. Column 5 uses the change in debt (Column 6: change in equity) as reported in the annual reports from the fiscal year prior to the loan application date to the fiscal year after the loan application date. All models are estimated using a linear model. For variable definitions, see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Table 3 The impact of the lender cutoff rule on firms’ financing    Loan acceptance  Change in loan volume with the bank  Change in total debt (all banks and nonbanks)  Change in equity     (1)  (2)  (3)  (4)  (5)  (6)  Dependent variable Model  Acceptance dummy (0/1) Linear  Time horizon: 3 months Linear  Time horizon: 12 months Linear  Time horizon: 24 months Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Methodology  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Parameter  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Inference                                      BelowCutOff (0/1)  –0.280***  (–12.27)  –0.062***  (–6.93)  –0.060***  (–5.55)  –0.070***  (–4.70)  –0.071***  (–3.29)  –0.015**  (–2.50)  Trends below/Above cutoff                                      (Rating-CutOff) x  –0.003  (–0.36)  0.028***  (4.33)  0.008  (1.27)  0.000  (0.00)  –0.006  (–0.37)  –0.007*  (–1.84)      BelowCutOff (0/1)                                      (Rating-CutOff) x  –0.028*  (–1.79)  –0.018**  (–2.80)  –0.020**  (–2.24)  –0.026**  (–2.76)  0.019  (1.29)  –0.009*  (–1.83)  (1- BelowCutOff (0/1))                                      Firm controls  Yes  Yes  Yes  Yes  Yes  Yes  Loan controls  Yes  Yes  Yes  Yes  Yes  Yes  Industry x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Region x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Diagnostics                    Adj. R$$^{\mathrm{2}}$$(%)  19.99  13.41  10.06  13.15  11.97  5.61  N  8,807  8,807  8,807  8,807  8,807  8,807     Loan acceptance  Change in loan volume with the bank  Change in total debt (all banks and nonbanks)  Change in equity     (1)  (2)  (3)  (4)  (5)  (6)  Dependent variable Model  Acceptance dummy (0/1) Linear  Time horizon: 3 months Linear  Time horizon: 12 months Linear  Time horizon: 24 months Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Methodology  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Parameter  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Inference                                      BelowCutOff (0/1)  –0.280***  (–12.27)  –0.062***  (–6.93)  –0.060***  (–5.55)  –0.070***  (–4.70)  –0.071***  (–3.29)  –0.015**  (–2.50)  Trends below/Above cutoff                                      (Rating-CutOff) x  –0.003  (–0.36)  0.028***  (4.33)  0.008  (1.27)  0.000  (0.00)  –0.006  (–0.37)  –0.007*  (–1.84)      BelowCutOff (0/1)                                      (Rating-CutOff) x  –0.028*  (–1.79)  –0.018**  (–2.80)  –0.020**  (–2.24)  –0.026**  (–2.76)  0.019  (1.29)  –0.009*  (–1.83)  (1- BelowCutOff (0/1))                                      Firm controls  Yes  Yes  Yes  Yes  Yes  Yes  Loan controls  Yes  Yes  Yes  Yes  Yes  Yes  Industry x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Region x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Diagnostics                    Adj. R$$^{\mathrm{2}}$$(%)  19.99  13.41  10.06  13.15  11.97  5.61  N  8,807  8,807  8,807  8,807  8,807  8,807  This table estimates the effect of the lender cutoff rule on credit supply using a regression discontinuity design. Column 1 uses the acceptance dummy as the dependent variable to test whether the lender cutoff rule is confirmed in the data. The acceptance dummy is equal to 1 if the bank makes a loan offer to the firm and equals 0 if the bank does not make a loan offer to the firm. Columns 2–4 provide results using the subsequent change in loan volume with the bank as the dependent variable. The subsequent change in the loan volume is measured as the logarithm of the ratio of the loan volume of the firm at the bank 3, 12, and 24 months after the loan application date divided by the loan volume of the firm at the bank 1 month prior to the loan application. Column 5 uses the change in debt (Column 6: change in equity) as reported in the annual reports from the fiscal year prior to the loan application date to the fiscal year after the loan application date. All models are estimated using a linear model. For variable definitions, see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Reassuringly, the cutoff rule is indeed borne out in the data: the coefficient on the cutoff dummy is equal to $$-0.28$$ ($$t$$-stat $$< -10$$), suggesting that the likelihood of an accept-decision drops by 28 percentage points at the cutoff rating. The following columns analyze how this drop in loan acceptance rates feeds through the firms’ financing structure (loan volume with the bank, total debt, and equity). 3.2.2 Loan volume with the bank How does the cutoff rating affect a firm’s loan volume at the bank? Column 2 of Table 3 looks at the change in loan volume from 1 month prior to 3 months after the loan application. The loan volume constitutes the total loan volume with the sample bank, that is, including loans granted prior to the sample period that are still outstanding at the time of interest (here: 3 months after the loan application) and including loans larger than EUR 1 million. Here and in the following, the change is measured relative to the firm’s total assets in the fiscal year prior to the loan application. Therefore, the results directly shed light on the economic importance, that is, on the loss in funding relative to the size of the firm’s balance sheet. The coefficient on the cutoff dummy is $$-0.062$$ and highly statistically significant. The coefficient is also economically significant: firms below the cutoff end up with a lower amount of funding from the sample bank equal to 6.2% of their total balance sheet size. This effect stems from both loan rejections (see Column 1 of Table 3) and the fact that risk management might accept a loan application, but only with a loan amount which is lower than that demanded by the firm, that is, from both the intensive and the extensive margins. With an average balance sheet size of EUR 5.2 million, a credit supply shock equal to 6.2% of total assets amounts to approximately EUR 320,000. The prior analyses looked at a rather short time window, that is, 1 month prior to 3 months after the loan application. It is important to analyze whether the same results carry over to longer time horizons, for example, 1 or 2 years after loan application. It is conceivable that a firm just below the cutoff rating migrates to a rating above the cutoff rating after a while; and is thus able to successfully reapply for a loan. For the identification of real effects, which are measured using annual report data, it is important that the discontinuity in the loan supply is nontransient. I thus repeat the regression using the change in loan volume from 1 month prior to 12 months (Column 3 of Table 3) and 24 months (Column 4 of Table 3) after the loan application. Results are very similar to Column 2, with the coefficient on the cutoff dummy ranging from $$-6.0{\%}$$ to $$-7.0{\%}$$ (significant at the 1% level in all specifications). I conclude that being below the cutoff rating at the time of a loan application has indeed a longer-lasting effect on loan supply from the sample bank.19 3.2.3 Substitution effect 1: Total debt The loan volume analyzed in Columns 2–4 of Table 3 only constitute the loan volume with the sample bank; that is, the results are uninformative as to whether the firm is able to substitute any funding shortfall by applying for a loan at another bank. Are firms able to substitute funding from the sample bank via other funding sources such as loans from other banks or equity capital? Column 5 of Table 3 sheds light on this question. The dependent variable is the change in total debt from the fiscal year prior to loan application to the fiscal year in the year following the loan application (i.e., the change is measured over 2 years). Total debt includes bank debt and trade payables; that is, the results shed light on substitution effects via other banks and via trade credit from suppliers. Again, the change is measured relative to the firm’s balance sheet size in the fiscal year prior to the loan application. The coefficient on the cutoff dummy is $$-7.1{\%}$$, suggesting that firms are not fully able to substitute from other banks or via trade credit. 3.2.4 Substitution effect 2: Equity As an alternative to debt funding, firms might choose to increase equity capital. Column 6 of Table 3 shows that this is not the case. Using the change in equity capital as the dependent variable gives a negative coefficient. If at all, firms below the cutoff decrease equity capital, but they certainly do not substitute the loss in debt funding by an increase in equity funding. Please note that the disclosure requirements are too coarse to allow distinguishing whether changes in equity capital are a result of lower earnings, lower retention rates, or lower external equity financing. Figure A2 plots the coefficients from Table 3 as a function of the bandwidth. Results from Figure A2 suggest that those effects that are highly significant in the $$\pm 2$$ notch specification are stable until approximately 0.5 notches bandwidth. Table A6 confirms the results from Table 3 in a robustness test using different econometric approaches (triangular kernel, higher-order polynomials, and a more granular industry definition: 99 industries instead of 14, controlling for past growth, no control variables, and different bandwidths). Taken together, these results imply that loan rejections have a nontransient effect on loan volumes with the bank. The loss in funding from the bank is not substituted by other funding sources (loans from other banks, trade credit, equity financing). 3.2.5 Results by size class Table 4 reproduces the regressions from Table 3 by size class. Firms are split into four quartiles by total assets in the year prior to the loan application. Table 4 only reports the key coefficient of interest, that is, the coefficient for the BelowCutOff dummy. Consistent with prior literature (Gertler and Gilchrist 1994), the effects are more pronounced for smaller firms. While acceptance rates drop significantly at the cutoff rating for all size classes, the change in loan volume with the bank and the change in total debt are only consistently significant for the first two quartiles. Table 4 The impact of the lender cutoff rule on firms’ financing: Split by size classes    Loan acceptance  Change in loan volume with the bank  Change in total debt (all banks and nonbanks)  Change in equity     (1)  (2)  (3)  (4)  (5)  (6)  Dependent variable Model  Acceptance dummy (0/1) Linear  Time horizon: 3 months Linear  Time horizon: 12 months Linear  Time horizon: 24 months Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Methodology  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Q1: Total assets $$\leqslant $$ EUR 1.5 mil BelowCutOff (0/1)  –0.325***  (–9.65)  –0.130***  (–5.38)  –0.131***  (–5.16)  –0.151***  (–4.04)  –0.118**  (–2.18)  –0.036**  (–2.02)  Controls and fixed effects used in Table 3  Yes  Yes  Yes  Yes  Yes  Yes  Q2: EUR 1.5mn < Total assets $$\leqslant $$ EUR 3 mil BelowCutOff (0/1)  –0.238***  (–5.83)  –0.048***  (–3.27)  –0.057***  (–2.79)  –0.060**  (–2.44)  –0.053*  (–1.96)  –0.008  (–0.57)  Controls and fixed effects used Table 3  Yes  Yes  Yes  Yes  Yes  Yes  Q3: EUR 3mn < Total assets $$\leqslant $$ EUR 5 mil BelowCutOff (0/1)  –0.297***  (–6.51)  –0.038***  (–3.51)  –0.017  (–1.37)  –0.001  (–0.05)  –0.029  (–1.03)  –0.011  (–0.97)  Controls and fixed effects used in Table 3  Yes  Yes  Yes  Yes  Yes  Yes  Q4: Total assets > EUR 5 mil BelowCutOff (0/1)  –0.229***  (–4.98)  –0.003  (–0.54)  0.009  (1.27)  0.006  (0.06)  –0.030  (–1.00)  0.004  (–.25)  Controls and fixed effects used in Table 3  Yes  Yes  Yes  Yes  Yes        Test for difference in coefficients (Q1-Q4)  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  Difference in coefficients  –0.096*  (3.33)  –0.127***  (28.20)  –0.140***  (28.94)  –0.157***  (16.13)  –0.088**  (3.29)  –0.040  (2.66)     Loan acceptance  Change in loan volume with the bank  Change in total debt (all banks and nonbanks)  Change in equity     (1)  (2)  (3)  (4)  (5)  (6)  Dependent variable Model  Acceptance dummy (0/1) Linear  Time horizon: 3 months Linear  Time horizon: 12 months Linear  Time horizon: 24 months Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Methodology  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Q1: Total assets $$\leqslant $$ EUR 1.5 mil BelowCutOff (0/1)  –0.325***  (–9.65)  –0.130***  (–5.38)  –0.131***  (–5.16)  –0.151***  (–4.04)  –0.118**  (–2.18)  –0.036**  (–2.02)  Controls and fixed effects used in Table 3  Yes  Yes  Yes  Yes  Yes  Yes  Q2: EUR 1.5mn < Total assets $$\leqslant $$ EUR 3 mil BelowCutOff (0/1)  –0.238***  (–5.83)  –0.048***  (–3.27)  –0.057***  (–2.79)  –0.060**  (–2.44)  –0.053*  (–1.96)  –0.008  (–0.57)  Controls and fixed effects used Table 3  Yes  Yes  Yes  Yes  Yes  Yes  Q3: EUR 3mn < Total assets $$\leqslant $$ EUR 5 mil BelowCutOff (0/1)  –0.297***  (–6.51)  –0.038***  (–3.51)  –0.017  (–1.37)  –0.001  (–0.05)  –0.029  (–1.03)  –0.011  (–0.97)  Controls and fixed effects used in Table 3  Yes  Yes  Yes  Yes  Yes  Yes  Q4: Total assets > EUR 5 mil BelowCutOff (0/1)  –0.229***  (–4.98)  –0.003  (–0.54)  0.009  (1.27)  0.006  (0.06)  –0.030  (–1.00)  0.004  (–.25)  Controls and fixed effects used in Table 3  Yes  Yes  Yes  Yes  Yes        Test for difference in coefficients (Q1-Q4)  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  Difference in coefficients  –0.096*  (3.33)  –0.127***  (28.20)  –0.140***  (28.94)  –0.157***  (16.13)  –0.088**  (3.29)  –0.040  (2.66)  This table estimates the effect of the lender cutoff rule on credit supply using a regression discontinuity design. Results are split by quartile of total assets in the fiscal year prior to the loan application date. Columns and models are the same as those used in Table 3, but only the coefficient on the BelowCutOff dummy is reported. For variable definitions, see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Table 4 The impact of the lender cutoff rule on firms’ financing: Split by size classes    Loan acceptance  Change in loan volume with the bank  Change in total debt (all banks and nonbanks)  Change in equity     (1)  (2)  (3)  (4)  (5)  (6)  Dependent variable Model  Acceptance dummy (0/1) Linear  Time horizon: 3 months Linear  Time horizon: 12 months Linear  Time horizon: 24 months Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Methodology  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Q1: Total assets $$\leqslant $$ EUR 1.5 mil BelowCutOff (0/1)  –0.325***  (–9.65)  –0.130***  (–5.38)  –0.131***  (–5.16)  –0.151***  (–4.04)  –0.118**  (–2.18)  –0.036**  (–2.02)  Controls and fixed effects used in Table 3  Yes  Yes  Yes  Yes  Yes  Yes  Q2: EUR 1.5mn < Total assets $$\leqslant $$ EUR 3 mil BelowCutOff (0/1)  –0.238***  (–5.83)  –0.048***  (–3.27)  –0.057***  (–2.79)  –0.060**  (–2.44)  –0.053*  (–1.96)  –0.008  (–0.57)  Controls and fixed effects used Table 3  Yes  Yes  Yes  Yes  Yes  Yes  Q3: EUR 3mn < Total assets $$\leqslant $$ EUR 5 mil BelowCutOff (0/1)  –0.297***  (–6.51)  –0.038***  (–3.51)  –0.017  (–1.37)  –0.001  (–0.05)  –0.029  (–1.03)  –0.011  (–0.97)  Controls and fixed effects used in Table 3  Yes  Yes  Yes  Yes  Yes  Yes  Q4: Total assets > EUR 5 mil BelowCutOff (0/1)  –0.229***  (–4.98)  –0.003  (–0.54)  0.009  (1.27)  0.006  (0.06)  –0.030  (–1.00)  0.004  (–.25)  Controls and fixed effects used in Table 3  Yes  Yes  Yes  Yes  Yes        Test for difference in coefficients (Q1-Q4)  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  Difference in coefficients  –0.096*  (3.33)  –0.127***  (28.20)  –0.140***  (28.94)  –0.157***  (16.13)  –0.088**  (3.29)  –0.040  (2.66)     Loan acceptance  Change in loan volume with the bank  Change in total debt (all banks and nonbanks)  Change in equity     (1)  (2)  (3)  (4)  (5)  (6)  Dependent variable Model  Acceptance dummy (0/1) Linear  Time horizon: 3 months Linear  Time horizon: 12 months Linear  Time horizon: 24 months Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Fiscal year prior to loan application to fiscal year after loan application Linear  Methodology  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Q1: Total assets $$\leqslant $$ EUR 1.5 mil BelowCutOff (0/1)  –0.325***  (–9.65)  –0.130***  (–5.38)  –0.131***  (–5.16)  –0.151***  (–4.04)  –0.118**  (–2.18)  –0.036**  (–2.02)  Controls and fixed effects used in Table 3  Yes  Yes  Yes  Yes  Yes  Yes  Q2: EUR 1.5mn < Total assets $$\leqslant $$ EUR 3 mil BelowCutOff (0/1)  –0.238***  (–5.83)  –0.048***  (–3.27)  –0.057***  (–2.79)  –0.060**  (–2.44)  –0.053*  (–1.96)  –0.008  (–0.57)  Controls and fixed effects used Table 3  Yes  Yes  Yes  Yes  Yes  Yes  Q3: EUR 3mn < Total assets $$\leqslant $$ EUR 5 mil BelowCutOff (0/1)  –0.297***  (–6.51)  –0.038***  (–3.51)  –0.017  (–1.37)  –0.001  (–0.05)  –0.029  (–1.03)  –0.011  (–0.97)  Controls and fixed effects used in Table 3  Yes  Yes  Yes  Yes  Yes  Yes  Q4: Total assets > EUR 5 mil BelowCutOff (0/1)  –0.229***  (–4.98)  –0.003  (–0.54)  0.009  (1.27)  0.006  (0.06)  –0.030  (–1.00)  0.004  (–.25)  Controls and fixed effects used in Table 3  Yes  Yes  Yes  Yes  Yes        Test for difference in coefficients (Q1-Q4)  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  $$\Delta $$ coeff.  $${\rm X}^{\mathrm{2}}$$  Difference in coefficients  –0.096*  (3.33)  –0.127***  (28.20)  –0.140***  (28.94)  –0.157***  (16.13)  –0.088**  (3.29)  –0.040  (2.66)  This table estimates the effect of the lender cutoff rule on credit supply using a regression discontinuity design. Results are split by quartile of total assets in the fiscal year prior to the loan application date. Columns and models are the same as those used in Table 3, but only the coefficient on the BelowCutOff dummy is reported. For variable definitions, see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. These results suggest that larger firms are able to cushion the effects of loan rejections, either because they have a more granular financing structure so that a single loan application constitutes a smaller amount of their total financing volume, or by reapplying for a loan at another bank. The existence of a credit supply shock is a necessary requirement for the following analysis, and I will thus focus on firms in the smallest two quartiles (firms with total assets $$\leqslant $$ EUR 3 million) in the following subsections. The firms in this size category are comparable to firms that are labeled “micro and small firms” by the European statistical agency (Eurostat). Micro and small firms according to the Eurostat category employ 42% of the workforce in Germany and 53% of the workforce in Europe. In the United States, the proportion is somehow lower (26% of the workforce employed in micro and small firms as defined by Eurostat). Taken together, this evidence suggests that the sample of firms in the smallest two quartiles by total assets is largely representative of firms employing about a quarter to half of the total workforce in Germany, Europe, and the United States. Given the limitations on data availability for smaller firms, these firms are usually underrepresented in academic studies.20 3.3 The impact of the lender cutoff rule on firms’ cash holdings The prior analysis has demonstrated that the lender cutoff rule restricts firms’ overall availability of funding, in particular for small firms. How does the loss in funding transmit to the asset side, that is, which assets are reduced as a response to funding shock? As a first item, I look at the impact on cash holdings. The theory on precautionary savings postulates that firms hold cash as a buffer against adverse cash flow shocks. One possible prediction of this theory is thus that firms use their cash holdings to cushion the credit supply shock induced by the lender cutoff rule. However, firms might as well increase their cash holdings as a result of the loan rejection: the loan rejection is likely to impact firms’ belief about the future availability of financing. The credit supply shock might therefore increase precautionary savings motives, as the value of cash is higher for credit-constrained firms than for unconstrained firms. I test these hypotheses in panel A of Table 5. Again, the regressions follow the regression discontinuity design as formulated in Equation (1). The dependent variable is the sum of cash and marketable securities, which I label cash and cash equivalents following common practice in the literature. Column 1 in Table 5 reports results for the total sample of small firms. Effects of loan rejections on cash holdings are insignificant. The results are, however, strikingly different when splitting the sample by the current ratio (current assets divided by current liabilities) of the firms in the year prior to the loan application. Column 2 reports results for firms with a low current ratio prior to the year of the loan application. Firms below the cutoff increase cash holdings by 2.5% of their total assets relative to firms above the cutoff. In contrast, firms below the cutoff with a high current ratio decrease their cash holdings by 3.2% of their total assets relative to firms above the cutoff. The difference between these two coefficients is highly significant at the 1% level. Table 5 The impact of the lender cutoff rule on firms’ cash holdings Dependent variable  A. Change in cash and cash equivalents  B. Change in current assets (excluding cash and cash equivalents)     (1)  (2)  (3)  (4)  (5)  (6)  Sample  All firms  Low liquidity (CA/CL $$\leqslant $$ 1.4)  High liquidity (CA/CL > 1.4)  All firms  Low liquidity (CA/CL $$\leqslant $$ 1.4)  High liquidity (CA/CL > 1.4)  Methodology  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Parameter  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  t-stat  Inference                                      BelowCutOff (0/1)  –0.002  (–0.24)  0.025**  (2.23)  –0.032**  (–2.48)  –0.048*  (–1.89)  –0.034  (–1.07)  –0.041  (–1.03)  Trends below/Above cutoff                                      (Rating-CutOff) x BelowCutOff (0/1)  0.005  (0.80)  –0.005  (–0.63)  0.014  (1.51)  0.006  (0.38)  0.007  (0.41)  –0.006  (–0.26)  (Rating-CutOff) x (1- BelowCutOff (0/1))  –0.001  (–0.07)  –0.006  (–1.00)  0.008  (0.67)  0.013  (0.69)  0.011  (0.53)  0.036  (1.00)  Firm controls  Yes  Yes  Yes  Yes  Yes  Yes  Loan controls  Yes  Yes  Yes  Yes  Yes  Yes  Industry x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Region x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Diagnostics                                      Adj. R$$^{\mathrm{2}}$$  4.28%  3.29%  7.48%  15.84%  12.37%  19.83%  N  4,714  2,279  2,435  4,714  2,279  2,435  Difference between BelowCutoOff(01/1)-        $$\Delta $$ coeff.  X$$^{\mathrm{2}}$$           $$\Delta $$ coeff.  X$$^{\mathrm{2}}$$     Low minus high liquidity        $$0.057^{***}$$  (11.38)           0.007  (0.02)     IV estimates                    First stage: Acceptance dummy  –0.286***  (–9.83)  –0.335***  (–9.32)  –0.225***  (–4.55)  –0.286***  (–9.83)  –0.335***  (–9.32)  –0.225***  (–4.55)  IV: using Acceptance dummy  0.007  (0.19)  –0.074**  (–2.14)  0.141*  (1.75)  0.168*  (1.75)  0.101  (1.19)  0.182  (0.80)  First stage: Change in loan volume  –0.074***  (–4.26)  –0.069***  (–4.21)  –0.079***  (–2.81)  –0.074***  (–4.26)  –0.069***  (–4.21)  –0.079***  (–2.81)  IV: Using change in loan volume  0.028  (0.24)  –0.364**  (–2.04)  0.401**  (2.11)  0.647**  (2.44)  0.497  (1.16)  0.515  (1.23)  Dependent variable  A. Change in cash and cash equivalents  B. Change in current assets (excluding cash and cash equivalents)     (1)  (2)  (3)  (4)  (5)  (6)  Sample  All firms  Low liquidity (CA/CL $$\leqslant $$ 1.4)  High liquidity (CA/CL > 1.4)  All firms  Low liquidity (CA/CL $$\leqslant $$ 1.4)  High liquidity (CA/CL > 1.4)  Methodology  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Parameter  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  t-stat  Inference                                      BelowCutOff (0/1)  –0.002  (–0.24)  0.025**  (2.23)  –0.032**  (–2.48)  –0.048*  (–1.89)  –0.034  (–1.07)  –0.041  (–1.03)  Trends below/Above cutoff                                      (Rating-CutOff) x BelowCutOff (0/1)  0.005  (0.80)  –0.005  (–0.63)  0.014  (1.51)  0.006  (0.38)  0.007  (0.41)  –0.006  (–0.26)  (Rating-CutOff) x (1- BelowCutOff (0/1))  –0.001  (–0.07)  –0.006  (–1.00)  0.008  (0.67)  0.013  (0.69)  0.011  (0.53)  0.036  (1.00)  Firm controls  Yes  Yes  Yes  Yes  Yes  Yes  Loan controls  Yes  Yes  Yes  Yes  Yes  Yes  Industry x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Region x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Diagnostics                                      Adj. R$$^{\mathrm{2}}$$  4.28%  3.29%  7.48%  15.84%  12.37%  19.83%  N  4,714  2,279  2,435  4,714  2,279  2,435  Difference between BelowCutoOff(01/1)-        $$\Delta $$ coeff.  X$$^{\mathrm{2}}$$           $$\Delta $$ coeff.  X$$^{\mathrm{2}}$$     Low minus high liquidity        $$0.057^{***}$$  (11.38)           0.007  (0.02)     IV estimates                    First stage: Acceptance dummy  –0.286***  (–9.83)  –0.335***  (–9.32)  –0.225***  (–4.55)  –0.286***  (–9.83)  –0.335***  (–9.32)  –0.225***  (–4.55)  IV: using Acceptance dummy  0.007  (0.19)  –0.074**  (–2.14)  0.141*  (1.75)  0.168*  (1.75)  0.101  (1.19)  0.182  (0.80)  First stage: Change in loan volume  –0.074***  (–4.26)  –0.069***  (–4.21)  –0.079***  (–2.81)  –0.074***  (–4.26)  –0.069***  (–4.21)  –0.079***  (–2.81)  IV: Using change in loan volume  0.028  (0.24)  –0.364**  (–2.04)  0.401**  (2.11)  0.647**  (2.44)  0.497  (1.16)  0.515  (1.23)  This table estimates the effect of the lender cutoff rule on cash holdings. Columns 1–3 provide results using cash and cash equivalents as the dependent variable. Column (1) reports results for all small firms, Column (2) reports results for low-liquidity firms (current-asset-to-current-liability ratio in the fiscal year prior to the loan application $$\leqslant $$ 1.4), and column (3) reports results for high-liquidity firms (current-asset-to-current-liability ratio in the fiscal year prior to the loan application > 1.4). Columns 4–6 provide results using current assets excluding cash and cash equivalents as the dependent variable. Column 4 reports results for all small firms; Column 5 reports results for low-liquidity firms (current-asset-to-current-liability ratio in the fiscal year prior to the loan application $$\leqslant $$ 1.4); and Column 6 reports results for high-liquidity firms (current-asset-to-current-liability ratio in the fiscal year prior to the loan application > 1.4). The upper part of the table provides reduced form estimates, the lower part provides IV estimates. All models are estimated using a linear model. For variable definitions, see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Table 5 The impact of the lender cutoff rule on firms’ cash holdings Dependent variable  A. Change in cash and cash equivalents  B. Change in current assets (excluding cash and cash equivalents)     (1)  (2)  (3)  (4)  (5)  (6)  Sample  All firms  Low liquidity (CA/CL $$\leqslant $$ 1.4)  High liquidity (CA/CL > 1.4)  All firms  Low liquidity (CA/CL $$\leqslant $$ 1.4)  High liquidity (CA/CL > 1.4)  Methodology  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Parameter  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  t-stat  Inference                                      BelowCutOff (0/1)  –0.002  (–0.24)  0.025**  (2.23)  –0.032**  (–2.48)  –0.048*  (–1.89)  –0.034  (–1.07)  –0.041  (–1.03)  Trends below/Above cutoff                                      (Rating-CutOff) x BelowCutOff (0/1)  0.005  (0.80)  –0.005  (–0.63)  0.014  (1.51)  0.006  (0.38)  0.007  (0.41)  –0.006  (–0.26)  (Rating-CutOff) x (1- BelowCutOff (0/1))  –0.001  (–0.07)  –0.006  (–1.00)  0.008  (0.67)  0.013  (0.69)  0.011  (0.53)  0.036  (1.00)  Firm controls  Yes  Yes  Yes  Yes  Yes  Yes  Loan controls  Yes  Yes  Yes  Yes  Yes  Yes  Industry x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Region x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Diagnostics                                      Adj. R$$^{\mathrm{2}}$$  4.28%  3.29%  7.48%  15.84%  12.37%  19.83%  N  4,714  2,279  2,435  4,714  2,279  2,435  Difference between BelowCutoOff(01/1)-        $$\Delta $$ coeff.  X$$^{\mathrm{2}}$$           $$\Delta $$ coeff.  X$$^{\mathrm{2}}$$     Low minus high liquidity        $$0.057^{***}$$  (11.38)           0.007  (0.02)     IV estimates                    First stage: Acceptance dummy  –0.286***  (–9.83)  –0.335***  (–9.32)  –0.225***  (–4.55)  –0.286***  (–9.83)  –0.335***  (–9.32)  –0.225***  (–4.55)  IV: using Acceptance dummy  0.007  (0.19)  –0.074**  (–2.14)  0.141*  (1.75)  0.168*  (1.75)  0.101  (1.19)  0.182  (0.80)  First stage: Change in loan volume  –0.074***  (–4.26)  –0.069***  (–4.21)  –0.079***  (–2.81)  –0.074***  (–4.26)  –0.069***  (–4.21)  –0.079***  (–2.81)  IV: Using change in loan volume  0.028  (0.24)  –0.364**  (–2.04)  0.401**  (2.11)  0.647**  (2.44)  0.497  (1.16)  0.515  (1.23)  Dependent variable  A. Change in cash and cash equivalents  B. Change in current assets (excluding cash and cash equivalents)     (1)  (2)  (3)  (4)  (5)  (6)  Sample  All firms  Low liquidity (CA/CL $$\leqslant $$ 1.4)  High liquidity (CA/CL > 1.4)  All firms  Low liquidity (CA/CL $$\leqslant $$ 1.4)  High liquidity (CA/CL > 1.4)  Methodology  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Local regression $$\pm 2$$ notches around cutoff  Parameter  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  $$t$$-stat  Coeff.  t-stat  Inference                                      BelowCutOff (0/1)  –0.002  (–0.24)  0.025**  (2.23)  –0.032**  (–2.48)  –0.048*  (–1.89)  –0.034  (–1.07)  –0.041  (–1.03)  Trends below/Above cutoff                                      (Rating-CutOff) x BelowCutOff (0/1)  0.005  (0.80)  –0.005  (–0.63)  0.014  (1.51)  0.006  (0.38)  0.007  (0.41)  –0.006  (–0.26)  (Rating-CutOff) x (1- BelowCutOff (0/1))  –0.001  (–0.07)  –0.006  (–1.00)  0.008  (0.67)  0.013  (0.69)  0.011  (0.53)  0.036  (1.00)  Firm controls  Yes  Yes  Yes  Yes  Yes  Yes  Loan controls  Yes  Yes  Yes  Yes  Yes  Yes  Industry x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Region x Time fixed effects  Yes  Yes  Yes  Yes  Yes  Yes  Diagnostics                                      Adj. R$$^{\mathrm{2}}$$  4.28%  3.29%  7.48%  15.84%  12.37%  19.83%  N  4,714  2,279  2,435  4,714  2,279  2,435  Difference between BelowCutoOff(01/1)-        $$\Delta $$ coeff.  X$$^{\mathrm{2}}$$           $$\Delta $$ coeff.  X$$^{\mathrm{2}}$$     Low minus high liquidity        $$0.057^{***}$$  (11.38)           0.007  (0.02)     IV estimates                    First stage: Acceptance dummy  –0.286***  (–9.83)  –0.335***  (–9.32)  –0.225***  (–4.55)  –0.286***  (–9.83)  –0.335***  (–9.32)  –0.225***  (–4.55)  IV: using Acceptance dummy  0.007  (0.19)  –0.074**  (–2.14)  0.141*  (1.75)  0.168*  (1.75)  0.101  (1.19)  0.182  (0.80)  First stage: Change in loan volume  –0.074***  (–4.26)  –0.069***  (–4.21)  –0.079***  (–2.81)  –0.074***  (–4.26)  –0.069***  (–4.21)  –0.079***  (–2.81)  IV: Using change in loan volume  0.028  (0.24)  –0.364**  (–2.04)  0.401**  (2.11)  0.647**  (2.44)  0.497  (1.16)  0.515  (1.23)  This table estimates the effect of the lender cutoff rule on cash holdings. Columns 1–3 provide results using cash and cash equivalents as the dependent variable. Column (1) reports results for all small firms, Column (2) reports results for low-liquidity firms (current-asset-to-current-liability ratio in the fiscal year prior to the loan application $$\leqslant $$ 1.4), and column (3) reports results for high-liquidity firms (current-asset-to-current-liability ratio in the fiscal year prior to the loan application > 1.4). Columns 4–6 provide results using current assets excluding cash and cash equivalents as the dependent variable. Column 4 reports results for all small firms; Column 5 reports results for low-liquidity firms (current-asset-to-current-liability ratio in the fiscal year prior to the loan application $$\leqslant $$ 1.4); and Column 6 reports results for high-liquidity firms (current-asset-to-current-liability ratio in the fiscal year prior to the loan application > 1.4). The upper part of the table provides reduced form estimates, the lower part provides IV estimates. All models are estimated using a linear model. For variable definitions, see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. In the bottom part of Table 5, I report fuzzy RDD results using either the acceptance dummy or the change in loan volume (Column 2 of Table 3) in the first stage. The resultant IV estimates scale the reduced form estimates by (1) the drop in the acceptance rate at the cutoff or (2) the drop in the loan volume granted at the cutoff. The results using the acceptance rate in the first stage of the IV implicitly assume that any treatment effect comes from loan rejections, but not from reductions in the loan volume. I therefore focus on the results using the drop in the loan volume granted as the first stage because it seems plausible that credit supply shocks have real effects on both the intensive and the extensive margins. For low-liquidity firms, a EUR 1 decrease in loan supply leads to a EUR 0.36 increase in cash holdings while for a high-liquidity firm, a EUR 1 decrease in loan supply leads to a EUR 0.40 drop in cash holdings. In other words, the original credit supply shock is either amplified by 36% (for low-liquidity firms) or dampened by 40% (high-liquidity firms). These magnitudes clearly seem economically significant. An increase in cash holdings after a credit supply shock implies that noncash assets need to be cut by more than the amount of the credit supply shock. The increase in cash holdi