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 holdings acts like an additional shock to the supply of funding for these firms. Firms have two primary options how to absorb this funding shock: First, firms can cut their other (noncash) current assets, for example, by collecting bills earlier and thereby decreasing trade receivables. Second, firms can cut down on investments, a strategy which has likely more severe consequences for employment at the firm. Panel B of Table 5 tests the first conjecture. The dependent variable is the change in current assets excluding cash and cash equivalents (i.e., the sum of inventories and account receivables). Differences between high- and low-liquidity firms are small and insignificant, suggesting that it is not the management of other current assets that distinguishes low- and high-liquidity firms. Rather, the credit supply shock and the increase in precautionary savings for low-liquidity firms manifests itself in a decrease in investments, as I will discuss in more detail in the next subsection. Table A7 provides results for four bins by liquidity, showing that the highest reductions in noncash asset growth, investment and employment occur in the lowest quartile by pre-application liquidity holdings while the largest drop in cash holdings occurs in the highest quartile by pre-application liquidity holdings. Table A6 confirms the results from Table 5 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). 3.4 The impact of the lender cutoff rule on asset growth, investment, and employment In the next step, I analyze variables that are usually summarized under “real effects”. In particular, these variables include asset growth, investments, and employment. Table 6 reports the results. Table 6 The impact of the lender cutoff rule on firms’ asset growth, investments, and employment Dependent variable A. Asset growth B. Noncash asset growth (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$|-stat Coeff. t-stat Inference BelowCutOff (0/1) –0.078*** (–3.98) –0.068*** (–2.64) –0.074** (–2.17) –0.075*** (–3.58) –0.093*** (–3.92) –0.041 (–0.81) Trends below/Above cutoff (Rating-CutOff) x BelowCutOff (0/1) 0.001 (0.11) 0.008 (0.47) –0.005 (–0.29) –0.003 (–0.22) 0.013 (0.97) –0.019 (–1.05) (Rating-CutOff) x (1- BelowCutOff (0/1)) 0.021 (1.57) 0.020 (1.16) 0.033 (1.34) 0.022 (1.59) 0.026 (1.51) 0.025 (0.79) Firm controls Yes Yes Yes Yes Yes Yes Loan controls Yes Yes Yes Yes Yes Yes Industry x Time fixed effects Yes Yes Yes Yes Yes Yes Region x Time fixed effects Yes Yes Yes Yes Yes Yes Diagnostics Adj. R|$^{\mathrm{2}}$| 15.84% 13.20% 19.78% 8.28% 11.47% 9.13% N 4,714 2,279 2,435 4,714 2,279 2,435 Difference between BelowCutoOff(01/1) |$\Delta $|Coeff. X|$^{\mathrm{2}}$| |$\Delta $|Coeff. X|$^{\mathrm{2}}$| - Low minus high liquidity 0.006 (0.02) –0.052 (1.85) IV estimates First stage: Acceptance dummy –0.286*** (–9.83) –0.335*** (–9.32) –0.225*** (–4.55) –0.286*** (–9.83) –0.335*** (–9.32) –0.225*** (–4.55) IV: using Acceptance dummy 0.269*** (3.83) 0.204*** (3.17) 0.324* (1.89) 0.262*** (3.12) 0.279*** (4.23) 0.182 (0.96) First stage: Change in loan volume –0.074*** (–4.26) –0.069*** (–4.21) –0.079*** (–2.81) –0.074*** (–4.26) –0.069*** (–4.21) –0.079*** (–2.81) IV: Using change in loan volume 1.047*** (5.02) 0.986*** (2.73) 0.942*** (2.72) 1.047*** (3.93) 1.349*** (3.34) 0.541 (1.39) Dependent variable A. Asset growth B. Noncash asset growth (1) (2) (3) (4) (5) (6) Sample All firms Low liquidity (CA/CL |$\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$|-stat Coeff. t-stat Inference BelowCutOff (0/1) –0.078*** (–3.98) –0.068*** (–2.64) –0.074** (–2.17) –0.075*** (–3.58) –0.093*** (–3.92) –0.041 (–0.81) Trends below/Above cutoff (Rating-CutOff) x BelowCutOff (0/1) 0.001 (0.11) 0.008 (0.47) –0.005 (–0.29) –0.003 (–0.22) 0.013 (0.97) –0.019 (–1.05) (Rating-CutOff) x (1- BelowCutOff (0/1)) 0.021 (1.57) 0.020 (1.16) 0.033 (1.34) 0.022 (1.59) 0.026 (1.51) 0.025 (0.79) Firm controls Yes Yes Yes Yes Yes Yes Loan controls Yes Yes Yes Yes Yes Yes Industry x Time fixed effects Yes Yes Yes Yes Yes Yes Region x Time fixed effects Yes Yes Yes Yes Yes Yes Diagnostics Adj. R|$^{\mathrm{2}}$| 15.84% 13.20% 19.78% 8.28% 11.47% 9.13% N 4,714 2,279 2,435 4,714 2,279 2,435 Difference between BelowCutoOff(01/1) |$\Delta $|Coeff. X|$^{\mathrm{2}}$| |$\Delta $|Coeff. X|$^{\mathrm{2}}$| - Low minus high liquidity 0.006 (0.02) –0.052 (1.85) IV estimates First stage: Acceptance dummy –0.286*** (–9.83) –0.335*** (–9.32) –0.225*** (–4.55) –0.286*** (–9.83) –0.335*** (–9.32) –0.225*** (–4.55) IV: using Acceptance dummy 0.269*** (3.83) 0.204*** (3.17) 0.324* (1.89) 0.262*** (3.12) 0.279*** (4.23) 0.182 (0.96) First stage: Change in loan volume –0.074*** (–4.26) –0.069*** (–4.21) –0.079*** (–2.81) –0.074*** (–4.26) –0.069*** (–4.21) –0.079*** (–2.81) IV: Using change in loan volume 1.047*** (5.02) 0.986*** (2.73) 0.942*** (2.72) 1.047*** (3.93) 1.349*** (3.34) 0.541 (1.39) Dependent variable C. Change in investment D. Change in employment (7) (8) (9) (10) (11) (12) Sample All firms Low liquidity (CA/CL |$\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$|-stat Coeff. t-stat Inference BelowCutOff (0/1) –0.024** (–2.53) –0.044*** (–3.47) –0.009 (–0.62) –0.018 (–0.78) –0.070** (–2.18) 0.015 (0.39) Trends below/Above cutoff (Rating-CutOff) x BelowCutOff (0/1) 0.002 (0.34) 0.011 (1.20) –0.002 (–0.35) –0.017 (–1.17) 0.018 (0.79) –0.026 (–1.38) (Rating-CutOff) x (1- BelowCutOff (0/1)) –0.003 (–0.31) 0.001 (0.16) 0.004 (0.40) –0.011 (–0.71) 0.019 (0.87) –0.045* (–1.81) Firm controls Yes Yes Yes Yes Yes Yes Loan controls Yes Yes Yes Yes Yes Yes Industry x Time fixed effects Yes Yes Yes Yes Yes Yes Region x Time fixed effects Yes Yes Yes Yes Yes Yes Diagnostics Adj. R|$^{\mathrm{2}}$| 6.48% 6.41% 6.72% 10.68% 14.15% 14.69% N 4,714 2,279 2,435 3,295 1,577 1,718 Difference between BelowCutoOff(01/1) |$\Delta $|Coeff. X|$^{\mathrm{2}}$| |$\Delta $|Coeff. X|$^{\mathrm{2}}$| - Low minus high liquidity –0.035* (3.24) –0.085* (2.78) IV estimates First stage: Acceptance dummy –0.286*** (–9.83) –0.335*** (–9.32) –0.225*** (–4.55) –0.227*** (–5.96) –0.231*** (–5.31) –0.209*** (–3.17) IV: using Acceptance dummy 0.085** (2.14) 0.131*** (2.93) 0.040 (0.53) 0.081 (0.68) 0.304** (2.24) –0.071 (–0.28) First stage: Change in loan volume –0.074*** (–4.26) –0.069*** (–4.21) –0.079*** (–2.81) –0.073*** (–4.46) –0.047*** (–2.64) –0.115*** (–4.40) IV: Using change in loan volume 0.330** (2.52) 0.644*** (2.61) 0.113 (0.72) 0.251 (0.83) 1.503*** (2.80) –0.129 (–0.24) Dependent variable C. Change in investment D. Change in employment (7) (8) (9) (10) (11) (12) Sample All firms Low liquidity (CA/CL |$\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$|-stat Coeff. t-stat Inference BelowCutOff (0/1) –0.024** (–2.53) –0.044*** (–3.47) –0.009 (–0.62) –0.018 (–0.78) –0.070** (–2.18) 0.015 (0.39) Trends below/Above cutoff (Rating-CutOff) x BelowCutOff (0/1) 0.002 (0.34) 0.011 (1.20) –0.002 (–0.35) –0.017 (–1.17) 0.018 (0.79) –0.026 (–1.38) (Rating-CutOff) x (1- BelowCutOff (0/1)) –0.003 (–0.31) 0.001 (0.16) 0.004 (0.40) –0.011 (–0.71) 0.019 (0.87) –0.045* (–1.81) Firm controls Yes Yes Yes Yes Yes Yes Loan controls Yes Yes Yes Yes Yes Yes Industry x Time fixed effects Yes Yes Yes Yes Yes Yes Region x Time fixed effects Yes Yes Yes Yes Yes Yes Diagnostics Adj. R|$^{\mathrm{2}}$| 6.48% 6.41% 6.72% 10.68% 14.15% 14.69% N 4,714 2,279 2,435 3,295 1,577 1,718 Difference between BelowCutoOff(01/1) |$\Delta $|Coeff. X|$^{\mathrm{2}}$| |$\Delta $|Coeff. X|$^{\mathrm{2}}$| - Low minus high liquidity –0.035* (3.24) –0.085* (2.78) IV estimates First stage: Acceptance dummy –0.286*** (–9.83) –0.335*** (–9.32) –0.225*** (–4.55) –0.227*** (–5.96) –0.231*** (–5.31) –0.209*** (–3.17) IV: using Acceptance dummy 0.085** (2.14) 0.131*** (2.93) 0.040 (0.53) 0.081 (0.68) 0.304** (2.24) –0.071 (–0.28) First stage: Change in loan volume –0.074*** (–4.26) –0.069*** (–4.21) –0.079*** (–2.81) –0.073*** (–4.46) –0.047*** (–2.64) –0.115*** (–4.40) IV: Using change in loan volume 0.330** (2.52) 0.644*** (2.61) 0.113 (0.72) 0.251 (0.83) 1.503*** (2.80) –0.129 (–0.24) This table estimates the effect of the lender cutoff rule on real effects using a regression discontinuity design. Results are reported for the sample of all small firms (first column in each panel) and are split by the median of liquidity (measured as the ratio of current assets to current liabilities in the fiscal year prior to the loan application date). Panel A provides results for asset growth; panel B provides results for noncash asset growth (noncash assets are defined as total asset minus cash and cash equivalents); panel C provides results for investment; and panel D provides results for employment. The upper part of the table provides reduced form estimates, the lower part provides IV estimates. All models are estimated using a linear model. For variable definitions, see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Table 6 The impact of the lender cutoff rule on firms’ asset growth, investments, and employment Dependent variable A. Asset growth B. Noncash asset growth (1) (2) (3) (4) (5) (6) Sample All firms Low liquidity (CA/CL |$\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$|-stat Coeff. t-stat Inference BelowCutOff (0/1) –0.078*** (–3.98) –0.068*** (–2.64) –0.074** (–2.17) –0.075*** (–3.58) –0.093*** (–3.92) –0.041 (–0.81) Trends below/Above cutoff (Rating-CutOff) x BelowCutOff (0/1) 0.001 (0.11) 0.008 (0.47) –0.005 (–0.29) –0.003 (–0.22) 0.013 (0.97) –0.019 (–1.05) (Rating-CutOff) x (1- BelowCutOff (0/1)) 0.021 (1.57) 0.020 (1.16) 0.033 (1.34) 0.022 (1.59) 0.026 (1.51) 0.025 (0.79) Firm controls Yes Yes Yes Yes Yes Yes Loan controls Yes Yes Yes Yes Yes Yes Industry x Time fixed effects Yes Yes Yes Yes Yes Yes Region x Time fixed effects Yes Yes Yes Yes Yes Yes Diagnostics Adj. R|$^{\mathrm{2}}$| 15.84% 13.20% 19.78% 8.28% 11.47% 9.13% N 4,714 2,279 2,435 4,714 2,279 2,435 Difference between BelowCutoOff(01/1) |$\Delta $|Coeff. X|$^{\mathrm{2}}$| |$\Delta $|Coeff. X|$^{\mathrm{2}}$| - Low minus high liquidity 0.006 (0.02) –0.052 (1.85) IV estimates First stage: Acceptance dummy –0.286*** (–9.83) –0.335*** (–9.32) –0.225*** (–4.55) –0.286*** (–9.83) –0.335*** (–9.32) –0.225*** (–4.55) IV: using Acceptance dummy 0.269*** (3.83) 0.204*** (3.17) 0.324* (1.89) 0.262*** (3.12) 0.279*** (4.23) 0.182 (0.96) First stage: Change in loan volume –0.074*** (–4.26) –0.069*** (–4.21) –0.079*** (–2.81) –0.074*** (–4.26) –0.069*** (–4.21) –0.079*** (–2.81) IV: Using change in loan volume 1.047*** (5.02) 0.986*** (2.73) 0.942*** (2.72) 1.047*** (3.93) 1.349*** (3.34) 0.541 (1.39) Dependent variable A. Asset growth B. Noncash asset growth (1) (2) (3) (4) (5) (6) Sample All firms Low liquidity (CA/CL |$\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$|-stat Coeff. t-stat Inference BelowCutOff (0/1) –0.078*** (–3.98) –0.068*** (–2.64) –0.074** (–2.17) –0.075*** (–3.58) –0.093*** (–3.92) –0.041 (–0.81) Trends below/Above cutoff (Rating-CutOff) x BelowCutOff (0/1) 0.001 (0.11) 0.008 (0.47) –0.005 (–0.29) –0.003 (–0.22) 0.013 (0.97) –0.019 (–1.05) (Rating-CutOff) x (1- BelowCutOff (0/1)) 0.021 (1.57) 0.020 (1.16) 0.033 (1.34) 0.022 (1.59) 0.026 (1.51) 0.025 (0.79) Firm controls Yes Yes Yes Yes Yes Yes Loan controls Yes Yes Yes Yes Yes Yes Industry x Time fixed effects Yes Yes Yes Yes Yes Yes Region x Time fixed effects Yes Yes Yes Yes Yes Yes Diagnostics Adj. R|$^{\mathrm{2}}$| 15.84% 13.20% 19.78% 8.28% 11.47% 9.13% N 4,714 2,279 2,435 4,714 2,279 2,435 Difference between BelowCutoOff(01/1) |$\Delta $|Coeff. X|$^{\mathrm{2}}$| |$\Delta $|Coeff. X|$^{\mathrm{2}}$| - Low minus high liquidity 0.006 (0.02) –0.052 (1.85) IV estimates First stage: Acceptance dummy –0.286*** (–9.83) –0.335*** (–9.32) –0.225*** (–4.55) –0.286*** (–9.83) –0.335*** (–9.32) –0.225*** (–4.55) IV: using Acceptance dummy 0.269*** (3.83) 0.204*** (3.17) 0.324* (1.89) 0.262*** (3.12) 0.279*** (4.23) 0.182 (0.96) First stage: Change in loan volume –0.074*** (–4.26) –0.069*** (–4.21) –0.079*** (–2.81) –0.074*** (–4.26) –0.069*** (–4.21) –0.079*** (–2.81) IV: Using change in loan volume 1.047*** (5.02) 0.986*** (2.73) 0.942*** (2.72) 1.047*** (3.93) 1.349*** (3.34) 0.541 (1.39) Dependent variable C. Change in investment D. Change in employment (7) (8) (9) (10) (11) (12) Sample All firms Low liquidity (CA/CL |$\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$|-stat Coeff. t-stat Inference BelowCutOff (0/1) –0.024** (–2.53) –0.044*** (–3.47) –0.009 (–0.62) –0.018 (–0.78) –0.070** (–2.18) 0.015 (0.39) Trends below/Above cutoff (Rating-CutOff) x BelowCutOff (0/1) 0.002 (0.34) 0.011 (1.20) –0.002 (–0.35) –0.017 (–1.17) 0.018 (0.79) –0.026 (–1.38) (Rating-CutOff) x (1- BelowCutOff (0/1)) –0.003 (–0.31) 0.001 (0.16) 0.004 (0.40) –0.011 (–0.71) 0.019 (0.87) –0.045* (–1.81) Firm controls Yes Yes Yes Yes Yes Yes Loan controls Yes Yes Yes Yes Yes Yes Industry x Time fixed effects Yes Yes Yes Yes Yes Yes Region x Time fixed effects Yes Yes Yes Yes Yes Yes Diagnostics Adj. R|$^{\mathrm{2}}$| 6.48% 6.41% 6.72% 10.68% 14.15% 14.69% N 4,714 2,279 2,435 3,295 1,577 1,718 Difference between BelowCutoOff(01/1) |$\Delta $|Coeff. X|$^{\mathrm{2}}$| |$\Delta $|Coeff. X|$^{\mathrm{2}}$| - Low minus high liquidity –0.035* (3.24) –0.085* (2.78) IV estimates First stage: Acceptance dummy –0.286*** (–9.83) –0.335*** (–9.32) –0.225*** (–4.55) –0.227*** (–5.96) –0.231*** (–5.31) –0.209*** (–3.17) IV: using Acceptance dummy 0.085** (2.14) 0.131*** (2.93) 0.040 (0.53) 0.081 (0.68) 0.304** (2.24) –0.071 (–0.28) First stage: Change in loan volume –0.074*** (–4.26) –0.069*** (–4.21) –0.079*** (–2.81) –0.073*** (–4.46) –0.047*** (–2.64) –0.115*** (–4.40) IV: Using change in loan volume 0.330** (2.52) 0.644*** (2.61) 0.113 (0.72) 0.251 (0.83) 1.503*** (2.80) –0.129 (–0.24) Dependent variable C. Change in investment D. Change in employment (7) (8) (9) (10) (11) (12) Sample All firms Low liquidity (CA/CL |$\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$|-stat Coeff. t-stat Inference BelowCutOff (0/1) –0.024** (–2.53) –0.044*** (–3.47) –0.009 (–0.62) –0.018 (–0.78) –0.070** (–2.18) 0.015 (0.39) Trends below/Above cutoff (Rating-CutOff) x BelowCutOff (0/1) 0.002 (0.34) 0.011 (1.20) –0.002 (–0.35) –0.017 (–1.17) 0.018 (0.79) –0.026 (–1.38) (Rating-CutOff) x (1- BelowCutOff (0/1)) –0.003 (–0.31) 0.001 (0.16) 0.004 (0.40) –0.011 (–0.71) 0.019 (0.87) –0.045* (–1.81) Firm controls Yes Yes Yes Yes Yes Yes Loan controls Yes Yes Yes Yes Yes Yes Industry x Time fixed effects Yes Yes Yes Yes Yes Yes Region x Time fixed effects Yes Yes Yes Yes Yes Yes Diagnostics Adj. R|$^{\mathrm{2}}$| 6.48% 6.41% 6.72% 10.68% 14.15% 14.69% N 4,714 2,279 2,435 3,295 1,577 1,718 Difference between BelowCutoOff(01/1) |$\Delta $|Coeff. X|$^{\mathrm{2}}$| |$\Delta $|Coeff. X|$^{\mathrm{2}}$| - Low minus high liquidity –0.035* (3.24) –0.085* (2.78) IV estimates First stage: Acceptance dummy –0.286*** (–9.83) –0.335*** (–9.32) –0.225*** (–4.55) –0.227*** (–5.96) –0.231*** (–5.31) –0.209*** (–3.17) IV: using Acceptance dummy 0.085** (2.14) 0.131*** (2.93) 0.040 (0.53) 0.081 (0.68) 0.304** (2.24) –0.071 (–0.28) First stage: Change in loan volume –0.074*** (–4.26) –0.069*** (–4.21) –0.079*** (–2.81) –0.073*** (–4.46) –0.047*** (–2.64) –0.115*** (–4.40) IV: Using change in loan volume 0.330** (2.52) 0.644*** (2.61) 0.113 (0.72) 0.251 (0.83) 1.503*** (2.80) –0.129 (–0.24) This table estimates the effect of the lender cutoff rule on real effects using a regression discontinuity design. Results are reported for the sample of all small firms (first column in each panel) and are split by the median of liquidity (measured as the ratio of current assets to current liabilities in the fiscal year prior to the loan application date). Panel A provides results for asset growth; panel B provides results for noncash asset growth (noncash assets are defined as total asset minus cash and cash equivalents); panel C provides results for investment; and panel D provides results for employment. 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. 3.4.1 Effects on asset growth, investments, and employment Panel A of Table 6 looks at asset growth. Firms below the cutoff decrease their assets by an average of 7.8% relative to firms above the cutoff. The effect is similar for firms with low liquidity and firms with high liquidity (Columns 2 and 3 in panel A). Panel B focuses on noncash assets only, that is, at total assets minus cash and cash equivalents. Firms with low liquidity cut their noncash assets by 9.3% (significant at the 1% level), while firms with high liquidity only cut their noncash assets by an insignificant 4.1%. The IV results at the bottom of the table show that low-liquidity firms cut their noncash assets by 135% of the original credit supply shock, while high-liquidity firms—using their cash buffers—only cut noncash assets by 54% of the original credit supply shock. These results suggest that precautionary savings motives amplify the credit supply shock for low-liquidity firms as these firms cut their noncash assets by more than their total assets. Panel C of Table 6, which looks at changes in investment, supports this narrative. While low-liquidity firms below the cutoff cut their investment by 4.4% (significant at the 1% level) relative to low-liquidity firms below the cutoff, the respective coefficient for high-liquidity firms is small (-0.9%) and insignificant. In relation to the credit supply shock, low-liquidity firms cut their investments by a significant EUR 0.64 per EUR 1 credit supply shock, while high-liquidity firms only cut their investments by an insignificant EUR 0.11 per EUR 1 credit supply shock. Results for employment are similar, with a negative and significant effect for low-liquidity firms (-7.0%, significant at the 5% level) and an insignificant effect for high-liquidity firms. The IV results at the bottom of the table suggest that the percentage cut in the workforce of low-liquidity firms is a factor of 1.5 larger than the percentage cut in loan-volume-to-total-assets; that is, employment drops somehow faster than the credit-supply-shock-induced drop in assets. These results suggest that employment losses after a credit supply shock are concentrated at firms with low-liquidity holdings, and are amplified by precautionary savings motives of these firms. Table A6 confirms the results from Table 6 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). Using a triangular kernel instead of a standard ordinary least squares (OLS)/rectangular kernel usually results in slightly higher standard errors (because observations are no longer equally weighted) but a very similar magnitude of the coefficient of interest. Higher-order polynomials sometimes lead to slightly lower and sometimes to slightly higher estimates of the treatment effect, but generally confirm the results from Table 6. Controlling for past asset growth is important to allay concerns that future growth opportunities are systematically different for firms above versus below the cutoff. Results are almost unchanged when controlling for past asset growth, suggesting that this is not the case. Using a more granular industry definition—99 industries instead of 14; dropping fixed effects or dropping all fixed effects; and controls apart from the running variable have little effect on the results—consistent with the logic of an RDD that variation across the threshold is as good as random. Shrinking the bandwidth lowers statistical significance, but there is no consistent change of the coefficient in either direction. Figure 3 View largeDownload slide Lending standards and real effects over time This figure depicts RDD estimates for key variables from Tables 3, 5, and 6 over time. The figures depict the key coefficient from the reduced-form RDD estimates (bold line) and |$\pm 1.65$| standard error bands (gray-shaded area). The regressions are based on the same specification (control variables and bandwidth) used in Tables 3, 5, and 6, the only difference being that the BelowCutOff dummy is interacted with a year dummy. Figure 3 View largeDownload slide Lending standards and real effects over time This figure depicts RDD estimates for key variables from Tables 3, 5, and 6 over time. The figures depict the key coefficient from the reduced-form RDD estimates (bold line) and |$\pm 1.65$| standard error bands (gray-shaded area). The regressions are based on the same specification (control variables and bandwidth) used in Tables 3, 5, and 6, the only difference being that the BelowCutOff dummy is interacted with a year dummy. 3.4.2 Time-series heterogeneity Figure 3 provides information on changes in lending standards over time as well as real effects over time. While the cutoff remained unchanged during the sample period, the drop in rejection rates at the cutoff is largest in 2009, suggesting that lending standards were tighter in 2009 than in the subsequent years.21 Changes in cash holdings and investments are also largest in 2009 and decay until 2012. This is consistent with rejection rates being higher in 2009 than in subsequent years, with the “real effect per loan rejection” being roughly stable over time. An exception is the employment in which effects are small in 2009, increase in 2010 and then slowly decay until 2012. The small effect in 2009 is consistent with a short-term worker scheme in 2009-2011 in Germany that subsidized wages for German firms if these firms reduced working hours per employee instead of laying off workers. At the height of the crisis, this scheme was used for almost 1.5 million workers in Germany and helped to cushion job losses at the beginning of the financial crisis (Brenke, Rinne, and Zimmermann 2012). Taken together, Figure 3 reinforces the view that firms with high and low liquidity react differently to credit supply shocks, with low-liquidity firms increasing their cash holdings and thereby amplifying real effects of credit supply shocks. Overall, the results suggest that pooling firms together provides an incomplete picture of the adjustment process after a credit supply shock. This adjustment process crucially depends on the liquidity of the firm: while firms with high liquidity are able to cushion credit supply shocks, firms with low liquidity need to cut their investment and see a significant decrease in employment. These real effects are amplified by precautionary savings motives: low-liquidity firms increase their cash holdings after a credit supply shock and cut their noncash assets by more than what the direct impact of the credit supply shock would imply. 3.5 Robustness: Panel data and dynamics over time The RDD analysis focuses on a cross-sectional analysis, comparing cash holdings and real effects for firms below and above the cutoff. In this subsection, I explore dynamics over time, comparing how treated firms (worse than the cutoff) develop relative to control group firms (better than the cutoff) in the years preceding and following the loan application. The corresponding difference-in-differences estimator is \begin{align} \label{eq1} y_{i,t} &=\beta_{-n} \cdot \textit{BelowCutOff}(0/1)\cdot \textit{Year}_{-n} +...+\beta_{n} \cdot \textit{BelowCutOff}(0/1)\cdot \textit{Year}_{n} \notag\\ &\quad + \textit{Controls} +\varepsilon \end{align} (2) where |$y_{i,t}$| is the variable of interest (e.g., cash holdings) of firm |$i$| in year |$t$|⁠, BelowCutOff(0/1) is a dummy equal to one if the rating is below the cutoff rating, Year|$_{X}$| is a variable which is equal to one if the observation is from a year which is X years away from the last year prior to the loan application (e.g., for a loan application from July 2010, Year|$_{0}$| is equal to one for data from the annual report as of the end of 2009, Year|$_{-1}$| is equal to data for data from the annual report as of the end of 2008, and Year|$_{+1}$| is equal to one for data from the annual report as of the end of 2010). Importantly, I estimate (2) using firm fixed effects, implying that the |$\beta $|-estimates are not affected by unobservable time-invariant firm characteristics. Furthermore, I control for industry x year fixed effects, one-digit ZIP code x year fixed effects, and relative-year fixed effects (Year|$_{-n}$| to Year|$_{n})$|⁠.22 Like in the RDD design, I estimate Equation (2) using a linear model and all standard errors are clustered at the branch level. The difference-in-differences specification provides two main additional insights over and above the RDD specification. First, I can analyze firm dynamics over time. Second, it allows using a narrower bandwidth around the cutoff and therefore treated and control group firms are even more similar in terms of credit quality. This is because in the difference-in-differences panel data setup, each firm occurs multiple times in the data set, thus allowing a more precise measurement of the various fixed effects. In the following, I use a sample which spans the 2 years before and the 2 years after the loan application and which focuses on firms |$\pm 1$| rating notch around the threshold.23 Results are reported in Table 7 for the key sample of interest—the low-liquidity firms—and for the key variables of interest from Tables 5 and 6. Cash holding, reported in Column 1, do not show any significant trend for the below-cutoff firms before the loan application relative to firms above the cutoff. After the loan application, rejected firms increase cash holdings gradually by 1.2% (Year 1 after the loan application, which is on average 6 months after the loan application) and 2.1% (Year 2 after the loan application, which is on average 18 months after the loan application). The RDD specification uses changes from Year|$_{0}$| to Year|$_{2}$| as the dependent variable, and it is therefore comforting to see that the Year|$_{2}$| result from the difference-in-differences estimator (⁠|$+$|2.1%) is similar to the results from the RDD specification (⁠|$+$|2.5%, see Column 2 of Table 5). Again, the results are also economically significant, implying an 18% increase relative to the unconditional mean cash holdings of 12% of total assets. The results suggest that rejected firms gradually build up cash buffers to their new target level, consistent with the idea that an immediate increase might require costly fire sales of existing assets. Table 7 Robustness: Differences-in-differences design (1) (2) (3) (4) Sample Low liquidity (CA/CL |$\leqslant $| 1.4) Low liquidity (CA/CL |$\leqslant $| 1.4) Low liquidity (CA/CL |$\leqslant $| 1.4) Low liquidity (CA/CL |$\leqslant $| 1.4) Dependent variable Change in cash and cash equivalents Noncash asset growth Investments Employment Methodology Diff-in-diff |$\pm 1$| notches around cutoff Diff-in-diff |$\pm 1$| notches around cutoff Diff-in-diff |$\pm 1$| notches around cutoff Diff-in-diff |$\pm 1$| notches around cutoff Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Inference BelowCutOff (0/1) x Year -1 0.007 (1.01) –0.013 (–0.49) 0.014 (1.55) 0.005 (0.23) BelowCutOff (0/1) x Year 0 Omitted (0.000) Omitted (0.000) Omitted (0.000) Omitted (0.000) BelowCutOff (0/1) x Year |$+$|1 0.012* (1.90) –0.029 (–0.93) –0.020* (–1.68) –0.024 (–0.94) BelowCutOff (0/1) x Year |$+$|2 0.021** (2.16) –0.127*** (–3.16) –0.034** (2.61) –0.066* (–1.88) Firm fixed effects Yes Yes Yes Yes Industry x Time fixed effects Yes Yes Yes Yes Region x Time fixed effects Yes Yes Yes Yes Relative-year fixed effects (Year -1 to Year |$+$|2) Yes Yes Yes Yes Diagnostics Adj. R|$^{\mathrm{2}}$|(%) 55.60 68.63 79.04 89.14 N 5,487 5,487 5,487 4,160 (1) (2) (3) (4) Sample Low liquidity (CA/CL |$\leqslant $| 1.4) Low liquidity (CA/CL |$\leqslant $| 1.4) Low liquidity (CA/CL |$\leqslant $| 1.4) Low liquidity (CA/CL |$\leqslant $| 1.4) Dependent variable Change in cash and cash equivalents Noncash asset growth Investments Employment Methodology Diff-in-diff |$\pm 1$| notches around cutoff Diff-in-diff |$\pm 1$| notches around cutoff Diff-in-diff |$\pm 1$| notches around cutoff Diff-in-diff |$\pm 1$| notches around cutoff Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Inference BelowCutOff (0/1) x Year -1 0.007 (1.01) –0.013 (–0.49) 0.014 (1.55) 0.005 (0.23) BelowCutOff (0/1) x Year 0 Omitted (0.000) Omitted (0.000) Omitted (0.000) Omitted (0.000) BelowCutOff (0/1) x Year |$+$|1 0.012* (1.90) –0.029 (–0.93) –0.020* (–1.68) –0.024 (–0.94) BelowCutOff (0/1) x Year |$+$|2 0.021** (2.16) –0.127*** (–3.16) –0.034** (2.61) –0.066* (–1.88) Firm fixed effects Yes Yes Yes Yes Industry x Time fixed effects Yes Yes Yes Yes Region x Time fixed effects Yes Yes Yes Yes Relative-year fixed effects (Year -1 to Year |$+$|2) Yes Yes Yes Yes Diagnostics Adj. R|$^{\mathrm{2}}$|(%) 55.60 68.63 79.04 89.14 N 5,487 5,487 5,487 4,160 This table estimates the effect of the lender cutoff rule on cash holdings and real effects using a differences-in-differences design. Results are reported for the sample of low-liquidity firms (measured as the ratio of current assets to current liabilities in the fiscal year prior to the loan application date). Column 1 uses Cash and cash equivalents as the dependent variable, and Columns 2–4 analyze real effects (noncash asset growth, investments, employment). The variables BelowCutOff (0/1) x Year X depict the development of the treated firms (rating worse than the cutoff) relative to the control groups (rating better than the cutoff) before and after the loan application. Year X refers to the year relative to the loan application. For example, Year 0 refers to the last year end before the loan application (e.g., for a loan application from July 2010, Year 0 refers to data from the annual report as of the end of 2009). The sample is restricted to firms with ratings |$\pm 1$| notches around the cutoff. All models are estimated using a linear model. For variable definitions, see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Table 7 Robustness: Differences-in-differences design (1) (2) (3) (4) Sample Low liquidity (CA/CL |$\leqslant $| 1.4) Low liquidity (CA/CL |$\leqslant $| 1.4) Low liquidity (CA/CL |$\leqslant $| 1.4) Low liquidity (CA/CL |$\leqslant $| 1.4) Dependent variable Change in cash and cash equivalents Noncash asset growth Investments Employment Methodology Diff-in-diff |$\pm 1$| notches around cutoff Diff-in-diff |$\pm 1$| notches around cutoff Diff-in-diff |$\pm 1$| notches around cutoff Diff-in-diff |$\pm 1$| notches around cutoff Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Inference BelowCutOff (0/1) x Year -1 0.007 (1.01) –0.013 (–0.49) 0.014 (1.55) 0.005 (0.23) BelowCutOff (0/1) x Year 0 Omitted (0.000) Omitted (0.000) Omitted (0.000) Omitted (0.000) BelowCutOff (0/1) x Year |$+$|1 0.012* (1.90) –0.029 (–0.93) –0.020* (–1.68) –0.024 (–0.94) BelowCutOff (0/1) x Year |$+$|2 0.021** (2.16) –0.127*** (–3.16) –0.034** (2.61) –0.066* (–1.88) Firm fixed effects Yes Yes Yes Yes Industry x Time fixed effects Yes Yes Yes Yes Region x Time fixed effects Yes Yes Yes Yes Relative-year fixed effects (Year -1 to Year |$+$|2) Yes Yes Yes Yes Diagnostics Adj. R|$^{\mathrm{2}}$|(%) 55.60 68.63 79.04 89.14 N 5,487 5,487 5,487 4,160 (1) (2) (3) (4) Sample Low liquidity (CA/CL |$\leqslant $| 1.4) Low liquidity (CA/CL |$\leqslant $| 1.4) Low liquidity (CA/CL |$\leqslant $| 1.4) Low liquidity (CA/CL |$\leqslant $| 1.4) Dependent variable Change in cash and cash equivalents Noncash asset growth Investments Employment Methodology Diff-in-diff |$\pm 1$| notches around cutoff Diff-in-diff |$\pm 1$| notches around cutoff Diff-in-diff |$\pm 1$| notches around cutoff Diff-in-diff |$\pm 1$| notches around cutoff Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Inference BelowCutOff (0/1) x Year -1 0.007 (1.01) –0.013 (–0.49) 0.014 (1.55) 0.005 (0.23) BelowCutOff (0/1) x Year 0 Omitted (0.000) Omitted (0.000) Omitted (0.000) Omitted (0.000) BelowCutOff (0/1) x Year |$+$|1 0.012* (1.90) –0.029 (–0.93) –0.020* (–1.68) –0.024 (–0.94) BelowCutOff (0/1) x Year |$+$|2 0.021** (2.16) –0.127*** (–3.16) –0.034** (2.61) –0.066* (–1.88) Firm fixed effects Yes Yes Yes Yes Industry x Time fixed effects Yes Yes Yes Yes Region x Time fixed effects Yes Yes Yes Yes Relative-year fixed effects (Year -1 to Year |$+$|2) Yes Yes Yes Yes Diagnostics Adj. R|$^{\mathrm{2}}$|(%) 55.60 68.63 79.04 89.14 N 5,487 5,487 5,487 4,160 This table estimates the effect of the lender cutoff rule on cash holdings and real effects using a differences-in-differences design. Results are reported for the sample of low-liquidity firms (measured as the ratio of current assets to current liabilities in the fiscal year prior to the loan application date). Column 1 uses Cash and cash equivalents as the dependent variable, and Columns 2–4 analyze real effects (noncash asset growth, investments, employment). The variables BelowCutOff (0/1) x Year X depict the development of the treated firms (rating worse than the cutoff) relative to the control groups (rating better than the cutoff) before and after the loan application. Year X refers to the year relative to the loan application. For example, Year 0 refers to the last year end before the loan application (e.g., for a loan application from July 2010, Year 0 refers to data from the annual report as of the end of 2009). The sample is restricted to firms with ratings |$\pm 1$| notches around the cutoff. All models are estimated using a linear model. For variable definitions, see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Results for noncash asset growth, investments, and employment are reported in Columns 2–4 of Table 7. In all cases, there is no evidence for a pre-event trend and effects gradually increase from Year|$_{1}$| to Year|$_{2}$| after the loan application. While firms below the cutoff increase cash, I find evidence that they cut noncash assets (by |$-2.9{\%}$| 1 year after the loan application, by |$-12.7{\%}$| 2 years after the loan application), cut investments (by |$-2.0{\%}$| of total assets 1 year after the loan application, by |$-3.4{\%}$| 2 years after the loan application) and employment (by |$-2.4{\%}$| 1 year after the loan application, by |$-6.6{\%}$| after the loan application). Taken together, the difference-in-differences results (1) do not find evidence for pre-event trends, (2) confirm the results from the RDD using a narrower bandwidth around the cutoff in a panel data setting, and (3) provide evidence that real effects from loan rejections gradually build up over the first 2 years after the loan rejection. 3.3 Robustness: Holdout sample of larger loans I provide a robustness test for loan applications larger than EUR 1 million for exactly the same sample period (2009–2012). For loan application above EUR 1 million, there is no cutoff rule and an additional review by a credit officer is required for each application independent of the rating (see also Footnote 2). If there are any omitted variable in the prior analysis that are correlated with future growth opportunities, we should also see a significant RDD estimate for this holdout sample. This sample provides a plausible robustness test for the key RDD assumption; that is, the assumption that in absence of a cutoff rule firms below and above the cutoff are similar in all respects. The robustness test is obviously only meaningful if the correlation between the ratings and the unobserved differences in growth opportunities is unrelated to loan size. The number of observations in this sample is 3,007 (i.e., smaller than for the main sample, which has 16,855 observations), and the interquartile range of loan volume ranges from EUR 1.54 to EUR 3.13 million. Figure A3 shows the distribution of loan applications for loans above EUR 1 million and the acceptance rate by rating. The distribution follows a very similar pattern as for the below-EUR1-million loans, while the acceptance rate does not show any drop at the rating of 7.5.|$^{\mathrm{\thinspace }}$|24 Table A8 reports the same tests used in Tables 3, 5, and 6 for the holdout sample of loan applications above EUR 1 million. For ease of comparison, Table A8 also reports the results from Tables 3, 5, and 6. In the holdout sample, 22 of the 24 specifications are statistically insignificant and in the two significant cases (holdout results for panel D of Table 6), coefficients go in the opposite direction compared to the results from Table 6. Standard errors are larger due to the smaller sample size. However, coefficients for low- versus high-liquidity firms are usually both positive or negative, suggesting that results from the holdout sample are not driven by higher standard errors only. Taken together, the holdout sample supports the evidence from the main sample, suggesting that differences in post-application growth rates across the cutoff in the main sample are indeed caused by the discrete change in credit constraints at the cutoff. 3.7 Alternative interpretations The results above are consistent with a precautionary savings narrative: firms learn about their ability to access debt markets and therefore increase cash holdings for precautionary savings motives after a loan rejection. Precautionary savings motives thereby amplify the credit supply shock as firms cut their noncash assets by more than the initial credit supply shock. Two alternative explanations come to mind: First, the loan rejection might not only be a signal about a firm’s ability to access debt markets, but also a signal about the quality of firms’ investment projects. While it seems plausible that firms update their beliefs about the quality of investment projects after a loan rejection, it is hard to find a narrative where the informativeness of this signal is correlated with cash holdings in the cross-section. Updating beliefs about investment projects should affect low-liquidity firms in a similar manner as high-liquidity firms, and therefore cannot explain the difference between low-liquidity and high-liquidity firms that I find. Second, a firm whose loan application is rejected might instead decide to retain earnings and accumulate cash until it is able to finance the investment project entirely with internal funds. In contrast to the precautionary savings narrative, the increase in cash holdings would not be precautionary, but would be for a specific purpose (i.e., the investment project that the firm wanted to finance with the loan in the first place). This alternative explanation is also consistent with different effects for high- versus low-liquidity firms because high-liquidity firms might be able to accumulate the necessary internal funds faster and subsequently spend the cash to invest into the originally planned project. However, this narrative is inconsistent with the fact that rejected firms don’t increase equity capital after a loan rejection (see Tables 3 and 4). If rejected firms would accumulate internal funds to finance the project, we should observe an accompanying increase in equity funding for rejected firms. Taken together, the fact that (1) only low-liquidity firms increase cash holdings after a loan rejection, and (2) rejected firms don’t increase equity funding is consistent with the precautionary motives narrative but inconsistent with the alternative narratives discussed above. These results therefore point to an important role for precautionary savings motives in the transmission of credit supply shocks and to significant real effects of credit rationing. 4. Conclusion Using exogenous variation induced by a lender cutoff rule, I analyze the real effects of loan rejections. 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 acceptance rate at the threshold. Using almost 17,000 loan applications by small and medium-sized enterprises at a major German bank, I compare the development of firms just above the cutoff rating to those firms that are just below the cutoff rating using a fuzzy regression discontinuity design. The results convey the importance of liquidity in the transmission of credit supply shocks. Real effects such as a reduction in asset growth, investments, and employment are concentrated among low-liquidity firms. Crucially, firms with low liquidity actually increase cash holdings after a credit supply shock. The results thus point to the amplifying effect of precautionary savings motives in the transmission of credit supply shocks. Figure A1 View largeDownload slide View largeDownload slide Control variables by rating This figure depicts the control variables at the fiscal year end prior to the loan application a function of the continuous rating for the sample of all loan applications between January 2009 and December 2012. Panel A depicts variables in levels, and panel B depicts changes in selected variables from 3 years prior to 1 year prior to the loan application (with “1 year prior” being the last observation before the loan application, for example, for a loan application in July 2012, panel B depicts changes from December 2009 to December 2011). For variable definitions, see Table 1. Figure A1 View largeDownload slide View largeDownload slide Control variables by rating This figure depicts the control variables at the fiscal year end prior to the loan application a function of the continuous rating for the sample of all loan applications between January 2009 and December 2012. Panel A depicts variables in levels, and panel B depicts changes in selected variables from 3 years prior to 1 year prior to the loan application (with “1 year prior” being the last observation before the loan application, for example, for a loan application in July 2012, panel B depicts changes from December 2009 to December 2011). For variable definitions, see Table 1. Figure A2 View largeDownload slide The impact of the lender cutoff rule on firms’ financing as a function of bandwidth This figure depicts the results from Table 3 for various functions of the bandwidth. The |$x$|-axis plots the bandwidth. The |$y$|-axis shows the coefficient of the BelowCutOff dummy. The gray-shaded areas depict |$\pm 1$|⁠.65 standard errors around the coefficient of the BelowCutOff dummy. Figures 1–6 correspond to Columns 1–6 in Table 3. Figure A2 View largeDownload slide The impact of the lender cutoff rule on firms’ financing as a function of bandwidth This figure depicts the results from Table 3 for various functions of the bandwidth. The |$x$|-axis plots the bandwidth. The |$y$|-axis shows the coefficient of the BelowCutOff dummy. The gray-shaded areas depict |$\pm 1$|⁠.65 standard errors around the coefficient of the BelowCutOff dummy. Figures 1–6 correspond to Columns 1–6 in Table 3. Figure A3 View largeDownload slide Holdout sample of loans > EUR 1 million This figure reports key graphs for the holdout sample of loans above EUR 1 million, where no cutoff at the rating of 7.5 exists. Panel A shows the distribution of ratings for the sample of loans >EUR 1 million. Panel B reports loan acceptance rates by rating for the sample of loans >EUR 1 million. Figure A3 View largeDownload slide Holdout sample of loans > EUR 1 million This figure reports key graphs for the holdout sample of loans above EUR 1 million, where no cutoff at the rating of 7.5 exists. Panel A shows the distribution of ratings for the sample of loans >EUR 1 million. Panel B reports loan acceptance rates by rating for the sample of loans >EUR 1 million. Table A1 Disclosure requirements by size class Size class 1 (Assets |$\leqslant $| EUR 350,000)|$^{\mathrm{a}}$| Size class 2 (EUR 350,000 < Assets < |$=$| EUR 4.84 mil)|$^{\mathrm{b}}$| Size class 3 (Assets > EUR 4.84 mil)|$^{\mathrm{a}}$| Assets A. Current assets x x x I. Inventory x x II. Trade receivables x x III. Marketable securities x x IV. Cash x x B. Investments x x x I. Intangible assets x x II. Property, plant, and equipment x x III. Financial investments x x C. Other assets (e.g., accruals, deferred tax assets) x x x Total assets x x x Liabilities A. Equity x x x B. Debt x x x 1. Bonds x 2. Bank loans x 3. Trade payables x 4. Other debt x C. Other liabilities (e.g., provisions, accruals, deferred tax liabilities) x x x Total liabilities x x x Size class 1 (Assets |$\leqslant $| EUR 350,000)|$^{\mathrm{a}}$| Size class 2 (EUR 350,000 < Assets < |$=$| EUR 4.84 mil)|$^{\mathrm{b}}$| Size class 3 (Assets > EUR 4.84 mil)|$^{\mathrm{a}}$| Assets A. Current assets x x x I. Inventory x x II. Trade receivables x x III. Marketable securities x x IV. Cash x x B. Investments x x x I. Intangible assets x x II. Property, plant, and equipment x x III. Financial investments x x C. Other assets (e.g., accruals, deferred tax assets) x x x Total assets x x x Liabilities A. Equity x x x B. Debt x x x 1. Bonds x 2. Bank loans x 3. Trade payables x 4. Other debt x C. Other liabilities (e.g., provisions, accruals, deferred tax liabilities) x x x Total liabilities x x x |$^{\mathrm{a}}$|Precise definition: Two out of the following criteria are fulfilled: (1) total assets |$\leqslant $| EUR 350,000, (2) revenues |$\leqslant $| EUR 700,000, or (3) number of employees |$\leqslant $| 10. |$^{\mathrm{b}}$|Precise definition: Not in size class 1, and two out of the following criteria are fulfilled: (1) total assets |$\leqslant $| EUR 4.84 million, (2) revenues |$\leqslant $| EUR 9.68 million, or (3) number of employees |$\leqslant $| 50. This table provides disclosure requirements by size class. An “x” denotes that the respective item needs to be disclosed by the firm in its annual report, and an empty field denotes that the respective items does not need to be disclosed by the firm in its annual report. Rules are according to §266 of Germany’s Commercial Code (“HGB”). Table A1 Disclosure requirements by size class Size class 1 (Assets |$\leqslant $| EUR 350,000)|$^{\mathrm{a}}$| Size class 2 (EUR 350,000 < Assets < |$=$| EUR 4.84 mil)|$^{\mathrm{b}}$| Size class 3 (Assets > EUR 4.84 mil)|$^{\mathrm{a}}$| Assets A. Current assets x x x I. Inventory x x II. Trade receivables x x III. Marketable securities x x IV. Cash x x B. Investments x x x I. Intangible assets x x II. Property, plant, and equipment x x III. Financial investments x x C. Other assets (e.g., accruals, deferred tax assets) x x x Total assets x x x Liabilities A. Equity x x x B. Debt x x x 1. Bonds x 2. Bank loans x 3. Trade payables x 4. Other debt x C. Other liabilities (e.g., provisions, accruals, deferred tax liabilities) x x x Total liabilities x x x Size class 1 (Assets |$\leqslant $| EUR 350,000)|$^{\mathrm{a}}$| Size class 2 (EUR 350,000 < Assets < |$=$| EUR 4.84 mil)|$^{\mathrm{b}}$| Size class 3 (Assets > EUR 4.84 mil)|$^{\mathrm{a}}$| Assets A. Current assets x x x I. Inventory x x II. Trade receivables x x III. Marketable securities x x IV. Cash x x B. Investments x x x I. Intangible assets x x II. Property, plant, and equipment x x III. Financial investments x x C. Other assets (e.g., accruals, deferred tax assets) x x x Total assets x x x Liabilities A. Equity x x x B. Debt x x x 1. Bonds x 2. Bank loans x 3. Trade payables x 4. Other debt x C. Other liabilities (e.g., provisions, accruals, deferred tax liabilities) x x x Total liabilities x x x |$^{\mathrm{a}}$|Precise definition: Two out of the following criteria are fulfilled: (1) total assets |$\leqslant $| EUR 350,000, (2) revenues |$\leqslant $| EUR 700,000, or (3) number of employees |$\leqslant $| 10. |$^{\mathrm{b}}$|Precise definition: Not in size class 1, and two out of the following criteria are fulfilled: (1) total assets |$\leqslant $| EUR 4.84 million, (2) revenues |$\leqslant $| EUR 9.68 million, or (3) number of employees |$\leqslant $| 50. This table provides disclosure requirements by size class. An “x” denotes that the respective item needs to be disclosed by the firm in its annual report, and an empty field denotes that the respective items does not need to be disclosed by the firm in its annual report. Rules are according to §266 of Germany’s Commercial Code (“HGB”). Table A2 Availability of data items: Test for discontinuity at the cutoff rating (1) (2) Dependent Model Change in firm balance sheet items is missing (0/1) Linear Change in the number of employees is missing (0/1) Linear Methodology Local regression |$\pm 2$| notches around cutoff Local regression |$\pm 2$| notches around cutoff Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Inference BelowCutOff (0/1) –0.005 (–0.44) 0.015 (0.76) Trends below/Above cutoff (Rating-CutOff) x BelowCutOff (0/1) –0.006 (–0.84) 0.006 (0.63) (Rating-CutOff) x (1- BelowCutOff (0/1)) 0.023** (2.56) 0.049*** (4.40) Firm controls Yes Yes Loan controls Yes Yes Industry x Time fixed effects Yes Yes Region x Time fixed effects Yes Yes Diagnostics Pseudo. R|$^{\mathrm{2}}$| / Adj. R|$^{\mathrm{2\thinspace }}$|(%) 3.16 7.98 N 10,127 10,127 (1) (2) Dependent Model Change in firm balance sheet items is missing (0/1) Linear Change in the number of employees is missing (0/1) Linear Methodology Local regression |$\pm 2$| notches around cutoff Local regression |$\pm 2$| notches around cutoff Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Inference BelowCutOff (0/1) –0.005 (–0.44) 0.015 (0.76) Trends below/Above cutoff (Rating-CutOff) x BelowCutOff (0/1) –0.006 (–0.84) 0.006 (0.63) (Rating-CutOff) x (1- BelowCutOff (0/1)) 0.023** (2.56) 0.049*** (4.40) Firm controls Yes Yes Loan controls Yes Yes Industry x Time fixed effects Yes Yes Region x Time fixed effects Yes Yes Diagnostics Pseudo. R|$^{\mathrm{2}}$| / Adj. R|$^{\mathrm{2\thinspace }}$|(%) 3.16 7.98 N 10,127 10,127 This table estimates the effect of the cutoff on the availability of annual report data. Annual report data are culled from the DAFNE data base from Bureau van Dijk in the fiscal year prior to the loan application and the fiscal year following the year of the loan application. The dependent variables is equal to one if any of these two annual reports are missing or if the respective data item in any these two annual reports is missing. Column 1 provides results for firms’ balance sheet items (dependent variable is equal to one if any of the balance sheet items is missing), and Column 2 provides results for the number of employees (dependent variable is equal to one of the number of employees is missing). All models are estimated using a linear model. For variable definitions, see Table 1. T-values based on standard errors clustered at the branch level are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Table A2 Availability of data items: Test for discontinuity at the cutoff rating (1) (2) Dependent Model Change in firm balance sheet items is missing (0/1) Linear Change in the number of employees is missing (0/1) Linear Methodology Local regression |$\pm 2$| notches around cutoff Local regression |$\pm 2$| notches around cutoff Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Inference BelowCutOff (0/1) –0.005 (–0.44) 0.015 (0.76) Trends below/Above cutoff (Rating-CutOff) x BelowCutOff (0/1) –0.006 (–0.84) 0.006 (0.63) (Rating-CutOff) x (1- BelowCutOff (0/1)) 0.023** (2.56) 0.049*** (4.40) Firm controls Yes Yes Loan controls Yes Yes Industry x Time fixed effects Yes Yes Region x Time fixed effects Yes Yes Diagnostics Pseudo. R|$^{\mathrm{2}}$| / Adj. R|$^{\mathrm{2\thinspace }}$|(%) 3.16 7.98 N 10,127 10,127 (1) (2) Dependent Model Change in firm balance sheet items is missing (0/1) Linear Change in the number of employees is missing (0/1) Linear Methodology Local regression |$\pm 2$| notches around cutoff Local regression |$\pm 2$| notches around cutoff Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Inference BelowCutOff (0/1) –0.005 (–0.44) 0.015 (0.76) Trends below/Above cutoff (Rating-CutOff) x BelowCutOff (0/1) –0.006 (–0.84) 0.006 (0.63) (Rating-CutOff) x (1- BelowCutOff (0/1)) 0.023** (2.56) 0.049*** (4.40) Firm controls Yes Yes Loan controls Yes Yes Industry x Time fixed effects Yes Yes Region x Time fixed effects Yes Yes Diagnostics Pseudo. R|$^{\mathrm{2}}$| / Adj. R|$^{\mathrm{2\thinspace }}$|(%) 3.16 7.98 N 10,127 10,127 This table estimates the effect of the cutoff on the availability of annual report data. Annual report data are culled from the DAFNE data base from Bureau van Dijk in the fiscal year prior to the loan application and the fiscal year following the year of the loan application. The dependent variables is equal to one if any of these two annual reports are missing or if the respective data item in any these two annual reports is missing. Column 1 provides results for firms’ balance sheet items (dependent variable is equal to one if any of the balance sheet items is missing), and Column 2 provides results for the number of employees (dependent variable is equal to one of the number of employees is missing). 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 A3 McCrary density test Bandwidth Bin size Jump estimate (SE) A. Density at rating of 7.5 Standard bandwidth 1.132 0.026 –0.019 (0.0551) Undersmoothing 0.566 0.026 –0.041 (0.0783) B. Density at rating of 7.5, firms with available balance sheet data only Standard bandwidth 1.008 0.028 –0.015 (0.0631) Undersmoothing 0.504 0.028 0.003 (0.0906) Bandwidth Bin size Jump estimate (SE) A. Density at rating of 7.5 Standard bandwidth 1.132 0.026 –0.019 (0.0551) Undersmoothing 0.566 0.026 –0.041 (0.0783) B. Density at rating of 7.5, firms with available balance sheet data only Standard bandwidth 1.008 0.028 –0.015 (0.0631) Undersmoothing 0.504 0.028 0.003 (0.0906) This table provides results of a McCrary density test for the internal rating at the cutoff rating of 7.5. Panel A provides results for the sample of all loan applications. Panel B provides results for the sample of all loan applications with available balance sheet data. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Table A3 McCrary density test Bandwidth Bin size Jump estimate (SE) A. Density at rating of 7.5 Standard bandwidth 1.132 0.026 –0.019 (0.0551) Undersmoothing 0.566 0.026 –0.041 (0.0783) B. Density at rating of 7.5, firms with available balance sheet data only Standard bandwidth 1.008 0.028 –0.015 (0.0631) Undersmoothing 0.504 0.028 0.003 (0.0906) Bandwidth Bin size Jump estimate (SE) A. Density at rating of 7.5 Standard bandwidth 1.132 0.026 –0.019 (0.0551) Undersmoothing 0.566 0.026 –0.041 (0.0783) B. Density at rating of 7.5, firms with available balance sheet data only Standard bandwidth 1.008 0.028 –0.015 (0.0631) Undersmoothing 0.504 0.028 0.003 (0.0906) This table provides results of a McCrary density test for the internal rating at the cutoff rating of 7.5. Panel A provides results for the sample of all loan applications. Panel B provides results for the sample of all loan applications with available balance sheet data. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Table A4 Test for discontinuity in pre-application level and growth variables (1) (2) Total sample Small firms Table shows coefficients for the BelowCutOff (0/1)-dummy Coeff. |$t$|-stat Coeff. |$t$|-stat Levels, measured in the year prior to the loan application (for variables from the annual report) or based on initial loan application (for loan terms) log (Loan amount in EUR ’000)* –0.012 (–0.18) –0.111* (–1.74) Collateralized (0/1)* –0.044 (–1.60) –0.027 (–0.72) log (Firm age in years)** –0.028 (–0.84) –0.023 (–0.43) log (Relationship age in years)** –0.090 (–1.31) 0.068 (–0.93) log (Revenues in EUR)** 0.152** (2.13) 0.064 (0.73) log (Total assets in EUR)** –0.043 (–0.70) –0.253 (–0.46) log (Number of employees)** –0.050 (–0.98) –0.058 (–1.13) Equity-to-asset ratio** –0.003 (–0.27) –0.010 (–0.50) EBIT margin 0.003 (0.59) 0.005 (0.80) Liquidity 0.047 (1.08) 0.025 (0.33) Changes, measured based on changes from 3 years prior to the year prior to the loan application Asset growth –0.031 (–1.31) –0.022 (–0.71) Investment growth –0.013 (–1.30) –0.010 (–0.68) Employment growth 0.004 (0.17) 0.021 (0.56) Revenue growth –0.004 (–0.12) 0.010 (0.21) (1) (2) Total sample Small firms Table shows coefficients for the BelowCutOff (0/1)-dummy Coeff. |$t$|-stat Coeff. |$t$|-stat Levels, measured in the year prior to the loan application (for variables from the annual report) or based on initial loan application (for loan terms) log (Loan amount in EUR ’000)* –0.012 (–0.18) –0.111* (–1.74) Collateralized (0/1)* –0.044 (–1.60) –0.027 (–0.72) log (Firm age in years)** –0.028 (–0.84) –0.023 (–0.43) log (Relationship age in years)** –0.090 (–1.31) 0.068 (–0.93) log (Revenues in EUR)** 0.152** (2.13) 0.064 (0.73) log (Total assets in EUR)** –0.043 (–0.70) –0.253 (–0.46) log (Number of employees)** –0.050 (–0.98) –0.058 (–1.13) Equity-to-asset ratio** –0.003 (–0.27) –0.010 (–0.50) EBIT margin 0.003 (0.59) 0.005 (0.80) Liquidity 0.047 (1.08) 0.025 (0.33) Changes, measured based on changes from 3 years prior to the year prior to the loan application Asset growth –0.031 (–1.31) –0.022 (–0.71) Investment growth –0.013 (–1.30) –0.010 (–0.68) Employment growth 0.004 (0.17) 0.021 (0.56) Revenue growth –0.004 (–0.12) 0.010 (0.21) This table corresponds to Figure A1. It provides regression discontinuity tests for the control variables as a function of the continuous rating between January 2009 and December 2012. Column 1 provides results for the sample of all loan applications, and Column 2 provides results for the sample of loan applications by small firms. Panel A depicts variables in levels, and panel B depicts changes in selected variables from 3 years prior to the year prior to the loan application (with “the year prior” being the last observation before the loan application, for example, for a loan application in July 2012, panel B depicts changes from December 2009 to December 2011). The regression discontinuity design is based on the same methodology used in Tables 3–6 (local linear regression |$\pm 2$| notches around cutoff) but without any control variables apart from the running variable. For variable definitions, see Table 1. Table A4 Test for discontinuity in pre-application level and growth variables (1) (2) Total sample Small firms Table shows coefficients for the BelowCutOff (0/1)-dummy Coeff. |$t$|-stat Coeff. |$t$|-stat Levels, measured in the year prior to the loan application (for variables from the annual report) or based on initial loan application (for loan terms) log (Loan amount in EUR ’000)* –0.012 (–0.18) –0.111* (–1.74) Collateralized (0/1)* –0.044 (–1.60) –0.027 (–0.72) log (Firm age in years)** –0.028 (–0.84) –0.023 (–0.43) log (Relationship age in years)** –0.090 (–1.31) 0.068 (–0.93) log (Revenues in EUR)** 0.152** (2.13) 0.064 (0.73) log (Total assets in EUR)** –0.043 (–0.70) –0.253 (–0.46) log (Number of employees)** –0.050 (–0.98) –0.058 (–1.13) Equity-to-asset ratio** –0.003 (–0.27) –0.010 (–0.50) EBIT margin 0.003 (0.59) 0.005 (0.80) Liquidity 0.047 (1.08) 0.025 (0.33) Changes, measured based on changes from 3 years prior to the year prior to the loan application Asset growth –0.031 (–1.31) –0.022 (–0.71) Investment growth –0.013 (–1.30) –0.010 (–0.68) Employment growth 0.004 (0.17) 0.021 (0.56) Revenue growth –0.004 (–0.12) 0.010 (0.21) (1) (2) Total sample Small firms Table shows coefficients for the BelowCutOff (0/1)-dummy Coeff. |$t$|-stat Coeff. |$t$|-stat Levels, measured in the year prior to the loan application (for variables from the annual report) or based on initial loan application (for loan terms) log (Loan amount in EUR ’000)* –0.012 (–0.18) –0.111* (–1.74) Collateralized (0/1)* –0.044 (–1.60) –0.027 (–0.72) log (Firm age in years)** –0.028 (–0.84) –0.023 (–0.43) log (Relationship age in years)** –0.090 (–1.31) 0.068 (–0.93) log (Revenues in EUR)** 0.152** (2.13) 0.064 (0.73) log (Total assets in EUR)** –0.043 (–0.70) –0.253 (–0.46) log (Number of employees)** –0.050 (–0.98) –0.058 (–1.13) Equity-to-asset ratio** –0.003 (–0.27) –0.010 (–0.50) EBIT margin 0.003 (0.59) 0.005 (0.80) Liquidity 0.047 (1.08) 0.025 (0.33) Changes, measured based on changes from 3 years prior to the year prior to the loan application Asset growth –0.031 (–1.31) –0.022 (–0.71) Investment growth –0.013 (–1.30) –0.010 (–0.68) Employment growth 0.004 (0.17) 0.021 (0.56) Revenue growth –0.004 (–0.12) 0.010 (0.21) This table corresponds to Figure A1. It provides regression discontinuity tests for the control variables as a function of the continuous rating between January 2009 and December 2012. Column 1 provides results for the sample of all loan applications, and Column 2 provides results for the sample of loan applications by small firms. Panel A depicts variables in levels, and panel B depicts changes in selected variables from 3 years prior to the year prior to the loan application (with “the year prior” being the last observation before the loan application, for example, for a loan application in July 2012, panel B depicts changes from December 2009 to December 2011). The regression discontinuity design is based on the same methodology used in Tables 3–6 (local linear regression |$\pm 2$| notches around cutoff) but without any control variables apart from the running variable. For variable definitions, see Table 1. Table A5 Stickiness of rating grades and reapplication rates by rejected firms A. Likelihood of moving across the cutoff Rating 1 year after application date Rating at application date Unchanged (%) Improved (%) Deteriorated (%) % above cutoff 1 81.23 0.00 18.77 99.48 2 59.83 13.63 26.54 99.80 3 60.88 16.71 22.40 99.50 4 63.12 19.17 17.71 98.84 5 63.08 18.15 18.76 96.73 6 65.91 19.21 14.88 94.06 7 65.34 21.71 12.95 87.05 8 63.09 24.32 12.59 24.32 9 63.88 24.32 11.80 14.54 10 69.05 23.77 7.18 9.20 11 70.40 24.64 4.96 6.79 Average 64.50 19.49 16.02 77.16 A. Likelihood of moving across the cutoff Rating 1 year after application date Rating at application date Unchanged (%) Improved (%) Deteriorated (%) % above cutoff 1 81.23 0.00 18.77 99.48 2 59.83 13.63 26.54 99.80 3 60.88 16.71 22.40 99.50 4 63.12 19.17 17.71 98.84 5 63.08 18.15 18.76 96.73 6 65.91 19.21 14.88 94.06 7 65.34 21.71 12.95 87.05 8 63.09 24.32 12.59 24.32 9 63.88 24.32 11.80 14.54 10 69.05 23.77 7.18 9.20 11 70.40 24.64 4.96 6.79 Average 64.50 19.49 16.02 77.16 B. Reapplication rates Within 1 year Within 2 years All rejected firms (%) Narrowly rejected firms (Rating |$=$| 8) (%) All rejected firms (%) Narrowly rejected firms (Rating |$=$| 8) (%) All firm sizes Reapplication rate of unsuccessful applicants 10.24 9.15 14.20 10.70 Successful reapplication rate of unsuccessful applicants 6.51 7.26 9.05 8.71 Small firms (⁠|$=$|sample of Table 5 and following) Reapplication rate of unsuccessful applicants 10.48 8.83 14.76 11.93 Successful reapplication rate of unsuccessful applicants 6.46 6.84 8.96 9.63 B. Reapplication rates Within 1 year Within 2 years All rejected firms (%) Narrowly rejected firms (Rating |$=$| 8) (%) All rejected firms (%) Narrowly rejected firms (Rating |$=$| 8) (%) All firm sizes Reapplication rate of unsuccessful applicants 10.24 9.15 14.20 10.70 Successful reapplication rate of unsuccessful applicants 6.51 7.26 9.05 8.71 Small firms (⁠|$=$|sample of Table 5 and following) Reapplication rate of unsuccessful applicants 10.48 8.83 14.76 11.93 Successful reapplication rate of unsuccessful applicants 6.46 6.84 8.96 9.63 This table provides information on the stickiness of rating grades (panel A) and on reapplication rates by rejected firms (panel B). Panel A is based on all firms in the sample between 2009 and 2011; with rating grades 1 year after the application ranging from 2010 to 2012. Panel B is based on all rejected firms between 2009 and 2011 (2009–2010 for the 2-year window); with reapplications from 2010 to 2012. For variable definitions, see Table 1. Table A5 Stickiness of rating grades and reapplication rates by rejected firms A. Likelihood of moving across the cutoff Rating 1 year after application date Rating at application date Unchanged (%) Improved (%) Deteriorated (%) % above cutoff 1 81.23 0.00 18.77 99.48 2 59.83 13.63 26.54 99.80 3 60.88 16.71 22.40 99.50 4 63.12 19.17 17.71 98.84 5 63.08 18.15 18.76 96.73 6 65.91 19.21 14.88 94.06 7 65.34 21.71 12.95 87.05 8 63.09 24.32 12.59 24.32 9 63.88 24.32 11.80 14.54 10 69.05 23.77 7.18 9.20 11 70.40 24.64 4.96 6.79 Average 64.50 19.49 16.02 77.16 A. Likelihood of moving across the cutoff Rating 1 year after application date Rating at application date Unchanged (%) Improved (%) Deteriorated (%) % above cutoff 1 81.23 0.00 18.77 99.48 2 59.83 13.63 26.54 99.80 3 60.88 16.71 22.40 99.50 4 63.12 19.17 17.71 98.84 5 63.08 18.15 18.76 96.73 6 65.91 19.21 14.88 94.06 7 65.34 21.71 12.95 87.05 8 63.09 24.32 12.59 24.32 9 63.88 24.32 11.80 14.54 10 69.05 23.77 7.18 9.20 11 70.40 24.64 4.96 6.79 Average 64.50 19.49 16.02 77.16 B. Reapplication rates Within 1 year Within 2 years All rejected firms (%) Narrowly rejected firms (Rating |$=$| 8) (%) All rejected firms (%) Narrowly rejected firms (Rating |$=$| 8) (%) All firm sizes Reapplication rate of unsuccessful applicants 10.24 9.15 14.20 10.70 Successful reapplication rate of unsuccessful applicants 6.51 7.26 9.05 8.71 Small firms (⁠|$=$|sample of Table 5 and following) Reapplication rate of unsuccessful applicants 10.48 8.83 14.76 11.93 Successful reapplication rate of unsuccessful applicants 6.46 6.84 8.96 9.63 B. Reapplication rates Within 1 year Within 2 years All rejected firms (%) Narrowly rejected firms (Rating |$=$| 8) (%) All rejected firms (%) Narrowly rejected firms (Rating |$=$| 8) (%) All firm sizes Reapplication rate of unsuccessful applicants 10.24 9.15 14.20 10.70 Successful reapplication rate of unsuccessful applicants 6.51 7.26 9.05 8.71 Small firms (⁠|$=$|sample of Table 5 and following) Reapplication rate of unsuccessful applicants 10.48 8.83 14.76 11.93 Successful reapplication rate of unsuccessful applicants 6.46 6.84 8.96 9.63 This table provides information on the stickiness of rating grades (panel A) and on reapplication rates by rejected firms (panel B). Panel A is based on all firms in the sample between 2009 and 2011; with rating grades 1 year after the application ranging from 2010 to 2012. Panel B is based on all rejected firms between 2009 and 2011 (2009–2010 for the 2-year window); with reapplications from 2010 to 2012. For variable definitions, see Table 1. Table A6 Robustness test RDD: Different econometric approaches ROBUSTNESS FOR TABLE 3 Dependent variable Loan acceptance Change in loan volume with the bank Change in total debt (all banks and nonbanks) Change in equity (1) (2) (3) (4) (5) (6) Acceptance dummy (0/1) Time horizon: 3 months Time horizon: 12 months Time horizon: 24 months Fiscal year prior to loan application to fiscal year after loan application Fiscal year prior to loan application to fiscal year after loan application Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Methodology Coefficients for the BelowCutOff (0/1)-dummy Rectangular kernel (same as used Table 3) –0.280*** (–12.27) –0.062*** (–6.93) –0.060*** (–5.55) –0.070*** (–4.70) –0.071*** (–3.29) –0.015** (–2.50) Triangular kernel –0.291*** (–11.76) –0.056*** (–4.89) –0.053*** (–4.20) –0.068*** (–4.09) –0.076*** (–3.36) –0.017** (–2.64) Higher-order polynomial, degree 2 –0.306*** (–11.62) –0.055*** (–4.72) –0.054*** (–4.02) –0.074*** (–3.89) –0.075*** (–2.68) –0.022** (–2.21) Higher-order polynomial, degree 3 –0.309*** (–8.95) –0.060*** (–3.88) –0.045*** (–2.72) –0.067*** (–2.98) –0.087** (–2.16) –0.014 (–1.03) More granular industry definition –0.283*** (–12.57) –0.063*** (–7.02) –0.061*** (–5.40) –0.070*** (–4.51) –0.071*** (–3.34) –0.014** (–2.19) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}}$|⁠) –0.278*** (–12.40) –0.064*** (–6.91) –0.057*** (–5.58) –0.069*** (–4.91) –0.076*** (–4.04) –0.016** (–2.43) No fixed effects, but with other controls –0.297*** (–13.04) –0.065*** (–7.13) –0.061*** (–5.66) –0.072*** (–4.69) –0.077*** (–3.71) –0.017*** (–2.77) No controls apart from running variable –0.309*** (–14.03) –0.076*** (–7.62) –0.072*** (–6.02) –0.093*** (–5.66) –0.083*** (–3.42) –0.015** (–2.52) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{a}}$| –0.299*** (–10.03) –0.055*** (–4.04) –0.053*** (–3.76) –0.066*** (–3.57) –0.064** (–2.34) –0.015 (–1.59) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{a}}$| –0.313*** (–6.79) –0.072*** (–4.24) –0.056*** (–2.87) –0.088*** (–3.35) –0.052 (–1.17) –0.001 (–0.08) ROBUSTNESS FOR TABLE 3 Dependent variable Loan acceptance Change in loan volume with the bank Change in total debt (all banks and nonbanks) Change in equity (1) (2) (3) (4) (5) (6) Acceptance dummy (0/1) Time horizon: 3 months Time horizon: 12 months Time horizon: 24 months Fiscal year prior to loan application to fiscal year after loan application Fiscal year prior to loan application to fiscal year after loan application Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Methodology Coefficients for the BelowCutOff (0/1)-dummy Rectangular kernel (same as used Table 3) –0.280*** (–12.27) –0.062*** (–6.93) –0.060*** (–5.55) –0.070*** (–4.70) –0.071*** (–3.29) –0.015** (–2.50) Triangular kernel –0.291*** (–11.76) –0.056*** (–4.89) –0.053*** (–4.20) –0.068*** (–4.09) –0.076*** (–3.36) –0.017** (–2.64) Higher-order polynomial, degree 2 –0.306*** (–11.62) –0.055*** (–4.72) –0.054*** (–4.02) –0.074*** (–3.89) –0.075*** (–2.68) –0.022** (–2.21) Higher-order polynomial, degree 3 –0.309*** (–8.95) –0.060*** (–3.88) –0.045*** (–2.72) –0.067*** (–2.98) –0.087** (–2.16) –0.014 (–1.03) More granular industry definition –0.283*** (–12.57) –0.063*** (–7.02) –0.061*** (–5.40) –0.070*** (–4.51) –0.071*** (–3.34) –0.014** (–2.19) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}}$|⁠) –0.278*** (–12.40) –0.064*** (–6.91) –0.057*** (–5.58) –0.069*** (–4.91) –0.076*** (–4.04) –0.016** (–2.43) No fixed effects, but with other controls –0.297*** (–13.04) –0.065*** (–7.13) –0.061*** (–5.66) –0.072*** (–4.69) –0.077*** (–3.71) –0.017*** (–2.77) No controls apart from running variable –0.309*** (–14.03) –0.076*** (–7.62) –0.072*** (–6.02) –0.093*** (–5.66) –0.083*** (–3.42) –0.015** (–2.52) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{a}}$| –0.299*** (–10.03) –0.055*** (–4.04) –0.053*** (–3.76) –0.066*** (–3.57) –0.064** (–2.34) –0.015 (–1.59) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{a}}$| –0.313*** (–6.79) –0.072*** (–4.24) –0.056*** (–2.87) –0.088*** (–3.35) –0.052 (–1.17) –0.001 (–0.08) |$^{\mathrm{a}}$|The number of observations is 8,807 for the specifications using |$\pm$| 2 notches, 5,041 for |$\pm$| 1 notch and 2,618 for |$\pm$| 0.5 notches. Table A6 Robustness test RDD: Different econometric approaches ROBUSTNESS FOR TABLE 3 Dependent variable Loan acceptance Change in loan volume with the bank Change in total debt (all banks and nonbanks) Change in equity (1) (2) (3) (4) (5) (6) Acceptance dummy (0/1) Time horizon: 3 months Time horizon: 12 months Time horizon: 24 months Fiscal year prior to loan application to fiscal year after loan application Fiscal year prior to loan application to fiscal year after loan application Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Methodology Coefficients for the BelowCutOff (0/1)-dummy Rectangular kernel (same as used Table 3) –0.280*** (–12.27) –0.062*** (–6.93) –0.060*** (–5.55) –0.070*** (–4.70) –0.071*** (–3.29) –0.015** (–2.50) Triangular kernel –0.291*** (–11.76) –0.056*** (–4.89) –0.053*** (–4.20) –0.068*** (–4.09) –0.076*** (–3.36) –0.017** (–2.64) Higher-order polynomial, degree 2 –0.306*** (–11.62) –0.055*** (–4.72) –0.054*** (–4.02) –0.074*** (–3.89) –0.075*** (–2.68) –0.022** (–2.21) Higher-order polynomial, degree 3 –0.309*** (–8.95) –0.060*** (–3.88) –0.045*** (–2.72) –0.067*** (–2.98) –0.087** (–2.16) –0.014 (–1.03) More granular industry definition –0.283*** (–12.57) –0.063*** (–7.02) –0.061*** (–5.40) –0.070*** (–4.51) –0.071*** (–3.34) –0.014** (–2.19) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}}$|⁠) –0.278*** (–12.40) –0.064*** (–6.91) –0.057*** (–5.58) –0.069*** (–4.91) –0.076*** (–4.04) –0.016** (–2.43) No fixed effects, but with other controls –0.297*** (–13.04) –0.065*** (–7.13) –0.061*** (–5.66) –0.072*** (–4.69) –0.077*** (–3.71) –0.017*** (–2.77) No controls apart from running variable –0.309*** (–14.03) –0.076*** (–7.62) –0.072*** (–6.02) –0.093*** (–5.66) –0.083*** (–3.42) –0.015** (–2.52) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{a}}$| –0.299*** (–10.03) –0.055*** (–4.04) –0.053*** (–3.76) –0.066*** (–3.57) –0.064** (–2.34) –0.015 (–1.59) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{a}}$| –0.313*** (–6.79) –0.072*** (–4.24) –0.056*** (–2.87) –0.088*** (–3.35) –0.052 (–1.17) –0.001 (–0.08) ROBUSTNESS FOR TABLE 3 Dependent variable Loan acceptance Change in loan volume with the bank Change in total debt (all banks and nonbanks) Change in equity (1) (2) (3) (4) (5) (6) Acceptance dummy (0/1) Time horizon: 3 months Time horizon: 12 months Time horizon: 24 months Fiscal year prior to loan application to fiscal year after loan application Fiscal year prior to loan application to fiscal year after loan application Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Methodology Coefficients for the BelowCutOff (0/1)-dummy Rectangular kernel (same as used Table 3) –0.280*** (–12.27) –0.062*** (–6.93) –0.060*** (–5.55) –0.070*** (–4.70) –0.071*** (–3.29) –0.015** (–2.50) Triangular kernel –0.291*** (–11.76) –0.056*** (–4.89) –0.053*** (–4.20) –0.068*** (–4.09) –0.076*** (–3.36) –0.017** (–2.64) Higher-order polynomial, degree 2 –0.306*** (–11.62) –0.055*** (–4.72) –0.054*** (–4.02) –0.074*** (–3.89) –0.075*** (–2.68) –0.022** (–2.21) Higher-order polynomial, degree 3 –0.309*** (–8.95) –0.060*** (–3.88) –0.045*** (–2.72) –0.067*** (–2.98) –0.087** (–2.16) –0.014 (–1.03) More granular industry definition –0.283*** (–12.57) –0.063*** (–7.02) –0.061*** (–5.40) –0.070*** (–4.51) –0.071*** (–3.34) –0.014** (–2.19) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}}$|⁠) –0.278*** (–12.40) –0.064*** (–6.91) –0.057*** (–5.58) –0.069*** (–4.91) –0.076*** (–4.04) –0.016** (–2.43) No fixed effects, but with other controls –0.297*** (–13.04) –0.065*** (–7.13) –0.061*** (–5.66) –0.072*** (–4.69) –0.077*** (–3.71) –0.017*** (–2.77) No controls apart from running variable –0.309*** (–14.03) –0.076*** (–7.62) –0.072*** (–6.02) –0.093*** (–5.66) –0.083*** (–3.42) –0.015** (–2.52) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{a}}$| –0.299*** (–10.03) –0.055*** (–4.04) –0.053*** (–3.76) –0.066*** (–3.57) –0.064** (–2.34) –0.015 (–1.59) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{a}}$| –0.313*** (–6.79) –0.072*** (–4.24) –0.056*** (–2.87) –0.088*** (–3.35) –0.052 (–1.17) –0.001 (–0.08) |$^{\mathrm{a}}$|The number of observations is 8,807 for the specifications using |$\pm$| 2 notches, 5,041 for |$\pm$| 1 notch and 2,618 for |$\pm$| 0.5 notches. ROBUSTNESS FOR TABLE 5 A. Change in cash and cash equivalents B. Change in current assets (excluding cash and cash equivalents) (1) (2) (3) (4) (5) (6) Sample All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Methodology Rectangular kernel (same used in Table 5) –0.002 (–0.24) 0.025** (2.23) –0.032** (–2.48) –0.048* (–1.89) –0.034 (–1.07) –0.041 (–1.03) Triangular kernel –0.004 (–0.42) 0.024** (2.53) –0.035** (–2.00) –0.054* (–1.86) –0.039 (–1.14) –0.041 (–0.88) Higher-order polynomial, degree 2 0.001 (0.12) 0.034*** (2.77) –0.024 (–1.36) –0.050* (–1.90) –0.029 (–0.89) –0.048 (–1.16) Higher-order polynomial, degree 3 –0.012 (–0.90) 0.020 (1.41) –0.043* (–1.85) –0.055 (–1.51) –0.046 (–1.16) –0.046 (–0.77) More granular industry definition –0.004 (–0.36) 0.025** (2.32) –0.036* (–1.72) –0.045* (–1.70) –0.039 (–1.19) –0.024 (–0.57) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}})$| –0.002 (–0.21) 0.023** (2.20) –0.035* (–1.67) –0.059** (–2.17) –0.024 (–0.90) –0.067 (–1.58) No fixed effects, but with other controls –0.005 (–0.53) 0.027** (2.66) –0.041** (–2.20) –0.054** (–2.08) –0.043 (–1.35) –0.055 (–1.40) No controls apart from running variable –0.005 (–0.49) 0.031*** (3.02) –0.039** (–2.20) –0.056** (–2.16) –0.045 (–1.56) –0.060 (–1.29) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{b}}$| –0.008 (–0.58) 0.017 (1.36) –0.027 (–1.05) –0.057 (–1.56) –0.057 (–1.31) –0.022 (–0.42) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{b}}$| –0.004 (–0.26) 0.022* (1.93) –0.030 (–1.22) –0.038 (–1.48) –0.027 (–0.96) –0.032 (–0.75) ROBUSTNESS FOR TABLE 5 A. Change in cash and cash equivalents B. Change in current assets (excluding cash and cash equivalents) (1) (2) (3) (4) (5) (6) Sample All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Methodology Rectangular kernel (same used in Table 5) –0.002 (–0.24) 0.025** (2.23) –0.032** (–2.48) –0.048* (–1.89) –0.034 (–1.07) –0.041 (–1.03) Triangular kernel –0.004 (–0.42) 0.024** (2.53) –0.035** (–2.00) –0.054* (–1.86) –0.039 (–1.14) –0.041 (–0.88) Higher-order polynomial, degree 2 0.001 (0.12) 0.034*** (2.77) –0.024 (–1.36) –0.050* (–1.90) –0.029 (–0.89) –0.048 (–1.16) Higher-order polynomial, degree 3 –0.012 (–0.90) 0.020 (1.41) –0.043* (–1.85) –0.055 (–1.51) –0.046 (–1.16) –0.046 (–0.77) More granular industry definition –0.004 (–0.36) 0.025** (2.32) –0.036* (–1.72) –0.045* (–1.70) –0.039 (–1.19) –0.024 (–0.57) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}})$| –0.002 (–0.21) 0.023** (2.20) –0.035* (–1.67) –0.059** (–2.17) –0.024 (–0.90) –0.067 (–1.58) No fixed effects, but with other controls –0.005 (–0.53) 0.027** (2.66) –0.041** (–2.20) –0.054** (–2.08) –0.043 (–1.35) –0.055 (–1.40) No controls apart from running variable –0.005 (–0.49) 0.031*** (3.02) –0.039** (–2.20) –0.056** (–2.16) –0.045 (–1.56) –0.060 (–1.29) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{b}}$| –0.008 (–0.58) 0.017 (1.36) –0.027 (–1.05) –0.057 (–1.56) –0.057 (–1.31) –0.022 (–0.42) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{b}}$| –0.004 (–0.26) 0.022* (1.93) –0.030 (–1.22) –0.038 (–1.48) –0.027 (–0.96) –0.032 (–0.75) |$^{\mathrm{b}}$|The number of observations is 4,714 in the first column of each panel (2,649 for |$\pm$| 1 notch, 1,328 for |$\pm$| 0.5 notches), 2,279 in the second column (1,316 for |$\pm$| 1 notch, 691 for |$\pm$| 0.5 notches) and 2,435 in the last column of each panel (1,333 for |$\pm$| 1 notch, 637 for |$\pm$| 0.5 notches). ROBUSTNESS FOR TABLE 5 A. Change in cash and cash equivalents B. Change in current assets (excluding cash and cash equivalents) (1) (2) (3) (4) (5) (6) Sample All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Methodology Rectangular kernel (same used in Table 5) –0.002 (–0.24) 0.025** (2.23) –0.032** (–2.48) –0.048* (–1.89) –0.034 (–1.07) –0.041 (–1.03) Triangular kernel –0.004 (–0.42) 0.024** (2.53) –0.035** (–2.00) –0.054* (–1.86) –0.039 (–1.14) –0.041 (–0.88) Higher-order polynomial, degree 2 0.001 (0.12) 0.034*** (2.77) –0.024 (–1.36) –0.050* (–1.90) –0.029 (–0.89) –0.048 (–1.16) Higher-order polynomial, degree 3 –0.012 (–0.90) 0.020 (1.41) –0.043* (–1.85) –0.055 (–1.51) –0.046 (–1.16) –0.046 (–0.77) More granular industry definition –0.004 (–0.36) 0.025** (2.32) –0.036* (–1.72) –0.045* (–1.70) –0.039 (–1.19) –0.024 (–0.57) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}})$| –0.002 (–0.21) 0.023** (2.20) –0.035* (–1.67) –0.059** (–2.17) –0.024 (–0.90) –0.067 (–1.58) No fixed effects, but with other controls –0.005 (–0.53) 0.027** (2.66) –0.041** (–2.20) –0.054** (–2.08) –0.043 (–1.35) –0.055 (–1.40) No controls apart from running variable –0.005 (–0.49) 0.031*** (3.02) –0.039** (–2.20) –0.056** (–2.16) –0.045 (–1.56) –0.060 (–1.29) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{b}}$| –0.008 (–0.58) 0.017 (1.36) –0.027 (–1.05) –0.057 (–1.56) –0.057 (–1.31) –0.022 (–0.42) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{b}}$| –0.004 (–0.26) 0.022* (1.93) –0.030 (–1.22) –0.038 (–1.48) –0.027 (–0.96) –0.032 (–0.75) ROBUSTNESS FOR TABLE 5 A. Change in cash and cash equivalents B. Change in current assets (excluding cash and cash equivalents) (1) (2) (3) (4) (5) (6) Sample All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Methodology Rectangular kernel (same used in Table 5) –0.002 (–0.24) 0.025** (2.23) –0.032** (–2.48) –0.048* (–1.89) –0.034 (–1.07) –0.041 (–1.03) Triangular kernel –0.004 (–0.42) 0.024** (2.53) –0.035** (–2.00) –0.054* (–1.86) –0.039 (–1.14) –0.041 (–0.88) Higher-order polynomial, degree 2 0.001 (0.12) 0.034*** (2.77) –0.024 (–1.36) –0.050* (–1.90) –0.029 (–0.89) –0.048 (–1.16) Higher-order polynomial, degree 3 –0.012 (–0.90) 0.020 (1.41) –0.043* (–1.85) –0.055 (–1.51) –0.046 (–1.16) –0.046 (–0.77) More granular industry definition –0.004 (–0.36) 0.025** (2.32) –0.036* (–1.72) –0.045* (–1.70) –0.039 (–1.19) –0.024 (–0.57) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}})$| –0.002 (–0.21) 0.023** (2.20) –0.035* (–1.67) –0.059** (–2.17) –0.024 (–0.90) –0.067 (–1.58) No fixed effects, but with other controls –0.005 (–0.53) 0.027** (2.66) –0.041** (–2.20) –0.054** (–2.08) –0.043 (–1.35) –0.055 (–1.40) No controls apart from running variable –0.005 (–0.49) 0.031*** (3.02) –0.039** (–2.20) –0.056** (–2.16) –0.045 (–1.56) –0.060 (–1.29) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{b}}$| –0.008 (–0.58) 0.017 (1.36) –0.027 (–1.05) –0.057 (–1.56) –0.057 (–1.31) –0.022 (–0.42) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{b}}$| –0.004 (–0.26) 0.022* (1.93) –0.030 (–1.22) –0.038 (–1.48) –0.027 (–0.96) –0.032 (–0.75) |$^{\mathrm{b}}$|The number of observations is 4,714 in the first column of each panel (2,649 for |$\pm$| 1 notch, 1,328 for |$\pm$| 0.5 notches), 2,279 in the second column (1,316 for |$\pm$| 1 notch, 691 for |$\pm$| 0.5 notches) and 2,435 in the last column of each panel (1,333 for |$\pm$| 1 notch, 637 for |$\pm$| 0.5 notches). ROBUSTNESS FOR TABLE 6 Low liquidity High liquidity Sample All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat A. Asset growth B. Noncash asset growth (1) (2) (3) (4) (5) (6) Methodology Rectangular kernel (same used in Table 6) –0.078*** (–3.98) –0.068*** (–2.64) –0.074** (–2.17) –0.075*** (–3.58) –0.093*** (–3.92) –0.041 (–0.81) Triangular kernel –0.079*** (–3.59) –0.067** (–2.46) –0.071* (–1.83) –0.075*** (–3.17) –0.091*** (–3.38) –0.035 (–0.91) Higher-order polynomial, degree 2 –0.074*** (–3.47) –0.058** (–2.11) –0.075** (–2.18) –0.075*** (–3.63) –0.092*** (–3.59) –0.050 (–1.44) Higher-order polynomial, degree 3 –0.089*** (–2.92) –0.075** (–2.15) –0.084* (–1.71) –0.077** (–2.37) –0.094*** (–2.93) –0.041 (–0.84) More granular industry definition –0.070*** (–3.52) –0.070** (–2.56) –0.049 (–1.46) –0.067*** (–3.13) –0.095*** (–3.74) –0.014 (–0.36) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}})$| –0.080*** (–3.69) –0.066** (–2.56) –0.067* (–1.81) –0.077*** (–3.46) –0.090*** (–3.83) –0.032 (–0.79) No fixed effects, but with other controls –0.085*** (–4.29) –0.076*** (–2.99) –0.084*** (–2.64) –0.080*** (–3.69) –0.104*** (–4.04) –0.044 (–1.24) No controls apart from running variable –0.086*** (–4.20) –0.077*** (–3.25) –0.086** (–2.34) –0.081*** (–3.70) –0.108*** (–4.55) –0.047 (–1.21) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{c}}$| –0.088*** (–3.21) –0.091** (–2.51) –0.049 (–1.07) –0.080** (–2.56) –0.109*** (–2.88) –0.022 (–0.46) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{c}}$| –0.086*** (–2.85) –0.095** (–2.61) –0.077 (–1.32) –0.083** (–2.52) –0.117** (–3.18) –0.047 (–0.79) C. Change in investment D. Change in employment (7) (8) (9) (10) (11) (12) Methodology Rectangular kernel (same used in Table 6) –0.024** (–2.53) –0.044*** (–3.47) –0.009 (–0.62) –0.018 (–0.78) –0.070** (–2.18) 0.015 (0.39) Triangular kernel –0.029** (–2.59) –0.040** (–2.25) –0.016 (–0.90) –0.022 (–0.89) –0.075** (–2.14) 0.028 (0.75) Higher-order polynomial, degree 2 –0.019 (–1.51) –0.032* (–1.90) –0.015 (–0.78) –0.026 (–1.14) –0.087** (–2.32) 0.031 (0.80) Higher-order polynomial, degree 3 –0.033** (–2.21) –0.040* (–1.73) –0.026 (–1.17) –0.019 (–0.64) –0.055 (–1.21) 0.037 (0.70) More granular industry definition –0.022** (–2.26) –0.040** (–2.33) –0.000 (–0.02) –0.020 (–0.81) –0.059 (–1.62) 0.017 (0.40) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}})$| –0.024** (–2.35) –0.034** (–2.03) –0.009 (–0.43) –0.013 (–0.59) –0.062* (–1.94) 0.029 (0.89) No fixed effects, but with controls –0.031*** (–2.96) –0.046*** (–2.86) –0.016 (–0.99) 0.027 (–1.12) –0.072** (–2.23) 0.033 (0.75) No controls apart from running variable –0.037*** (–3.49) –0.047*** (–3.05) –0.023 (–1.41) –0.018 (–0.73) –0.061* (–1.86) 0.032 (0.73) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{d}}$| –0.036** (–2.04) –0.048* (–1.72) –0.025 (–0.97) –0.023 (–0.78) –0.089** (–2.00) 0.054 (1.08) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{d}}$| –0.068** (–2.64) –0.083** (–2.37) –0.057 (–1.39) –0.035 (–0.79) –0.057 (–0.87) 0.024 (0.40) ROBUSTNESS FOR TABLE 6 Low liquidity High liquidity Sample All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat A. Asset growth B. Noncash asset growth (1) (2) (3) (4) (5) (6) Methodology Rectangular kernel (same used in Table 6) –0.078*** (–3.98) –0.068*** (–2.64) –0.074** (–2.17) –0.075*** (–3.58) –0.093*** (–3.92) –0.041 (–0.81) Triangular kernel –0.079*** (–3.59) –0.067** (–2.46) –0.071* (–1.83) –0.075*** (–3.17) –0.091*** (–3.38) –0.035 (–0.91) Higher-order polynomial, degree 2 –0.074*** (–3.47) –0.058** (–2.11) –0.075** (–2.18) –0.075*** (–3.63) –0.092*** (–3.59) –0.050 (–1.44) Higher-order polynomial, degree 3 –0.089*** (–2.92) –0.075** (–2.15) –0.084* (–1.71) –0.077** (–2.37) –0.094*** (–2.93) –0.041 (–0.84) More granular industry definition –0.070*** (–3.52) –0.070** (–2.56) –0.049 (–1.46) –0.067*** (–3.13) –0.095*** (–3.74) –0.014 (–0.36) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}})$| –0.080*** (–3.69) –0.066** (–2.56) –0.067* (–1.81) –0.077*** (–3.46) –0.090*** (–3.83) –0.032 (–0.79) No fixed effects, but with other controls –0.085*** (–4.29) –0.076*** (–2.99) –0.084*** (–2.64) –0.080*** (–3.69) –0.104*** (–4.04) –0.044 (–1.24) No controls apart from running variable –0.086*** (–4.20) –0.077*** (–3.25) –0.086** (–2.34) –0.081*** (–3.70) –0.108*** (–4.55) –0.047 (–1.21) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{c}}$| –0.088*** (–3.21) –0.091** (–2.51) –0.049 (–1.07) –0.080** (–2.56) –0.109*** (–2.88) –0.022 (–0.46) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{c}}$| –0.086*** (–2.85) –0.095** (–2.61) –0.077 (–1.32) –0.083** (–2.52) –0.117** (–3.18) –0.047 (–0.79) C. Change in investment D. Change in employment (7) (8) (9) (10) (11) (12) Methodology Rectangular kernel (same used in Table 6) –0.024** (–2.53) –0.044*** (–3.47) –0.009 (–0.62) –0.018 (–0.78) –0.070** (–2.18) 0.015 (0.39) Triangular kernel –0.029** (–2.59) –0.040** (–2.25) –0.016 (–0.90) –0.022 (–0.89) –0.075** (–2.14) 0.028 (0.75) Higher-order polynomial, degree 2 –0.019 (–1.51) –0.032* (–1.90) –0.015 (–0.78) –0.026 (–1.14) –0.087** (–2.32) 0.031 (0.80) Higher-order polynomial, degree 3 –0.033** (–2.21) –0.040* (–1.73) –0.026 (–1.17) –0.019 (–0.64) –0.055 (–1.21) 0.037 (0.70) More granular industry definition –0.022** (–2.26) –0.040** (–2.33) –0.000 (–0.02) –0.020 (–0.81) –0.059 (–1.62) 0.017 (0.40) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}})$| –0.024** (–2.35) –0.034** (–2.03) –0.009 (–0.43) –0.013 (–0.59) –0.062* (–1.94) 0.029 (0.89) No fixed effects, but with controls –0.031*** (–2.96) –0.046*** (–2.86) –0.016 (–0.99) 0.027 (–1.12) –0.072** (–2.23) 0.033 (0.75) No controls apart from running variable –0.037*** (–3.49) –0.047*** (–3.05) –0.023 (–1.41) –0.018 (–0.73) –0.061* (–1.86) 0.032 (0.73) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{d}}$| –0.036** (–2.04) –0.048* (–1.72) –0.025 (–0.97) –0.023 (–0.78) –0.089** (–2.00) 0.054 (1.08) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{d}}$| –0.068** (–2.64) –0.083** (–2.37) –0.057 (–1.39) –0.035 (–0.79) –0.057 (–0.87) 0.024 (0.40) |$^{\mathrm{d}}$|The number of observations in panel C is 4,714 in the first column (2,649 for |$\pm$| 1 notch, 1,328 for |$\pm$| 0.5 notches), 2,279 in the second column (1,316 for |$\pm$| 1 notch, 691 for |$\pm$| 0.5 notches) and 2,435 in the last column of (1,333 for |$\pm$| 1 notch, 637 for |$\pm$| 0.5 notches). The number of observations in panel D is 3,295 in the first column (1,835 for |$\pm$| 1 notch, 941 for |$\pm$| 0.5 notches), 1,577 in the second column (922 for |$\pm$| 1 notch, 495 for |$\pm$| 0.5 notches) and 1,718 in the last column (913 for |$\pm$| 1 notch, 446 for |$\pm$| 0.5 notches). This table provides a robustness test for Tables 3, 5, and 6 using different econometric approaches. Econometric approaches include a rectangular kernel (like in Tables 3, 5, and 6), a triangular kernel, higher-order polynomials, different sets of controls, and a smaller bandwidth. Only the results for the main coefficient of interest, the BelowCutOff dummy, are reported below. Results are reported for the sample of all small firms (first column in each panel) and are split by the median of liquidity (measured as the ratio of current assets to current liabilities in the fiscal year prior to the loan application date). For variable definitions, see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. ROBUSTNESS FOR TABLE 6 Low liquidity High liquidity Sample All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat A. Asset growth B. Noncash asset growth (1) (2) (3) (4) (5) (6) Methodology Rectangular kernel (same used in Table 6) –0.078*** (–3.98) –0.068*** (–2.64) –0.074** (–2.17) –0.075*** (–3.58) –0.093*** (–3.92) –0.041 (–0.81) Triangular kernel –0.079*** (–3.59) –0.067** (–2.46) –0.071* (–1.83) –0.075*** (–3.17) –0.091*** (–3.38) –0.035 (–0.91) Higher-order polynomial, degree 2 –0.074*** (–3.47) –0.058** (–2.11) –0.075** (–2.18) –0.075*** (–3.63) –0.092*** (–3.59) –0.050 (–1.44) Higher-order polynomial, degree 3 –0.089*** (–2.92) –0.075** (–2.15) –0.084* (–1.71) –0.077** (–2.37) –0.094*** (–2.93) –0.041 (–0.84) More granular industry definition –0.070*** (–3.52) –0.070** (–2.56) –0.049 (–1.46) –0.067*** (–3.13) –0.095*** (–3.74) –0.014 (–0.36) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}})$| –0.080*** (–3.69) –0.066** (–2.56) –0.067* (–1.81) –0.077*** (–3.46) –0.090*** (–3.83) –0.032 (–0.79) No fixed effects, but with other controls –0.085*** (–4.29) –0.076*** (–2.99) –0.084*** (–2.64) –0.080*** (–3.69) –0.104*** (–4.04) –0.044 (–1.24) No controls apart from running variable –0.086*** (–4.20) –0.077*** (–3.25) –0.086** (–2.34) –0.081*** (–3.70) –0.108*** (–4.55) –0.047 (–1.21) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{c}}$| –0.088*** (–3.21) –0.091** (–2.51) –0.049 (–1.07) –0.080** (–2.56) –0.109*** (–2.88) –0.022 (–0.46) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{c}}$| –0.086*** (–2.85) –0.095** (–2.61) –0.077 (–1.32) –0.083** (–2.52) –0.117** (–3.18) –0.047 (–0.79) C. Change in investment D. Change in employment (7) (8) (9) (10) (11) (12) Methodology Rectangular kernel (same used in Table 6) –0.024** (–2.53) –0.044*** (–3.47) –0.009 (–0.62) –0.018 (–0.78) –0.070** (–2.18) 0.015 (0.39) Triangular kernel –0.029** (–2.59) –0.040** (–2.25) –0.016 (–0.90) –0.022 (–0.89) –0.075** (–2.14) 0.028 (0.75) Higher-order polynomial, degree 2 –0.019 (–1.51) –0.032* (–1.90) –0.015 (–0.78) –0.026 (–1.14) –0.087** (–2.32) 0.031 (0.80) Higher-order polynomial, degree 3 –0.033** (–2.21) –0.040* (–1.73) –0.026 (–1.17) –0.019 (–0.64) –0.055 (–1.21) 0.037 (0.70) More granular industry definition –0.022** (–2.26) –0.040** (–2.33) –0.000 (–0.02) –0.020 (–0.81) –0.059 (–1.62) 0.017 (0.40) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}})$| –0.024** (–2.35) –0.034** (–2.03) –0.009 (–0.43) –0.013 (–0.59) –0.062* (–1.94) 0.029 (0.89) No fixed effects, but with controls –0.031*** (–2.96) –0.046*** (–2.86) –0.016 (–0.99) 0.027 (–1.12) –0.072** (–2.23) 0.033 (0.75) No controls apart from running variable –0.037*** (–3.49) –0.047*** (–3.05) –0.023 (–1.41) –0.018 (–0.73) –0.061* (–1.86) 0.032 (0.73) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{d}}$| –0.036** (–2.04) –0.048* (–1.72) –0.025 (–0.97) –0.023 (–0.78) –0.089** (–2.00) 0.054 (1.08) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{d}}$| –0.068** (–2.64) –0.083** (–2.37) –0.057 (–1.39) –0.035 (–0.79) –0.057 (–0.87) 0.024 (0.40) ROBUSTNESS FOR TABLE 6 Low liquidity High liquidity Sample All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat A. Asset growth B. Noncash asset growth (1) (2) (3) (4) (5) (6) Methodology Rectangular kernel (same used in Table 6) –0.078*** (–3.98) –0.068*** (–2.64) –0.074** (–2.17) –0.075*** (–3.58) –0.093*** (–3.92) –0.041 (–0.81) Triangular kernel –0.079*** (–3.59) –0.067** (–2.46) –0.071* (–1.83) –0.075*** (–3.17) –0.091*** (–3.38) –0.035 (–0.91) Higher-order polynomial, degree 2 –0.074*** (–3.47) –0.058** (–2.11) –0.075** (–2.18) –0.075*** (–3.63) –0.092*** (–3.59) –0.050 (–1.44) Higher-order polynomial, degree 3 –0.089*** (–2.92) –0.075** (–2.15) –0.084* (–1.71) –0.077** (–2.37) –0.094*** (–2.93) –0.041 (–0.84) More granular industry definition –0.070*** (–3.52) –0.070** (–2.56) –0.049 (–1.46) –0.067*** (–3.13) –0.095*** (–3.74) –0.014 (–0.36) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}})$| –0.080*** (–3.69) –0.066** (–2.56) –0.067* (–1.81) –0.077*** (–3.46) –0.090*** (–3.83) –0.032 (–0.79) No fixed effects, but with other controls –0.085*** (–4.29) –0.076*** (–2.99) –0.084*** (–2.64) –0.080*** (–3.69) –0.104*** (–4.04) –0.044 (–1.24) No controls apart from running variable –0.086*** (–4.20) –0.077*** (–3.25) –0.086** (–2.34) –0.081*** (–3.70) –0.108*** (–4.55) –0.047 (–1.21) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{c}}$| –0.088*** (–3.21) –0.091** (–2.51) –0.049 (–1.07) –0.080** (–2.56) –0.109*** (–2.88) –0.022 (–0.46) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{c}}$| –0.086*** (–2.85) –0.095** (–2.61) –0.077 (–1.32) –0.083** (–2.52) –0.117** (–3.18) –0.047 (–0.79) C. Change in investment D. Change in employment (7) (8) (9) (10) (11) (12) Methodology Rectangular kernel (same used in Table 6) –0.024** (–2.53) –0.044*** (–3.47) –0.009 (–0.62) –0.018 (–0.78) –0.070** (–2.18) 0.015 (0.39) Triangular kernel –0.029** (–2.59) –0.040** (–2.25) –0.016 (–0.90) –0.022 (–0.89) –0.075** (–2.14) 0.028 (0.75) Higher-order polynomial, degree 2 –0.019 (–1.51) –0.032* (–1.90) –0.015 (–0.78) –0.026 (–1.14) –0.087** (–2.32) 0.031 (0.80) Higher-order polynomial, degree 3 –0.033** (–2.21) –0.040* (–1.73) –0.026 (–1.17) –0.019 (–0.64) –0.055 (–1.21) 0.037 (0.70) More granular industry definition –0.022** (–2.26) –0.040** (–2.33) –0.000 (–0.02) –0.020 (–0.81) –0.059 (–1.62) 0.017 (0.40) Controlling for past asset growth (t|$_{\mathrm{-3}} \to $| t|$_{\mathrm{-1}})$| –0.024** (–2.35) –0.034** (–2.03) –0.009 (–0.43) –0.013 (–0.59) –0.062* (–1.94) 0.029 (0.89) No fixed effects, but with controls –0.031*** (–2.96) –0.046*** (–2.86) –0.016 (–0.99) 0.027 (–1.12) –0.072** (–2.23) 0.033 (0.75) No controls apart from running variable –0.037*** (–3.49) –0.047*** (–3.05) –0.023 (–1.41) –0.018 (–0.73) –0.061* (–1.86) 0.032 (0.73) Half bandwidth (⁠|$\pm$| 1 notch)|$^{\mathrm{d}}$| –0.036** (–2.04) –0.048* (–1.72) –0.025 (–0.97) –0.023 (–0.78) –0.089** (–2.00) 0.054 (1.08) Quarter bandwidth (⁠|$\pm$| 0.5 notch)|$^{\mathrm{d}}$| –0.068** (–2.64) –0.083** (–2.37) –0.057 (–1.39) –0.035 (–0.79) –0.057 (–0.87) 0.024 (0.40) |$^{\mathrm{d}}$|The number of observations in panel C is 4,714 in the first column (2,649 for |$\pm$| 1 notch, 1,328 for |$\pm$| 0.5 notches), 2,279 in the second column (1,316 for |$\pm$| 1 notch, 691 for |$\pm$| 0.5 notches) and 2,435 in the last column of (1,333 for |$\pm$| 1 notch, 637 for |$\pm$| 0.5 notches). The number of observations in panel D is 3,295 in the first column (1,835 for |$\pm$| 1 notch, 941 for |$\pm$| 0.5 notches), 1,577 in the second column (922 for |$\pm$| 1 notch, 495 for |$\pm$| 0.5 notches) and 1,718 in the last column (913 for |$\pm$| 1 notch, 446 for |$\pm$| 0.5 notches). This table provides a robustness test for Tables 3, 5, and 6 using different econometric approaches. Econometric approaches include a rectangular kernel (like in Tables 3, 5, and 6), a triangular kernel, higher-order polynomials, different sets of controls, and a smaller bandwidth. Only the results for the main coefficient of interest, the BelowCutOff dummy, are reported below. Results are reported for the sample of all small firms (first column in each panel) and are split by the median of liquidity (measured as the ratio of current assets to current liabilities in the fiscal year prior to the loan application date). For variable definitions, see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Table A7 Robustness test RDD: Four bins by liquidity (1) (2) (3) (4) (5) Sample All firms Q1 (highest liquidity) Q2 Q3 Q4 (lowest liquidity) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Dependent variable Variables where results in Table 5/Table 6 are decreasing in liquidity Cash holdings –0.002 (–0.24) –0.059** (–2.42) 0.008 (0.36) 0.036* (1.86) 0.015 (1.15) Variables where results in Table 5/Table 6 are decreasing in liquidity Noncash asset growth –0.075*** (–3.58) –0.044 (–0.62) –0.017 (–0.39) –0.080* (–1.84) –0.097** (–2.51) Investment –0.024** (–2.53) –0.033 (–1.29) 0.004 (0.29) –0.028 (–1.38) –0.058*** (–2.73) Employment –0.018 (–0.78) 0.073 (1.31) 0.015 (0.37) –0.013 (–0.27) –0.094** (–2.02) (1) (2) (3) (4) (5) Sample All firms Q1 (highest liquidity) Q2 Q3 Q4 (lowest liquidity) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Dependent variable Variables where results in Table 5/Table 6 are decreasing in liquidity Cash holdings –0.002 (–0.24) –0.059** (–2.42) 0.008 (0.36) 0.036* (1.86) 0.015 (1.15) Variables where results in Table 5/Table 6 are decreasing in liquidity Noncash asset growth –0.075*** (–3.58) –0.044 (–0.62) –0.017 (–0.39) –0.080* (–1.84) –0.097** (–2.51) Investment –0.024** (–2.53) –0.033 (–1.29) 0.004 (0.29) –0.028 (–1.38) –0.058*** (–2.73) Employment –0.018 (–0.78) 0.073 (1.31) 0.015 (0.37) –0.013 (–0.27) –0.094** (–2.02) This table provides a robustness test for Tables 5 and 6 using four bins by liquidity (Q1-Q4) instead of just two bins by liquidity (high/low liquidity). Only the results for the main coefficient of interest, the BelowCutOff dummy, are reported below, and the econometric specification is the same as that used in Tables 5 and 6. Results are reported for all columns where Table 6 and 6 reveal a significant difference between high- and low-liquidity firms. For variable definitions see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Table A7 Robustness test RDD: Four bins by liquidity (1) (2) (3) (4) (5) Sample All firms Q1 (highest liquidity) Q2 Q3 Q4 (lowest liquidity) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Dependent variable Variables where results in Table 5/Table 6 are decreasing in liquidity Cash holdings –0.002 (–0.24) –0.059** (–2.42) 0.008 (0.36) 0.036* (1.86) 0.015 (1.15) Variables where results in Table 5/Table 6 are decreasing in liquidity Noncash asset growth –0.075*** (–3.58) –0.044 (–0.62) –0.017 (–0.39) –0.080* (–1.84) –0.097** (–2.51) Investment –0.024** (–2.53) –0.033 (–1.29) 0.004 (0.29) –0.028 (–1.38) –0.058*** (–2.73) Employment –0.018 (–0.78) 0.073 (1.31) 0.015 (0.37) –0.013 (–0.27) –0.094** (–2.02) (1) (2) (3) (4) (5) Sample All firms Q1 (highest liquidity) Q2 Q3 Q4 (lowest liquidity) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Dependent variable Variables where results in Table 5/Table 6 are decreasing in liquidity Cash holdings –0.002 (–0.24) –0.059** (–2.42) 0.008 (0.36) 0.036* (1.86) 0.015 (1.15) Variables where results in Table 5/Table 6 are decreasing in liquidity Noncash asset growth –0.075*** (–3.58) –0.044 (–0.62) –0.017 (–0.39) –0.080* (–1.84) –0.097** (–2.51) Investment –0.024** (–2.53) –0.033 (–1.29) 0.004 (0.29) –0.028 (–1.38) –0.058*** (–2.73) Employment –0.018 (–0.78) 0.073 (1.31) 0.015 (0.37) –0.013 (–0.27) –0.094** (–2.02) This table provides a robustness test for Tables 5 and 6 using four bins by liquidity (Q1-Q4) instead of just two bins by liquidity (high/low liquidity). Only the results for the main coefficient of interest, the BelowCutOff dummy, are reported below, and the econometric specification is the same as that used in Tables 5 and 6. Results are reported for all columns where Table 6 and 6 reveal a significant difference between high- and low-liquidity firms. For variable definitions see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Table A8 Holdout sample of loans > EUR 1 million Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Loan acceptance Change in loan volume with the bank Change in total debt (all banks and nonbanks) Change in equity Acceptance dummy (0/1) Time horizon: 3 months Time horizon: 12 months Time horizon: 24 months Fiscal year prior to loan application to fiscal year after loan application Fiscal year prior to loan application to fiscal year after loan application BelowCutOff (0/1) – Table 3 –0.280*** (–12.27) –0.062*** (–6.93) –0.060*** (–5.55) –0.070*** (–4.70) –0.071*** (–3.29) –0.015** (–2.50) BelowCutOff(0/1) – Volume > EUR 1 mil 0.038 (0.85) 0.025 (0.65) –0.010 (–0.19) –0.002 (–0.03) 0.086 (1.43) –0.025 (–1.57) Table 5: Cash holdings A. Change in cash and cash equivalents B. Change in current assets (excluding cash and cash equivalents) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) BelowCutOff (0/1) – Table 5 –0.002 (–0.24) 0.025** (2.23) –0.032** (–2.48) –0.048* (–1.89) –0.034 (–1.07) –0.041 (–1.03) BelowCutOff(0/1) – Volume > EUR 1 mil 0.012 (0.97) –0.003 (–0.26) 0.034 (1.07) 0.048 (0.91) 0.026 (0.41) 0.038 (0.46) Table 6: Real effects A. Asset growth B. Noncash asset growth All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) BelowCutOff (0/1) – Table 6 –0.078*** (–3.98) –0.068*** (–2.64) –0.074** (–2.17) –0.075*** (–3.58) –0.093*** (–3.92) –0.041 (–0.81) BelowCutOff(0/1) – Volume > EUR 1 mil 0.034 (0.72) 0.011 (0.21) 0.032 (0.41) 0.013 (0.26) 0.010 (0.20) –0.013 (–0.18) C. Change in investment D: Change in employment All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) BelowCutOff (0/1) – Table 6 –0.024** (–2.53) –0.044*** (–3.47) –0.009 (–0.62) –0.018 (–0.78) –0.070** (–2.18) 0.015 (0.39) BelowCutOff(0/1) – Volume > EUR 1 mil –0.016 (–0.54) –0.026 (–0.79) –0.014 (–0.28) 0.078* (1.81) 0.101* (1.84) 0.037 (0.61) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Loan acceptance Change in loan volume with the bank Change in total debt (all banks and nonbanks) Change in equity Acceptance dummy (0/1) Time horizon: 3 months Time horizon: 12 months Time horizon: 24 months Fiscal year prior to loan application to fiscal year after loan application Fiscal year prior to loan application to fiscal year after loan application BelowCutOff (0/1) – Table 3 –0.280*** (–12.27) –0.062*** (–6.93) –0.060*** (–5.55) –0.070*** (–4.70) –0.071*** (–3.29) –0.015** (–2.50) BelowCutOff(0/1) – Volume > EUR 1 mil 0.038 (0.85) 0.025 (0.65) –0.010 (–0.19) –0.002 (–0.03) 0.086 (1.43) –0.025 (–1.57) Table 5: Cash holdings A. Change in cash and cash equivalents B. Change in current assets (excluding cash and cash equivalents) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) BelowCutOff (0/1) – Table 5 –0.002 (–0.24) 0.025** (2.23) –0.032** (–2.48) –0.048* (–1.89) –0.034 (–1.07) –0.041 (–1.03) BelowCutOff(0/1) – Volume > EUR 1 mil 0.012 (0.97) –0.003 (–0.26) 0.034 (1.07) 0.048 (0.91) 0.026 (0.41) 0.038 (0.46) Table 6: Real effects A. Asset growth B. Noncash asset growth All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) BelowCutOff (0/1) – Table 6 –0.078*** (–3.98) –0.068*** (–2.64) –0.074** (–2.17) –0.075*** (–3.58) –0.093*** (–3.92) –0.041 (–0.81) BelowCutOff(0/1) – Volume > EUR 1 mil 0.034 (0.72) 0.011 (0.21) 0.032 (0.41) 0.013 (0.26) 0.010 (0.20) –0.013 (–0.18) C. Change in investment D: Change in employment All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) BelowCutOff (0/1) – Table 6 –0.024** (–2.53) –0.044*** (–3.47) –0.009 (–0.62) –0.018 (–0.78) –0.070** (–2.18) 0.015 (0.39) BelowCutOff(0/1) – Volume > EUR 1 mil –0.016 (–0.54) –0.026 (–0.79) –0.014 (–0.28) 0.078* (1.81) 0.101* (1.84) 0.037 (0.61) This table reports results from a holdout sample of loan applications larger than EUR 1 million in which the bank does not use the same cutoff as for the main sample. This table uses exactly the same RDD specifications used in Tables 3, 5, and 6 but only reports the key coefficient of interests. For ease of comparison, this table also reports the original coefficient estimate from Tables 3, 5, and 6. For variable definitions see Table 1. T-values, based on standard errors clustered at the branch level, are reported in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Table A8 Holdout sample of loans > EUR 1 million Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Loan acceptance Change in loan volume with the bank Change in total debt (all banks and nonbanks) Change in equity Acceptance dummy (0/1) Time horizon: 3 months Time horizon: 12 months Time horizon: 24 months Fiscal year prior to loan application to fiscal year after loan application Fiscal year prior to loan application to fiscal year after loan application BelowCutOff (0/1) – Table 3 –0.280*** (–12.27) –0.062*** (–6.93) –0.060*** (–5.55) –0.070*** (–4.70) –0.071*** (–3.29) –0.015** (–2.50) BelowCutOff(0/1) – Volume > EUR 1 mil 0.038 (0.85) 0.025 (0.65) –0.010 (–0.19) –0.002 (–0.03) 0.086 (1.43) –0.025 (–1.57) Table 5: Cash holdings A. Change in cash and cash equivalents B. Change in current assets (excluding cash and cash equivalents) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) BelowCutOff (0/1) – Table 5 –0.002 (–0.24) 0.025** (2.23) –0.032** (–2.48) –0.048* (–1.89) –0.034 (–1.07) –0.041 (–1.03) BelowCutOff(0/1) – Volume > EUR 1 mil 0.012 (0.97) –0.003 (–0.26) 0.034 (1.07) 0.048 (0.91) 0.026 (0.41) 0.038 (0.46) Table 6: Real effects A. Asset growth B. Noncash asset growth All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) BelowCutOff (0/1) – Table 6 –0.078*** (–3.98) –0.068*** (–2.64) –0.074** (–2.17) –0.075*** (–3.58) –0.093*** (–3.92) –0.041 (–0.81) BelowCutOff(0/1) – Volume > EUR 1 mil 0.034 (0.72) 0.011 (0.21) 0.032 (0.41) 0.013 (0.26) 0.010 (0.20) –0.013 (–0.18) C. Change in investment D: Change in employment All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) BelowCutOff (0/1) – Table 6 –0.024** (–2.53) –0.044*** (–3.47) –0.009 (–0.62) –0.018 (–0.78) –0.070** (–2.18) 0.015 (0.39) BelowCutOff(0/1) – Volume > EUR 1 mil –0.016 (–0.54) –0.026 (–0.79) –0.014 (–0.28) 0.078* (1.81) 0.101* (1.84) 0.037 (0.61) Parameter Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Coeff. |$t$|-stat Loan acceptance Change in loan volume with the bank Change in total debt (all banks and nonbanks) Change in equity Acceptance dummy (0/1) Time horizon: 3 months Time horizon: 12 months Time horizon: 24 months Fiscal year prior to loan application to fiscal year after loan application Fiscal year prior to loan application to fiscal year after loan application BelowCutOff (0/1) – Table 3 –0.280*** (–12.27) –0.062*** (–6.93) –0.060*** (–5.55) –0.070*** (–4.70) –0.071*** (–3.29) –0.015** (–2.50) BelowCutOff(0/1) – Volume > EUR 1 mil 0.038 (0.85) 0.025 (0.65) –0.010 (–0.19) –0.002 (–0.03) 0.086 (1.43) –0.025 (–1.57) Table 5: Cash holdings A. Change in cash and cash equivalents B. Change in current assets (excluding cash and cash equivalents) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) BelowCutOff (0/1) – Table 5 –0.002 (–0.24) 0.025** (2.23) –0.032** (–2.48) –0.048* (–1.89) –0.034 (–1.07) –0.041 (–1.03) BelowCutOff(0/1) – Volume > EUR 1 mil 0.012 (0.97) –0.003 (–0.26) 0.034 (1.07) 0.048 (0.91) 0.026 (0.41) 0.038 (0.46) Table 6: Real effects A. Asset growth B. Noncash asset growth All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) BelowCutOff (0/1) – Table 6 –0.078*** (–3.98) –0.068*** (–2.64) –0.074** (–2.17) –0.075*** (–3.58) –0.093*** (–3.92) –0.041 (–0.81) BelowCutOff(0/1) – Volume > EUR 1 mil 0.034 (0.72) 0.011 (0.21) 0.032 (0.41) 0.013 (0.26) 0.010 (0.20) –0.013 (–0.18) C. Change in investment D: Change in employment All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) All firms Low liquidity (CA/CL |$\leqslant $| 1.4) High liquidity (CA/CL > 1.4) BelowCutOff (0/1) – Table 6 –0.024** (–2.53) –0.044*** (–3.47) –0.009 (–0.62) –0.018 (–0.78) –0.070** (–2.18) 0.015 (0.39) BelowCutOff(0/1) – Volume > EUR 1 mil –0.016 (–0.54) –0.026 (–0.79) –0.014 (–0.28) 0.078* (1.81) 0.101* (1.84) 0.037 (0.61) This table reports results from a holdout sample of loan applications larger than EUR 1 million in which the bank does not use the same cutoff as for the main sample. This table uses exactly the same RDD specifications used in Tables 3, 5, and 6 but only reports the key coefficient of interests. For ease of comparison, this table also reports the original coefficient estimate from Tables 3, 5, and 6. 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. I wish to thank the editor, Philip Strahan, and two anonymous referees for comments that improved the quality of the paper significantly. I also wish to thank Heitor Almeida, Narly Dwarkasing, Reint Gropp, Andre Guettler, Martin Hellwig, Rainer Haselmann, Daniel Paravisini, Benedikt Ruprecht, Larissa Schäfer, Enrico Sette, Sascha Steffen, Daniel Streitz, Sebastian Pfeil, and Burcin Yurtoglu; an anonymous referee for the ECB Working Paper Series; and the participants of the 2016 AFA meetings in San Francisco, the 2016 EFA meetings in Oslo, the 2016 FIRS meetings in Lisbon, the 2015 German Finance Association (DGF) Meetings, the 2015 Banking Workshop in Muenster, the Swiss Winter Finance Conference in Lenzerheide, and research seminars at the University of Bonn, IWH Halle, WHU, University of Frankfurt, Ulm University, Tilburg University, Frankfurt School of Finance & Management, and Copenhagen Business School for valuable comments and suggestions. This work was supported by the Lamfalussy Fellowship Program sponsored by the ECB. Any views expressed are only those of the author and do not necessarily represent the views of the ECB or the Eurosystem. Footnotes 1 Manipulation of the internal rating might be another reason firms just below and just above the cutoff are not similar. For example, if ratings of firms with a positive outlook would be revised by the loan officer to a rating slightly above the cutoff, then firms slightly below and slightly above the cutoff are no longer comparable. Loan officers are not compensated based on loan volume but on ex post loan performance, and the rating does not include soft information. Thus, there are no incentives to manipulate the rating (like in, e.g., Berg, Puri, and Rocholl 2017), and there is no (conscious or unconscious) influence of soft information. A formal McCrary density test confirms the no-manipulation assumption. 2 For loan applications above EUR 1 million, there is no cutoff rule, and an additional review by a credit officer is required for each application independent of the rating. Dropping loan applications with a loan volume of exactly EUR 1 million, which might be strategically chosen as to avoid the additional review, does not significantly alter any of the following results. 3 Firms with unlimited liability of the owners (typically sole proprietorships, such as self-employed consultants, architects, or physicians) are handled in a different segment (private customers) by the bank. Financial firms (banks and insurance companies) are handled in a separate segment as well. 4 See Udell (1989) and Berg (2015) for a detailed description of the loan review function of risk management. 5 The data at hand allow me to differentiate between outright rejections and reductions in the loan volume granted (vis-à-vis the loan volume applied for) by the risk management department. However, the decision to reject versus reduce loan volume is likely endogenous (with firms with a poor outlook likely being rejected) and therefore does not allow for a credible distinction between the extensive and the intensive margin. Data on changes to other loan terms, such as maturity or covenants, are not available. 6 Ruckes (2004) and Bubb and Kaufman (2014) provide a theoretical motivation for the use of such cutoff rules in the loan application. The key argument in both papers is that the lender must bear a fixed cost per applicant for the additional screening process by the risk management department. This fixed-cost assumption implies that the additional information outweighs the fixed costs only for loan applicants below a particular threshold of the hard-information rating. 7 Loan officers can reject loan applications for ratings between 1 and 7. Discussions with loan officers suggest that these rejections are mainly due to technical reasons: after entering the applicant’s data into the system, a loan officer would communicate the terms and conditions of the loan offer to the client if the rating is in the 1–7 range. If an applicant directly decides not to take up the loan offer, most loan officers hit the “reject” button in the loan application system instead of formally making a loan offer. The identification strategy is unaffected by this fact as I only use a BelowCutOff dummy, but not the accept/reject decision itself, in the empirical section. 8 The common identifier is the Creditreform ID. Creditreform is the dominant private credit registry for firms in Germany, and therefore both the bank and Bureau van Dijk have this item available. 9 In addition to these main items, firms need to disclose deferred tax assets, deferred tax liabilities, and accruals. 10 An exception is the variable noncash asset growth. For this variable, I define the winsorized version as the winsorized version of asset growth minus the winsorized version of the change in cash and cash equivalents scaled by lagged total assets. This ensures that the balance sheet identity cash|$+$|noncash assets|$=$|total assets also holds for the winsorized version, that is, winsorized cash growth |$+$| winsorized noncash asset growth |$=$| winsorized asset growth. 11 The correlation between the data collected by the bank and Bureau van Dijk’s data exceeds 95% for all characteristics in the sample of firms where both characteristics are available. 12 Please note that (1) is a cross-sectional regression and not a panel regression and the subindex |$t$| simply denotes the time of the loan application. Each loan application constitutes one observation in Equation (1). Section 3.5 presents a difference-in-differences panel data specification. 13 The rule-of-thumb bandwidth selector trades off bias versus precision. Ceteris paribus, it therefore calls for a larger bandwidth if precision is low and lower bandwidth if precision is high. Compared to the fixed bandwidth that I use, regression-specific bandwidths are therefore more “socialistic”: strong results—where the precision of the estimate is high and the bandwidth selector therefore suggests a low bandwidth—get weaker and weak results—where the precision of the estimate is low and the selector therefore suggests a wider bandwidth—get stronger. 14 The EIBT margin is highly correlated with the return on assets (RoA). Using the return on assets instead of the EBIT margin provides very similar results in all of the following tables. 15 Firms are grouped into 14 different industries (agriculture, building and construction, consulting, retail sales, wholesale trade, health care, hotels/restaurants/travel, IT, manufacturing, median/publishing/education, services, utilities, chemical and pharmaceutical industry, and other). The three largest industries in the sample are manufacturing (4,622 observations), wholesale trade (3,730 observations), chemical and pharmaceutical industry (1,045 observations), and services (1,028 observations). 16 The bank has approximately 100 branches. Clustering by branches accounts for both regional correlations among borrowers and for branch-specific traits of the banking organization. 17 Furthermore, loan officers are not volume incentivized so that any incentives to manipulate hard information as documented in Berg, Puri, and Rocholl (2017) are muted. Sorting by firms is also not likely, because this would require firms to know exactly whether they have a rating just right or just left of the threshold. This knowledge is unlikely given that firms do not have access to the exact formula behind the bank’s internal rating. 18 Please note that other loan characteristics, such as loan type, loan purpose, or loan pricing information, are not available. I report the loan amount and a collateral dummy in Figure A1, and don’t find any evidence for a discontinuity in these loan terms. There is also no evidence of manipulation of the rating; see the McCrary test in Table A3. It seems plausible to assume that other loan characteristics are also similar across the cutoff. 19 Rejected firms that successfully reapply for a loan in the next period would attenuate the results of the comparison of below versus above-cutoff firms. Table A5 provides a direct test by showing that ratings are slowly moving, and successful reapplications are rare in my sample. Any attenuation bias caused by successful reapplications is therefore likely to be small. 20 As a comparison, other researchers who use Bureau van Dijk’s database for SMEs report median assets of EUR 3.42 million (Popov and Rocholl (Forthcoming)) and EUR 0.73 million (De Marco (Forthcoming)). Median assets in this paper are EUR 2.58 million for the total sample and EUR 1.45 million for the sample of small firms, values that are therefore broadly in line with prior research on small- and medium-sized enterprises. 21 The change in lending standard is consistent with the change in lending standards reported by German SMEs in ECB’s Survey on the Access to Finance of Enterprises, see http://sdw.ecb.europa.eu/quickview.do?SERIES_KEY=235.SAFE.H.DE.SME.A.0.0.0.Q7B.FBLN.S4.AL.WP&resetSettings.x=0&resetSettings.y=0&start=&end=&trans=N. 22 In contrast to the RDD, I do not include further control variables (loan and firm characteristics). Firm characteristics in the year prior to the loan application are absorbed by firm fixed effects. Firm characteristics in the years after the loan applications obviously constitute bad controls. Loan characteristics are only observed in the year of the loan application and are therefore also absorbed by firm fixed effects. 23 Loan applications are from 2009 to 2012. Annual report data are available from 2007 until 2014, implying that data for the 2 years before and after the loan application is the longest time period that I can consistently analyze. Results using |$\pm 2$| notches around the threshold are usually significantly stronger than the results reported in Table 7. 24 Loan officers are aware that there is no longer a clear cutoff in the likelihood of rejections in this holdout sample of larger loans, and loan officer incentives to manipulate ratings can therefore differ in the holdout sample. However, concerns about differences in loan officer behavior are somehow mitigated by the fact that the distribution of ratings for the holdout sample follows a very similar pattern as that in the main sample of the paper. References Acharya, V., Almeida, H. and Campello. M. 2007 . Is cash negative debt? A hedging perspective on corporate financial policies. Journal of Financial Intermediation 16 : 515 – 54 . Google Scholar Crossref Search ADS Acharya, V., Eisert, T. Eufinger, C. and Hirsch. C. Forthcoming . Real effects of the sovereign debt crisis in Europe: Evidence from syndicated loans. Review of Financial Studies . Almeida, H., and Campello. M. 2001 . Financial constraints and investment-cash flow sensitivities: New research directions. Working Paper . Almeida, H., Campello, M. and Weisbach. M. 2004 . The cash flow sensitivity of cash. Journal of Finance 59 : 1777 – 804 . Google Scholar Crossref Search ADS Almeida, H., Campello, M. Laranjeira, B. and Weisbenner. S. 2012 . Corporate debt maturity and the real effects of the 2007 credit crisis. Critical Finance Review 1 : 3 – 58 . Google Scholar Crossref Search ADS Balduzzi, P., Brancati, E. and Schiantarelli. F. 2017 . Financial markets, banks’ cost of funding, and firms’ decisions: lessons from two crises. Journal of Financial Intermediation . Advance Access published September 28 , 2017 , https://doi.org/10.1016/j.jfi.2017.09.004 . Banerjee, A. V., and Duflo. E. 2014 . Do firms want to borrow more? Testing credit constraints using a directed lending program. Review of Economic Studies 81 : 572 – 607 . Google Scholar Crossref Search ADS Bates, T. W., Kahle, K. M. and Stulz. R.M. 2009 . Why do US firms hold so much more cash than they used to? Journal of Finance 64 : 1985 – 2021 . Google Scholar Crossref Search ADS Becker, B., and Ivashina. V. 2014 . Cyclicality of credit supply: Firm level evidence. Journal of Monetary Economics 62 : 76 – 93 . Google Scholar Crossref Search ADS Bentolila, S., Jansen, M. Jiminéz, G. and Ruano. S. 2017 . When credit dries up: Job losses in the great recession. Journal of the European Economic Association . Advance Access published August 17 , 2017 , https://doi.org/10.1093/jeea/jvx021 . Berg, T. 2015 . Playing the devil’s advocate: The causal effect of risk management on loan quality. Review of Financial Studies 28 : 3367 – 406 . Google Scholar Crossref Search ADS Berg, T., Puri, M., and Rocholl. J. 2017 . Loan officer incentives, internal ratings and default rates. Working Paper. Berger, A., and Udell. G. 1992 . Some evidence on the empirical significance of credit rationing. Journal of Political Economy 100 : 1047 – 77 . Google Scholar Crossref Search ADS Bernanke, B. 1983 . Non-monetary effects of the financial crisis in the propagation of the Great Depression. American Economic Review 73 : 257 – 76 . Brenke, K., Rinne, U. and Zimmermann. K. 2012 . Short-time work: The German answer to the great recession. IZA Working Paper. Bubb, R., and Kaufman, A. A. 2014 . Securitization and moral hazard: Evidence from credit score cutoff rules. Journal of Monetary Economics 63 : 1 – 18 . Google Scholar Crossref Search ADS Campello, M., Graham, J. R. and Harvey. C. R. 2010 . The real effects of financial constraints: Evidence from a financial crisis. Journal of Financial Economics 97 : 470 – 87 . Google Scholar Crossref Search ADS Chodorow-Reich, G. 2014 . The employment effects of credit market disruptions: Firm-level evidence from the 2008-2009 financial crisis. Quarterly Journal of Economics 129 : 1 – 59 . Google Scholar Crossref Search ADS Cingano, F., Manaresi, F. and Sette. E. 2013 . Does credit crunch investments down? Review of Financial Studies 29 : 2737 – 73 . Google Scholar Crossref Search ADS De Marco, F. Forthcoming . Bank lending and the European sovereign debt crisis. Journal of Financial and Quantitative Analysis . Duchin, R., Ozbas, O. and Sensoy. B. A. 2010 . Costly external finance, corporate investment, and the subprime mortgage crisis. Journal of Financial Economics 97 : 418 – 35 . Google Scholar Crossref Search ADS Fan, J., and Gijbels. I. 1996 . Local polynomial modelling and its applications. In Monographs on statistics and applied probability . Boca Raton, FL : CRC Press. Faulklender, M., and Petersen. M. 2012 . Investment and capital constraints: Repatriations under the American Jobs Creation Act. Review of Financial Studies 25 : 3351 – 88 . Google Scholar Crossref Search ADS Fazzari, S., Hubbard, R. and Petersen. B. 1988 . Financing constraints and corporate investment. Brooking Papers on Economic Activity 1 : 141 – 206 . Google Scholar Crossref Search ADS Gan, J. 2007 . The real effects of asset market bubbles: Loan- and firm-level evidence of a lending channel. Review of Financial Studies 20 : 1941 – 73 . Google Scholar Crossref Search ADS Gertler, M., and Gilchrist. S. 1994 . Monetary policy, business cycles, and the behavior of small manufacturing firms. Quarterly Journal of Economics 109 : 309 – 40 . Google Scholar Crossref Search ADS Hahn, J., Todd, P. and van der Klaauw. W. 2001 . Identification and estimation of treatment effects with a regression-discontinuity design. Econometrica 69 : 201 – 9 . Google Scholar Crossref Search ADS Han, S., and Qiu. J. 2007 . Corporate precautionary cash holdings. Journal of Corporate Finance 13 : 43 – 57 . Google Scholar Crossref Search ADS Jensen, M. C., and Meckling. W. H. 1976 . Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3 : 305 – 60 . Google Scholar Crossref Search ADS Jiménez, G., Ongena, S. Peydró, J. and Saurina. J. 2012 . Credit supply and monetary policy: Identifying the bank balance-sheet channel with loan applications. American Economic Review 102 : 2301 – 6 . Google Scholar Crossref Search ADS Kashyap, A., and Stein. J. 2000 . What do a million observations on banks say about the transmission of monetary policy. American Economic Review 90 : 407 – 28 . Google Scholar Crossref Search ADS Keynes, J. M. 1936 . The general theory of employment, interest and money . India : Atlantic Publishers & Distributors. Lamont, O. 1997 . Cash flow and investment: Evidence from internal capital markets. Journal of Finance 52 : 83 – 109 . Google Scholar Crossref Search ADS Lee, D. S., and Lemieux, T. 2009 . Regression discontinuity designs in economics. Working Paper, NBER. Lin, H., and Paravisini. D. 2011 . The Effect of Financing Constraints on Risk. Review of Finance 17 : 229 – 59 . Google Scholar Crossref Search ADS McCrary, J. 2008 . Manipulation of the running variable in the regression discontinuity design: A density test. Journal of Econometrics 142 : 698 – 714 . Google Scholar Crossref Search ADS Popov, A., and Rocholl. J. Forthcoming. Financing constraints, employment, and labor compensation. Journal of Financial Intermediation . Rauh, J. 2006 . Investment and financing constraints: Evidence from the funding of corporate pension plans. Journal of Finance 61 : 33 – 75 . Google Scholar Crossref Search ADS Riddick, L. A., and Whited. T. M. 2009 . The corporate propensity to save. Journal of Finance 64 : 1729 – 66 . Google Scholar Crossref Search ADS Ruckes, M. 2004 . Bank competition and credit standards. Review of Financial Studies 17 : 1073 – 102 . Google Scholar Crossref Search ADS Sofianos, G., Wachtel, P. and Melnik. A. 1990 . Loan commitments and monetary policy. Journal of Banking & Finance 14 : 677 – 89 . Google Scholar Crossref Search ADS Stiglitz, J., and Weiss. A. 1981 . Credit rationing in markets with imperfect information. American Economic Review 71 : 393 – 410 . Thistlewaite, D., and Campbell. D. 1960 . Regression-discontinuity analysis: an alternative to the ex-post facto experiment. Journal of Educational Psychology 51 : 309 – 31 . Google Scholar Crossref Search ADS Udell, G. F. 1989 . Loan quality, commercial loan review and loan officer contracting. Journal of Banking & Finance 13 : 367 – 82 . Google Scholar Crossref Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Review of Financial Studies Oxford University Press

Got Rejected? Real Effects of Not Getting a Loan