of  

or
Sign in to continue reading...

An Anatomy of a Sudden Stop and the Credit Supply Channel

Salih Fendoglu
By Salih Fendoglu
5 years ago
An Anatomy of a Sudden Stop and the Credit Supply Channel

Sales


Create FREE account or Login to add your comment
Comments (0)


Transcription

  1. An Anatomy of a Sudden Stop and the Credit Supply Channel Salih FENDO ĞLU Steven ONGENA November 2018 Working Paper No: 18/18
  2. © Central Bank of the Republic of Turkey 2018 Address: Central Bank of the Republic of Turkey Head Office Structural Economic Research Department İstiklal Caddesi No: 10 Ulus, 06050 Ankara, Turkey Phone: +90 312 507 80 04 Facsimile: +90 312 507 78 96 The views expressed in this working paper are those of the author(s) and do not necessarily represent the official views of the Central Bank of the Republic of Turkey. The Working Paper Series are externally refereed.
  3. An Anatomy of a Sudden Stop and the Credit Supply Channel Salih Fendo ˘glu*, Steven Ongena** Abstract How does a sudden stop affect the local supply of credit by domestic banks? To answer this question, we study the Turkish credit register around Lehman. We find that the credit supply channel during a sudden stop works through not only banks’ ex-ante level of reliance on foreign funding (as is often documented in extant research), but also the difficulty of banks in rolling over their maturing foreign debt (which we capture by remaining maturity of and premium on their foreign funding), and for banks with ex-ante more optimistic exchange rate expectations, their holding of ex-ante riskier loan portfolios. Our research thereby uncovers the latter two channels as a significant source of amplification in contracting credit supply during a sudden stop. Keywords: Sudden Stop; Credit Supply Channel, Emerging Market Economies. JEL Codes: E44; F34; F41. ∗ Central Bank of the Republic of Turkey. ∗∗ University of Zurich, Swiss Finance Institute, KU Leuven, CEPR. E-mail addresses: salih.fendoglu@tcmb.gov.tr; steven.ongena@bf.uzh.ch Acknowledgments: Ongena acknowledges financial support from ERC ADG 2016 - GA 740272 lending. The views expressed here are only those of the authors and should not be attributed to the Central Bank of the Republic of Turkey. November 28, 2018
  4. Non-Technical Summary In this paper , we study a particular channel through which a sudden reversal of international capital flows, often called ``sudden stop'', may affect an emerging market economy: the credit supply channel. Our key contribution in this regard is to broaden the understanding on banks' ex-ante exposure to a sudden stop shock (or, in a more general setting, banks' exposure to a liquidity shock), and how it matters for the credit supply dynamics in the aftermath. We study the Turkish credit register around Lehman (2008 – 2009). Our findings are as follows: Banks that ex-ante rely more on short-term foreign funding or pay ex-ante higher cross-border interest premium and banks that were ex-ante more optimistic about the external funding conditions prior to the sudden stop cut back their supply of credit more strongly in the aftermath of a sudden stop. The channel works through not only the dry-up of foreign funding, as shown in the previous literature in related contexts, but also the difficulty of banks in rolling over their maturing foreign debt as well as, for the case of ex-ante more optimistic banks, their holding of riskier loan portfolios prior to the sudden stop. Moreover, the credit supply channel affects firms differently. In particular, the reduction in credit supply is stronger for smaller, more indebted, and non-exporter firms, and these effects are further amplified by banks with ex-ante shorter maturity foreign funding or ex-ante more optimistic expectations. Finally, firms’ indirect exposure to a sudden stop has real repercussions in the aftermath. In particular, firms that ex-ante work more with banks that rely more on global liquidity prior to the sudden stop cannot make up for the loss in bank credit supply by borrowing from less affected banks, borrowing from other firms, or issuing other forms of debt. Eventually, such firms reduce their investment and employment more.
  5. 1 . Introduction Sudden reversals of international capital flows, “sudden stops”, happen fast, and when they happen, they have detrimental effects on the domestic economy (Calvo, 1998; Calvo et al., 2008; Mendoza, 2010). In this paper, we study a particular channel through which a sudden stop may affect an emerging market economy: the credit supply channel. Our key contribution is to broaden the understanding on banks’ ex-ante exposure to a sudden stop shock (or, in a more general setting, banks’ exposure to a liquidity shock), and how it matters for the credit supply dynamics in the aftermath. A natural candidate measure of banks’ exposure to a sudden stop is by how much they rely on foreign funding prior to the sudden stop. This, however, may underestimate the ‘true’ exposure. Given that banks’ supply of credit depends fundamentally on their expectation of and ability to access to liquidity (Diamond and Rajan, 2006; Allen and Gale, 2007; Morais et al., 2018), the cross-sectional variation across banks in the difficulty of re-financing their maturing foreign debt during the sudden stop should also matter. For example, banks with a shorter time-to-maturity or higher premium for their foreign funding prior to the sudden stop would find it particularly costly to refinance their maturing debt during the sudden stop, and thus, are potentially more likely to reduce their supply of credit. We expect this channel, which is visited in the literature in different contexts (see, e.g., Stein, 2012; Segura and Suarez, 2017), to be operational for sudden stops as well. In addition, banks that were more optimistic about the external funding conditions prior to the sudden stop may ex-ante hold different, possibly riskier, loan portfolios. In turn, their capacity to continue lending may become more sensitive to a sudden stop. The first challenge is then to have a finer measure of bank exposure to a sudden stop, namely, the time-to-maturity and premium of banks’ foreign funding as well as banks’ expectations about future external funding conditions prior to a sudden stop. Presenting a second challenge is the fact that factors that affect domestic banks’ access to foreign funding may simultaneously affect non-financial firms. For instance, a currency depreciation, which is particularly sharp during sudden stops, may deteriorate the strength of the firms’ balance sheets (particularly for firms with lower foreign currency assets). Similarly, gloomy economic prospects may affect the firms’ expected income stream or their perceived ability to pay their debts back. Therefore, the observed reduction in credit during a sudden stop is driven not only by 2
  6. supply side effects but also the demand side . A third and final challenge is the fact that sudden stops usually occur after a period of financial imbalances, e.g., Mexico during 1993-1994, East Asian economies 1996-1997, Russia 1997-1998, and Argentina and Turkey during 1999-2001. Such imbalances generally sow the seeds of a subsequent reversal in capital flows (Mendoza and Terrones, 2012; Eichengreen and Gupta, 2016). In other words, sudden stops may not always come entirely unexpected. We address these three challenges as follows. To address the first challenge, we use (i) the loan-level database on the universe of cross-border transactions of banks operating in Turkey to calculate the average remaining maturity of and the interest premium on their foreign funding; and (ii) unique (but proprietary) survey data on USD/Turkish lira exchange rate expectations to gauge banks’ expectations about future external funding conditions.1 To address the second challenge, we use the Turkish Credit Register which provides loan-level details on virtually all corporate loans extended by all banks operating in Turkey. By using the Credit Register, we are able to control for demand side effects and identify the supply side. Finally, to address the third challenge, we take the Lehman Brothers’ collapse in September 2008, an event exogenous to Turkey. Turkey provides an excellent laboratory to study the impact of a sudden stop on an emerging market economy through the credit supply channel. First, banks are the main providers of finance (with equity financing playing a negligible role for majority of firms). Second, global liquidity is an important funding source for banks, comprising about 16% of their size in terms of assets. Indeed, concurrent with the drop in foreign-currency wholesale funding, we observe a sharp drop in aggregate credit after September 2008 (Figure 1(A)). As foreign wholesale funding came practically to a halt after September 2008, annual growth in aggregate domestic credit dropped from over 30% to almost nil within a year. Moreover, as Figure 1(B) suggests, banks that rely more on foreign wholesale funding prior to the sudden stop appear to provide less credit afterwards. Figure 2 further shows that banks that have shorter time to maturity on their foreign wholesale funding in September 2008 reduce their credit more strongly afterwards (the left panel). Moreover, for banks that ex-ante were paying high cross-border premia, particularly for domestic banks, we 1 The Central Bank of the Republic of Turkey (CBRT) collects, in each month, data on exchange rate expectations of financial institutions and large non-financial corporations. The question states “What does your institution expect the USD-Turkish lira exchange rate to be at the end of current month, at the end of the current year, and after 12 months?” We use bank-level expectations about the end-of-the-year exchange rate prior to the sudden stop. See Appendix A for details. 3
  7. also find a stronger drop in credit (the middle panel). Lastly, banks that were expecting a more valued domestic currency in December 2008, i.e. those that were more surprised about the reversal in external funding conditions, provide less credit afterwards (the right panel). The results are, in general, stronger for domestic banks (black dashed lines). To this end, we identify the effect of a sudden stop on the domestic credit supply. In particular, we match the Credit Register with the bank-level data on remaining maturity and premium on foreign funding, exchange rate expectations, as well as key bank characteristics (most importantly, banks’ non-core foreign currency liabilities, capital adequacy and liquidity ratios, and size). We then explore how banks with different ex-ante exposures to the sudden stop in September 2008 change their supply of a similar type of loan to the same firm in the following year.2 As discussed above, a sharper inference requires a richer perspective on banks’ exposure to the sudden stop. We do so by exploiting the cross-sectional variation across banks not only in their ex-ante level of reliance on foreign funding, as is often done in extant research, but also in the remaining maturity and premium on foreign funding, as well as in exchange rate expectations prior to the sudden stop. We find robust evidence that a sudden stop leads to a sharper reduction in credit supply by banks with a higher ex-ante reliance on foreign funding, and importantly, by banks facing higher difficulties in rolling over their maturing debt (namely, banks with shorter remaining maturity and higher premium on their foreign funding), and by banks that were ex-ante more optimistic about the external funding conditions (namely, banks that expect a more valued domestic currency).3 Our results are not only statistically significant but also economically large. For instance, a bank with 10% higher ex-ante foreign funding reduces its credit supply by 9.1% more.4 Moreover, a bank with a 1-year shorter time to maturity for its foreign funding cuts back its supply of credit by an additional 10.5%. The premium on foreign funding matters as well: by an additional 9% drop in credit supply by a bank with a 1% point higher cross-border premium. Moreover, banks that are more surprised about the sudden reversal in external funding conditions, i.e., those that were more optimistic, reduce their supply of credit significantly, and by an additional 9.2%. 2 By additionally controlling for loan types, we address the concern that banks with higher ex-ante exposure to the sudden stop may have granted different types of loans prior to the sudden stop, and in this regard, we make sure that our results are not driven by potential differences in the types of loans extended by highly exposed banks. 3 Throughout the text, we use “foreign funding”, “foreign wholesale funding”, or “foreign-currency wholesale funding” interchangeably. 4 The hypothetical value that we choose for brevity, i.e., 10%, corresponds to roughly half of the interquartile range of banks’ foreign wholesale funding. 4
  8. Evaluating these impacts this time at the interquartiles ranges for each respective exposure metric , i.e., maturity, premium, and surprise, reveals which metric matters more. Maturity seems to play a noticeably larger role (a bank at the first quartile of maturity of foreign funding reduces its credit supply by 24% more compared to a bank at the 3rd quartile), compared to price (a bank at the 3rd quartile of premium of foreign funding reduces its credit supply by 9.6% more compared to a bank at the first quartile) or surprise (a bank at the 3rd quartile of realized-minus-the-expectedexchange-rate reduces its credit supply by 9.2% more compared to a bank at the first quartile, as reported above). Our second set of results shows that these exposure metrics also play a role for the heterogeneity of credit supply dynamics across different firms. Banks with a higher reliance on foreign wholesale funding cut their supply of credit more strongly to smaller firms (in terms of asset size or employment), highly leveraged firms (those with higher debt-to-equity or short-term debt-to-total debt ratios), and less strongly for exporters. In most cases, we find that banks with ex-ante shorter maturity foreign funding or more optimistic expectations about the external funding conditions significantly amplify these effects. Finally, the reduction in credit supply is binding and has real consequences. Firms that borrowed more from banks with higher ex-ante reliance on foreign funding cannot compensate for the reduction in bank credit supply by switching to less affected banks or by increasing their borrowing from other firms (via trade credit), or through other forms of market debt. Eventually, firms that ex-ante borrowed more from affected banks reduce their total financial debt by 2.3%. Giving overall binding financial constraints, these firms, particularly those that depend on bank credit, eventually reduce their investment and employment more (by 2.3% and 0.6%, respectively). To further augment our work, we pursue several additional analyses. First, we focus on domestic banks, and show that the results are numerically stronger. A domestic bank that rely 10% more on foreign funding reduces its credit supply to a given firm by 12%. Moreover, similar as above, we evaluate the economic impacts of each exposure metric at their interquartiles ranges. The surprise dimension now plays a larger role. The reduction in credit supply is 18.5% for domestic banks with a shorter time-to-maturity foreign funding, 13% for higher premium domestic banks, and 19% for more surprised domestic banks. Second, we explore how banks adjust their credit limits for a given firm. Since credit limits are a form of commitment by banks that promise lines of credit, it has a sharper notion of credit supply. 5
  9. Exploring adjustments in credit limits is important also due to the fact that such adjustments play an important role in the theoretical literature on sudden stops (Mendoza, 2010).5 Numerically, a bank with 10% higher ex-ante foreign funding reduces its credit limit to a firm by 7.6%. Banks with 1-year shorter maturity or 1% higher premium on their foreign funding reduce their credit limits by 5.5% and 8.9% more. These dimensions, however, do not significantly matter when we additionally control for their degree of surprise about the sudden stop. A bank that is more surprised about the sudden stop reduces its credit limit by 14.1%. Third, the external margin matters significantly. Banks with higher ex-ante reliance on foreign wholesale funding and those with shorter time-to-maturity or higher cross-border premium are less likely to establish a new lending relationship with a firm during the sudden stop. On the termination margin, banks with shorter maturity or higher cross-border premium on their foreign funding are more likely to terminate their relationships with firms. Moreover, more surprised banks are less likely to establish a new lending relationship with a firm and are more likely to terminate an existing relationship with a firm. Overall, banks’ adjustments on the external margin are economically large. More surprised banks, for instance, are 3.2% less likely to start a new lending relationship with a firm (whereas the average probability of starting a new lending relationship during this period is 16%), and 6.6% additionally more likely to terminate its relationship with a firm (whereas the average is 22%). We conclude our external margin analyses by showing that the external margin works differently on different firms. Banks with higher ex-ante foreign funding prior to the sudden stop are less likely to establish a new lending relationship with smaller, younger, more indebted, or non-exporting firms. Moreover, shorter maturity or higher premium on bank foreign funding, as well as higher degree of surprise about the reversal in external funding conditions seem to play a significant role, and in general, amplify these effects significantly. Finally, we explore an additional channel –besides the dry-up of foreign liquidity and the higher difficulty of rolling over foreign debt– that may contribute to how banks adjust their credit supply during the sudden stop: the characteristics of their loan portfolios prior to the sudden stop. 5 Though, we depart from the key mechanism in this strand of literature and focus on the supply-side adjustments. Mendoza (2010) show that the deterioration in asset prices following sudden stops leads to a lower credit limit (or tighter borrowing constraints), which in turn, forces borrowers to fire sale their assets, which then reduces the limit further and feeds into an amplified downward spiral. In our paper, we control for possible deterioration in firms’ quality of assets and focus on the supply side. Since credit lines expose banks to liquidity and credit risks (Agarwal et al., 2006), one could expect supply side adjustments on credit limits to take place as well. 6
  10. The reduction in credit supply by banks that rely more on short-term foreign funding appears mainly related to their difficulty in rolling over their foreign funding during the sudden stop rather than such banks working with riskier firms . For the surprise dimension, we find strong evidence that banks that were more optimistic about the external funding conditions prior to the sudden stop are likely to have extended credit to riskier firms. In particular, firms that were granted a loan prior to the sudden stop by more optimistic banks are more likely to be smaller, younger, more indebted, more short-term indebted, or non-exporter, or, have defaulted on a loan at the bank recently and are more likely to default in the near future. For the premium, we find supportive evidence for both views (roll-over risks and riskier loan portfolio). Our paper is related to a large body of literature on the international transmission of liquidity shocks but with key differences. In particular, following a liquidity shock emanating from outside, local banks that rely on such funds and cannot easily substitute them with other funding sources are shown to transmit the shock more strongly to their clients.6 Our main contribution to this large strand of literature is to broaden the understanding on banks’ exposure to a ‘liquidity’ shock (in our case, the sudden stop). Namely, we show that the maturity and the price of funding, as well as banks’ expectations about future funding conditions, play a significant role, and in some cases, as important as the ex-ante level of reliance on that funding source, for the credit supply dynamics and heterogeneity of credit supply across different firms in the aftermath of funding shock. Moreover, these additional exposure metrics serve as a significant source of amplification, e.g., shorter maturity funded or more surprised banks reducing their credit supply significantly more, and even more so for smaller firms. Let us further highlight our main contribution over some of the related papers above. Ongena et al. (2015), for instance, show in a multi-country setting that internationally-borrowing domestic 6 Examples include the effect of the 1998 Russian default on the domestic credit supply of Peruvian banks (Schnabl, 2012), the effect of ECB monetary policy on the credit supply of Spanish banks (Jimenez et al., 2012), the effect of the European interbank freeze in August 2007 on the credit supply of Portuguese banks (Iyer et al., 2014), the effect of US monetary policy on the credit supply and risk-taking behavior of Bolivian banks (Ioannidou et al., 2015), the international transmission of US monetary policy through global banks’ activating their internal capital markets (Cetorelli and Goldberg, 2012) or through global banks’ lending to non-affiliated parties abroad (Temesvary et al., 2018), the effect of global uncertainty (the VIX) on the credit supply of banks in Turkey (di Giovanni et al., 2018), the effect of global liquidity conditions (measured by the VIX or the US monetary policy) on the effectiveness of domestic monetary policy transmission in Turkey (Fendoglu et al., 2018), the effect of Lehman Brothers’ collapse on bank credit supply and firm performances in Eastern Europe and Central Asia (Ongena et al., 2015), and the effect of monetary policy in financial center economies on the credit supply of foreign banks in Mexico (Morais et al., 2018), to name a few. 7
  11. and foreign-owned banks reduce their credit more during the global financial crisis than domestic banks that are funded only locally , and that the reduction of credit has real effects on firms’ performances. Improving upon Ongena et al. (2015), we exploit Credit Register data and broaden the perspective on exposure to a sudden stop that enable us to better identify the credit supply channel. Similar to us, di Giovanni et al. (2018) and Fendoglu et al. (2018) also use the Turkish Credit Register. di Giovanni et al. (2018) show that when global uncertainty (VIX) is lower, locally-owned banks with higher level of non-core liabilities increase their supply of credit and reduce their loan rates. Fendoglu et al. (2018) show that easier global liquidity conditions (measured by lower VIX or expansionary US monetary policy) limit domestic monetary policy transmission. While our primary focus is different, we contribute to this strand of literature by uncovering additional channels through which global financial conditions transmit to the domestic economy through banks, namely, the price, maturity and expectations channels, and show that they significantly amplify domestic credit supply dynamics, and affect access to credit across different types of firms. The paper proceeds as follows: Section 2 presents details about the databases and the empirical strategy. Section 3 provides empirical results. Section 4 provide further insights and several robustness analyses. Section 5 concludes. 2. Data and Empirical Strategy The Turkish Credit Register (CR) provides confidential and detailed information about the universe of loans granted by all banks operating in Turkey. The CR is maintained and supervised by The Banks Association of Turkey Risk Center. Banks are obliged to report their outstanding loans with the universe of firms they are working with, along with various details about the loans. The CR is exhaustively comprehensive, as there is practically no reporting threshold.7 We mainly focus on the period of September 2008 to September 2009, the former coinciding with the Lehman Brothers’ collapse and the latter by and large coinciding with the lowest growth in aggregate domestic credit after September 2008 (Figure 1(A)).8 In total, we have 298,304 firms (or 144,121 firms with multiple banking relationships), 23 different types of loans (e.g. domestic vs. foreigncurrency denominated, non-cash loans (letters of guarantee, accreditation, and others), financial leasing, etc.), and 31 banks. 7 The results are robust if we exclude loans below a certain threshold, e.g., 1,000 TRY or 5,000 TRY. 8 Later, we also study different end periods for robustness. 8
  12. We match the CR with the monthly balance sheet , income statement, and supervisory bankcapital datasets. Our focus bank variable is the non-core foreign-currency liabilities in proportion to total assets (in short, foreign wholesale funding or foreign funding). We further consider for each bank the following control variables: capital adequacy ratio (total bank capital to riskweighted assets), liquidity ratio (ratio of liquid assets to total assets), size (in natural logarithm of total assets), profitability (the ratio of net profit to total assets, or in short, return on assets (ROA)), and non-performing loans ratio (the ratio of total non-performing loans to total loans (NPL)). Further, we include the maturity and the price of foreign funding, as well the degree of banks’ surprise about the reversal in external funding conditions (we discuss them below). We use these bank controls exhaustively, in levels and interactions, to provide a clearer picture about the transmission of sudden stop on domestic credit supply. To properly identify the credit supply channel, we study firms with multiple banking relationships and saturate the model with firm fixed effects (Khwaja and Mian, 2008). Moreover, to control for the possibility that banks with higher ex-ante reliance on foreign funding might have extended different types of loans prior to the sudden stop, we include loan type fixed effects. Baseline Specification - For our baseline specification, we estimate variants of the following equation: ∆L b f a,post = β Bank Foreign Fundingb,pre + OTHER TERMS + µ f + ζ a + ε b f a,post (1) where β = γ0 + γ1 Maturityb,pre + γ2 Premiumb,pre + γ3 Surpriseb,pre OTHER TERMS = ζBank Foreign Fundingb,pre ⊗ Capitalb,pre , Liqb,pre , Sizeb,pre , ... ... ROAb,pre , NPLb,pre , R b f ,pre + ... + γ ⊗ Maturityb,pre , Premiumb,pre , Surpriseb,pre + η ⊗ Capitalb,pre , Liqb,pre , Sizeb,pre , ROAb,pre , NPLb,pre + R b f ,pre where ∆L b f a,post is the log change in outstanding credit provided by bank b to firm f in loan type a from September 2008 to September 2009. In later analyses, we use the log change in credit limit committed by bank b to firm f for loan-type a over the same period, as the dependent 9
  13. variable . In subsequent sections, we further replace the dependent variable, ∆L b f a,post , with (i) the new lending margin (the probability of bank b establishing a new lending relationship with firm f in September 2009 that bank b has not been working with in September 2008); and (ii) the termination margin (the probability of bank b terminating its existing relationship with firm f ). “Bank Foreign Fundingb,pre ” is the ratio of non-core foreign-currency liabilities to total assets ratio, and Bb,pre ≡ Capitalb,pre , Liqb,pre , Sizeb,pre , ROAb,pre , NPLb,pre are the bank controls, all measured as of September 2008 (hence the subscript pre). R b f ,pre is a measure of the strength of the bank-firm relationship, namely the ratio of outstanding balance of firm f at bank b to total outstanding bank loans the firm f has in September 2008. µ f stand for firm fixed effects, and ζ a denote loan-type fixed effects. “Maturityb,pre ” denotes average time-to-maturity on foreign wholesale funding of bank b (measured ex-ante).9 To calculate the maturity, we use the Financial Sector Database (MS) supervised by the BRSA and the CBRT that provides transaction-level details about the universe of crossborder borrowing by all banks operating in Turkey. The details include unique borrower and lender identifiers, volume, interest rate, currency of denomination, and origination and termination dates for each transaction. For bank b, we calculate average time-to-maturity on its cross-border foreigncurrency liabilities, using loan volumes as weights (e.g. smaller loans receive lower weight) and the termination dates.10 “Premiumb,pre ” denotes weighted average interest premium that bank b pays above the London Interbank Borrowing Rate (LIBOR) of the same currency with the closest maturity.11 For instance, if a domestic bank is charged with a 2% interest rate on a USD-denominated cross-border loan with a maturity of 5.5 months, we subtract the 2% from the 6-month USD-LIBOR rate.12 We then calculate for bank b the weighted average of these spreads, using loan volumes as weights. Lastly, “Surpriseb,pre ” denotes the log difference between the realized exchange rate at December 2008 and what bank b expects in September 2008 about the USD/TRY exchange rate for December 2008. A higher value implies more optimistic expectations about future foreign funding conditions, and in turn, being more surprised with the sudden stop. Appendix A provides further 9 We use the term “Maturity” and “Time-to-Maturity” interchangeably throughout the text. 10 The results are robust if we consider the maturity at the loan origination (available upon request). 11 We use the term “Premium” and “Cross-Border Interest Rate Premium” interchangeably throughout the text. 12 We use 1-, 3-, 6-, and 12-month LIBOR rates for three major currencies (USD, Euro, and British pound). These currencies constitute more than 99% of cross-border borrowing of banks operating in Turkey in September 2008. 10
  14. details on the database and how expectations are used . Firm Heterogeneity - To this end, we explore whether the effect of sudden stop on credit supply differs across different firms. In doing so, we match the CR with the Central Bank of Turkey’s Company Accounts Database (CAD), that provides balance sheets and income statements of a large set of non-financial firms. The CAD is annual, and therefore, we use end-of-2008 information on firms. In total, we have detailed information about 11,818 firms (and 10,475 firms with multiple banking relationships). In sum, we estimate variants of the following equation: ∆L b f a,post = β1 Bank Foreign Fundingb,pre ∗ I(Firm(k) < ki ) + ... f ,pre > ki ) + ... + β2 Bank Foreign Fundingb,pre ∗ I(Firm(k) f ,pre ... + Bb,pre + R b f ,pre + µ f + ζ a + ε b f a (2) where Firm(k) is a key firm characteristic and ki is its ith decile across firms (where i = 1, 2, 3, ..., f ,pre or 9). For firm characteristics (k), we consider firm size (measured by log of total assets, and alternatively, by log of number of employees), age (log of years since the incorporation of the firm), indebtedness (the ratio of total debt to total equity, and alternatively, the ratio of short-term debt to total debt), and export status (the ratio of overseas sales to total sales). I(Firm(k) < ki ) is an indicator variable that is equal to 1 if the firm characteristic k is lower f ,pre than the ith decile, and 0 otherwise. For example, consider firm size (k ≡ size) and i = 5 (the median). Equation (2) then boils down to the hypothesis of how banks with higher ex-ante higher foreign funding change their supply of credit to firms below the median size ( β1 ) and for firms above the median ( β2 ). To account for possible non-linearities, we consider each of the deciles. We also allow β1 and β2 to be a function of “Maturity”,“Premium”, and “Surprise”. For “Maturity” for instance, our full empirical model includes two triple interactions with the Maturity, and incorporates all the remaining interaction and level terms. That is, ∆L b f a,post = β1 Bank Foreign Fundingb,pre ∗ Maturityb,pre ∗ I(Firm(k) < ki ) + ... f ,pre + β2 Bank Foreign Fundingb,pre ∗ Maturityb,pre ∗ I(Firm(k) > ki ) + ... f ,pre + OTHER TERMS + µ f + ζ a + ε b f a where 11 (3)
  15. OTHER TERMS = ζ1 Bank Foreign Fundingb,pre ⊗ Premiumb,pre , Surpriseb,pre , Capitalb,pre , ... ... Liqb,pre , Sizeb,pre , ROAb,pre , NPLb,pre , R b f ,pre + ... > ki ) + ... < ki ), I(Firm(k) + ζ2 Bank Foreign Fundingb,pre ⊗ I(Firm(k) f ,pre f ,pre + ζ3 Maturityb,pre ⊗ I(Firm(k) < ki ), I(Firm(k) > ki ) + ... f ,pre f ,pre + Bb,pre + R b f ,pre + γ ⊗ Maturityb,pre , Premiumb,pre , Surpriseb,pre where we report estimates for β1 and β2 (and whether they are statistically significant). By doing so, we uncover an additional source of amplification, e.g., whether banks with shorter maturity foreign funding reduce their supply of credit strongly to large firms as well. In subsequent sections, we then study credit lines and external margin of credit supply across different firms. Real Effects - Finally, we study whether the change in credit supply is binding, or has real outcomes. Note that firms may obtain credit from less affected banks, or even substitute bank loans with other forms of market debt (e.g. trade credit). In this regard, a potential reduction in credit supply may not be binding. Even more starkly, even if the reduction were binding, it is not a priori obvious whether that would lead to significant real economic outcomes. To address these questions, we estimate variants of the following equation: ∆Yf ,post = β1 I Bank Dependent f ,pre Firm’s Exposure to Banks’ Foreign Funding f ,pre + ... + β2 I Bank Non-Dependent f ,pre Firm’s Exposure to Banks’ Foreign Funding f ,pre + ... + Fb,pre + µ s + ε f (4) where ∆Yf ,post is the change in the outcome variable for firm f from year 2008 to 2009. The outcome variables are total bank credit, trade credit, total financial debt, fixed assets and employment. “Firm’s Exposure to Banks’ Foreign Funding f ,pre ” is the weighted average of foreign wholesale funding of banks that firm f was working with in September 2008.13 This variable serves as an instrument for firm f ’s exposure to the sudden stop shock through its relationship with banks. We interact this instrument with indicator variables that reflect by how much firm f re13 For each firm, we calculate the average of foreign wholesale funding of banks that the firm is borrowing from, using outstanding loan balances as weights. For example, foreign funding of banks with smaller loans to the firm receive a lower weight. 12
  16. lies on bank credit , as one would expect the exposure to matter significantly more if firm f is bank dependent. Namely, I Bank Dependent f ,pre takes a value 1 if firm f ’s total bank credit in proportion to its total financial debt is above the NACE 2-digit industry median, and 0 otherwise, that is, whether or not firm f relies heavily on bank credit compared to its peers. Similarly, I Bank Non-Dependent f ,pre takes a value 1 if firm f does not rely on bank credit compared to its sector median, and 0 otherwise. Fb,pre is the set of firm characteristics, namely, size, age, debtto-equity ratio, the ratio of short-term debt to total debt, and the export status. Lastly, to control for sector specific changes in outcome variables, we include NACE 2-digit sector fixed effects (µ s ). Accordingly, we test the following two hypotheses: (i) whether a bank dependent firm with higher exposure to banks’ foreign wholesale funding experience a significant change in the outcome variable (H0 : β1 = 0); and (ii) whether being bank dependent matters (H0 : β1 = β2 ). 3. Empirical Results 3.1. Baseline Results Table 2 presents the baseline results. We start with the universe of firms (column 1), and from column 2 and onward, we focus on firms with multiple banking relationships. We successively saturate the model with “Bank Foreign Funding” in interaction with “Maturity”,“Premium”, and “Surprise”. Column (8) is the most saturated specification, which includes the interaction of bank foreign funding with maturity, price, and surprise, as well as with all the bank controls (capital adequacy, liquidity, size, ROA, and NPL), and the strength of the bank-firm relationship, as well as all the respective variables in levels. The results show that banks’ ex-ante reliance on foreign wholesale funding prior to the sudden stop plays a statistically significant, robust, and economically important role for how they adjust their credit supply to a given firm after the sudden stop. Numerically, a bank with 10% higher foreign funding reduces its supply of a similar type of loan to a given firm by 9.1% (column 2). Moreover, a bank with a 1-year shorter time-to-maturity on its foreign funding reduces its credit supply by an additional 4.2% (column 3). Similarly, a bank with a 1% point higher premium on its foreign funding cuts its credit supply by an additional 2.1% (column 4). In column (5), we include both the maturity and price, and in column (6), we further control for bank controls and the strength of the bank-firm relationship in interaction with bank foreign funding. The effects are 13
  17. numerically larger : an additional 10.5% reduction in credit supply by shorter maturity banks, and additional 17.9% by higher premium banks. In columns (7) and (8), we study whether the degree of surprise matters. Column (7) includes the surprise in interaction with bank foreign funding. Column (8) is the most saturated specification. Previous results are qualitatively robust. Moreover, the extent at which a bank is surprised about the reversal in foreign funding conditions matters. A bank that is more surprised (the 3th quartile) reduces its supply of a similar type of loan to a given firm by 9.2% more strongly (compared to a bank that is less surprised –the first quartile–). Do Bank Characteristics Matter? In Table 3, we study whether the reduction in credit supply by highly exposed banks differs across different banks. In particular, we include bank controls and the strength of the bank-firm relationship in interaction with bank foreign wholesale funding one by one (columns (1) to (7)). In columns (8) and (9), we further include maturity, price and surprise in interaction with bank foreign funding –essentially replicating columns (6) and (8) of Table 2, respectively– for ease of assessing whether the interaction terms change signs when they are individually added into the analysis compared to the most saturated specifications. The results suggest that well-capitalized banks reduce their supply of credit less strongly (column 1). This result resonates well with the bank lending literature that bank capital serves as an important cushion for their clients against market liquidity shocks (see, e.g. Jimenez et al., 2012). Interestingly, we also find that more liquid or larger banks reduce their credit supply more strongly (see columns (2) and (3), and columns (7) to (9)).14 We interpret this finding as banks that are more liquid or larger might be holding a riskier loan portfolio prior to the sudden stop that leads them to reduce their credit supply more strongly (we explore this possibility in the next Section). Or, banks that were relatively liquid prior the sudden stop may be more inclined to make precautionary savings (by increasing their liquid assets more after the sudden stop for precautionary reasons). We do not find significant or robust differences with respect to bank return-on-assets, non-performing loans, or the strength of the bank-firm relationship. 14 This result is also qualitatively robust to saturating the model further with firm×bank-type fixed effects to enforce comparison across loans to the same firm by banks with the same type (available upon request). 14
  18. 3 .2. Does The Reduction in Credit Supply Differ Across Different Firms? We now explore whether the reduction in credit supply is borne differently across different firms. Exploring different types of firms serves two purposes: First, it helps for better identification. For example, banks with higher reliance on global liquidity may not be reducing their credit supply to large firms, unless such banks have shorter maturity on their foreign funding (this is indeed the case, as we show below). Second, to understand what types of firms are affected the most during sudden stops help reach further policy insights. Figure 3 presents the results. The first column reports the baseline estimates (by estimating equation (2)). The second column reports estimates for the maturity dimension (equation (3)). Similarly, the third and the fourth columns study the price and surprise dimensions. We test whether banks with shorter time-to-maturity or higher premium on their foreign funding or those that are more surprised reduce its credit supply differently across different firms. The baseline results (first column) show that the sudden stop and the associated reduction is credit supply is felt the most by smaller firms (those with lower assets or lower employment), indebted firms (those with higher debt-to-equity or short-term debt-to-total debt ratios), or nonexporter firms. Moreover, the results are qualitatively robust across different deciles. Numerically, while very small firms (those with total assets less than the first decile) experience a reduction in credit supply by 8.5% by highly exposed banks, larger firms (those with total assets higher than the first decile) experience a reduction in credit supply significantly less (by about 3%). Moreover, firms larger than the median size do not experience a statistically significant drop in credit supply. These results qualitatively carry through when we consider firm size in terms of employment, i.e. firms with fewer employees experience a statistically significant and economically larger drop in credit supply. We do not find an economically large or robust evidence on whether firm age matters for the reduction in credit supply. Shorter maturity on bank foreign funding amplifies the responses (second column). For example, banks with a 1-year shorter time-to maturity on their foreign funding reduce their credit supply even more strongly for smaller firms. In particular, a small firm (with total assets less than the median) experiences an additional 5.5% reduction in credit supply if the bank were ex-ante holding 1-year shorter maturity foreign funding. Moreover, while we confirm existing literature that large firms in general do not experience a significant reduction in credit supply by more ‘exposed’ banks after a liquidity shock (see, e.g., Khwaja and Mian, 2008; Iyer et al., 2014), our 15
  19. findings further uncover that banks that ex-ante rely more on short-term foreign funding (or banks with ex-ante more optimistic expectations about the external funding conditions, as we show next) reduce their supply of credit to large firms as well. Large firms (those with total assets above the 9th decile) still experience a drop in credit supply (by about 4%). We find similar results when we consider firms with different number of employees. For firm age, we do not find economically large or strong evidence. We also find that banks that are charged with a higher cross-border premium prior to the sudden stop reduce their supply of credit more strongly to smaller or younger firms (third column). Yet, these results are significant in only a few cases. Moreover, more surprised banks reduce their supply of credit more strongly to smaller or younger firms (fourth column). A more surprised bank, for instance, reduces its credit supply by an additional 12% to very small firms (those below the first decile), and by an additional 8.5% to very large firms (those above the ninth decile). The results are qualitatively similar when we consider firm size in terms of number of employees. Moreover, very young firms (those with an age less than the first decile) experience an additional 12.8% drop in credit supply if the bank was more surprised, while older firms experience a milder reduction (of about 9%). For other deciles, the differences are economically similar. Next, we continue with other firms characteristics: firm indebtedness and export status. We find robust evidence that firms that are more indebted, those that hold mostly short-term debt, and those that are relatively non-exporter, experience a significantly stronger reduction in credit supply by banks with higher ex-ante reliance on global liquidity. Moreover, banks with shorter time-to-maturity or those that are more surprised reduces their credit supply more strongly for more indebted firms. Banks with higher ex-ante cross-border premium appear to behave not differentially across firms with different indebtedness or export status. In sum, the results show that maturity and expectations channels matter for heterogeneous credit supply responses. Shorter maturity and more surprised banks appear to reinforce our baseline results significantly, eventually making smaller, indebted or non-exporter firms being hit even stronger. 3.3. Binding Financial Constraints and Real Effects So far, the evidence suggests that banks with a higher degree of reliance on global liquidity reduce their supply of a similar loan to a firm more strongly, and that these effects differ across 16
  20. different firms . These results do not imply that the reduction in credit supply poses a binding constraint on firms’ overall financial needs. For example, smaller firms –which experience the most significant reduction in bank credit supply as we show above– may obtain trade credit from large firms (for which the reduction in bank credit supply is milder). To address these concerns, we explore firm-level aggregate outcomes by estimating equation (4). Table 4 presents the results. Column (1) shows that firms that borrowed more from banks that have higher ex-ante reliance on global liquidity reduces their total bank credit more (by 11%).15 That is, these firms cannot easily switch to less affected banks. Moreover, these firms raise their trade credit from other firms by 4.7% more (column (2)). Yet, they had to reduce their total financial debt (that includes bank credit, total credit, and all other forms of market debt) by 2.3% more. In other words, such firms face overall binding financial constraints. Eventually, they reduce their fixed assets by 0.7% more, and particularly more strongly if they depend on bank credit (by 2.4%). Moreover, such firms reduce their number of employees by 0.6%. 4. Further Discussions and Robustness 4.1. Domestic Banks The results are qualitatively robust when we focus on domestic banks. In particular, we reestimate our baseline specification by focusing on domestic banks only. Table 5 presents the results. A domestic bank with 10% higher foreign funding prior to the sudden stop cuts its supply of a similar type of loan to a given firm by 12.2% in the aftermath, by additional 8.5% if the bank has 1-year shorter time-to-maturity on its foreign funding, an additional 12.8% if has 1% point higher cross-border premium, and an additional 28% if the bank was more surprised. The biggest difference compared to baseline results appears on the surprise margin.16 15 In calculating the economic magnitudes, we compare two firms, one that is more exposed to banks’ ex-ante foreign wholesale funding (evaluated at the third quartile) versus the one that is less exposed (evaluated at the first quartile). These quartiles correspond to 0.134 and 0.101, respectively. 16 Since domestic banks, on average, rely less on foreign funding, with an interquartile range of [0.10-0.215] compared to the range of [0.082-0.252] for the full sample of banks, we can re-adjust our hypothetical value of 10% –that we use to label a bank as highly dependent on foreign funding–. Namely, consider a value of 6.8% instead of 10% for ‘higher foreign funding’ (by comparing the interquartile ranges). A domestic bank with a 6.8% higher foreign funding, cuts back its supply of a similar type of loan to a given firm by 8.3%, by an additional 5.7% if has 1-year shorter time-to-maturity on its foreign funding, by an additional 8.7% if has 1% point higher cross-border premium, and by an additional 19% if more surprised. 17
  21. 4 .2. Change in Credit Limits? A reduction in credit limits during the sudden stop would signal a powerful deterioration in firms’ ease of doing business, since credit lines provide firms with access to credit to help them manage their working capital, which may be particularly hard during unexpected liquidity shortfalls, and help reduce asymmetric information about the firm (Kanatas, 1987; Berger and Udell, 1995; Agarwal et al., 2006). Moreover, adjustments on credit limits would imply tighter borrowing constraints on firms (as shown in the theoretical literature on sudden stops, see, e.g., Mendoza (2010)). In this regard, studying how banks adjust their credit limits during a sudden stop is important by itself. The Credit Register provides not only the outstanding balance of a firm at a bank for a loan type (as we have been using previously in our baseline estimates), but also the associated credit limit at the same level. To have an even finer inference for the supply side effects, we now discard the credit relationships for which outstanding balance is equal to the credit limit (which one may call as spot loans). We focus on the remaining, the ones for which outstanding credit balance is less than the committed limit, and study how banks adjust their credit limits for this sample. Table 6 presents the results. A firm borrowing from at least two banks, the bank with 10% higher foreign funding reduces its credit limit to a given firm by 7.6% more strongly (column 1). Moreover, shorter maturity banks or banks that pay a higher premium on their foreign funding reduce their credit limits significantly more strongly by 5.5% and 8.9%, respectively (column 2). When we additionally include bank surprise, however, maturity and price lose their significance and economic relevance. A more surprised bank reduces its credit limit to a given firm by 14%. From columns (4) to (10), we include bank characteristics in interaction with bank foreign funding one-by-one. Previous results are by and large robust. Well capitalized, less liquid or smaller banks reduce their credit limits less. In addition, we find that banks with higher NPL ratio decrease their credit limits less, potentially implying that such banks engage in loan evergreening (e.g. to help firms meet their loan obligations by not decreasing credit limits as strongly as other banks). Moreover, banks that provide a larger share of loans to a firm appear to reduce their credit limit more strongly. Moreover, supply side adjustments on credit limits are felt differently across different firms (Figure 4). In the majority of the cases, we find that banks reduce their credit limits more strongly for smaller firms (in terms of asset size or employment), more indebted firms, and non-exporters. 18
  22. Banks ’ maturity or price of their foreign funding do not seem to play a statistically significant role. A more surprised bank reduces its credit limits additionally more strongly for smaller and more indebted firms. Interestingly, the results also suggest that more surprised banks reduce their credit limits more strongly for exporters. 4.3. Do Bank Types Matter? Next, we explore whether different types of banks adjust their credit supply differently. There are four types of banks in our dataset: State, private domestic, foreign, and development banks. For each type, we define an indicator variable that is equal to 1 if the bank is of that type, and 0 otherwise. We include bank type indicator variables in levels and in interaction with bank foreign funding (Table 7). Results suggest that higher ex-ante reliance on foreign funding matters the most for private domestic banks, and to some degree, for state banks. State banks (compared to other types of banks) do not change their supply of credit in an economically relevant way if they were highly exposed to the sudden stop. However, they make adjustments on their credit limits: they reduce credit limits to a firm more strongly if they have higher ex-ante foreign funding. Private domestic banks, on the other hand, reduce their credit supply strongly (e.g. a private domestic bank with 10% higher foreign funding reduces its supply of a similar type of loan to a given firm by 36%, a value substantially higher than our baseline estimates). They do, however, make smaller adjustments on their credit limits (by 6.9%). Regarding foreign banks, while they seem to reduce their credit supply more (in levels), they reduce their credit supply less if they are ex-ante more exposed to global liquidity. Finally, development banks reduce their supply significantly less compared to other types of banks (both in levels or in the interaction with foreign funding). 4.4. External Margin: New Lending and Termination Another margin of adjustment for banks in response to the sudden stop is to decide whether to start a lending relationship with a new client (firm) that they were not previously working with, or whether to terminate an existing relationship with an existing client after the sudden stop. In short, we label these margins as “new lending” and “termination”, respectively. To study these margins, we simply change the dependent variable in our baseline specification, and use (i) Nb f ,post , an indicator variable for new lending that takes a value 1 if bank b was not working with firm f in September 2008 but establishes a new lending relationship with the firm in 19
  23. September 2009 ; or (ii) Tb f ,post , an indicator variable for termination that takes a value 1 if bank b was working with firm f in September 2008 but terminates that relationship with the firm in September 2009. Similar as above, we also incorporate “Bank Foreign Funding” in interaction with other bank controls and the strength of the relationship.17 Note that we now aggregate the data set to bank-firm level (rather than the bank-firm loan-level as in the baseline model), since bank-firm relationship entails different loan types.18 Table 8 presents the results for the new lending margin. A bank with 10% higher foreign funding is 4.7% less likely to a establish a new lending relationship with a firm. This magnitude is economically relevant, given the fact that average probability of a bank starting a new lending relationship with a firm is 16% during this period. Moreover, banks with lower maturity or higher premium on their foreign funding are less likely to start a new lending relationship with a firm (by 1.2% and 2.2%, respectively). When we further control for bank surprise, the maturity dimension remains qualitatively robust, but not so for the price dimension. Moreover, a surprised bank is 3.2% additionally less likely to start a new lending relationship with a firm. Lastly, inline with the internal margin, well-capitalized, less liquid or smaller banks appear more likely to start working with new firms. On the termination margin, we find that banks with higher reliance on foreign funding are, on average, less likely to terminate their existing relationship with a given firm (Table 9, column 1). Combining this result with the new lending margin, we can infer that banks with ex-ante higher foreign funding are on average reluctant to change the set of firms they are working with, and that these banks may find it potentially costly to terminate their existing relationships with firms. On the price, maturity and surprise dimensions, we find that banks with lower maturity or higher premium on their foreign funding, or those that are more surprised, are more likely to terminate their existing relationship with a firm. Moreover, well capitalized, liquid, or more profitable banks appear more likely to terminate their existing relationships with firms. This result by and large resonates with our previous results on the internal margin. Other bank characteristics do not seem to matter in a robust and statistically and economically significant way. 17 For the new lending analyses, we naturally exclude R b f , pr e , the strength of the bank-firm relationship, since it does not exist. 18 For example, a bank may terminate its existing relationship with a firm for a certain loan type, while continuing its relationship through a different loan type. 20
  24. 4 .4.1. Bank Types Moreover, different types of banks behave differently at the external margin. As shown in Table 10, state, domestic private, or foreign banks are less likely to establish a new lending relationship with a firm. Development banks behave otherwise: they are more likely to start a new lending relationship with a firm (see columns (1) to (5)). On the termination margin (columns (6) to (10)), it is mostly the domestic banks that terminate their relationships with firms. Moreover, it appears that foreign banks are much less likely to terminate their relationships with firms. This may suggest that foreign banks potentially find it more costly to re-establish their relationship with a firm after it is terminated. 4.4.2. Firm Heterogeneity The external margin matters differently for different types of firms (Figures 5 and 6). Figure 5, in particular, shows that it is less likely for a bank with higher foreign funding to establish a new lending relationship with a smaller, younger firm, or with a firm with higher short-term debt-to-total debt ratio, or with an non-exporting firm. In other words, such firms appear to find it increasingly difficult to have access to bank credit during the sudden stop. Moreover, banks with shorter time-to-maturity or higher premium on their foreign funding are, in general, additionally less likely to start a new lending relationship with a small, young, more indebted, and more short-term indebted firm. In a similar vein, we find that more surprised banks are less likely to start working with a small firm (in terms of number of employees but not particularly so when we consider firm total assets), and on average, with more indebted firms. Interestingly, we also find that, more surprised banks are less likely to start a new lending relationship with an exporting firm (though, the difference is mild). Figure 6 presents the termination margin. By and large consistent with our previous results, a bank with higher reliance on global liquidity is on average less likely to terminate its existing relationship with small, young, indebted, or non-exporting firms. Other dimensions on the bank exposure suggest the following: Banks with shorter maturity foreign funding are more likely to terminate its relationship with more indebted firms (as one would expect), yet, are also less likely to terminate its relationship with smaller, younger or non-exporting firms. Moreover, banks with higher premium on their foreign funding too behave differentially across different firms: these banks are more likely to terminate their relationships with smaller firms (in terms of total assets), 21
  25. firms with higher short-term debt , or non-exporting firms. Lastly, more surprised banks are more likely to end their relationship with smaller, more short-term indebted or non-exporting firms. To summarize, the results suggest that smaller, younger, more indebted or non-exporting firms are on average find it more difficult to have new access to bank credit during the sudden stop. For the termination margin, the evidence is mixed. Moreover, shorter maturity or higher premium on bank foreign funding, as well as higher degree of surprise about the reversal in external funding conditions seem to play a role, and in most cases, amplify the effects. 4.5. Further Discussions on Maturity, Price and Surprise Dimensions So far, the evidence suggests that banks with shorter time-to-maturity or higher premium on their foreign funding reduce their supply of credit to a firm more strongly (and differentially across different firms). A potential reason, as we have discussed in the Introduction, is that such banks may find it more difficult to roll over their foreign wholesale debt during the sudden stop. Indeed, existing literature shows in different contexts how banks with shorter term funding face increasing roll-over risks during liquidity shortfalls, and present evidence for how banks’ refinancing needs may lead to inefficient outcomes during such episodes (Stein, 2012; Segura and Suarez, 2017). A potential alternative channel behind our results is related to loan portfolio risks. Namely, our exposure metrics, maturity, price or surprise, may be systematically related to riskiness of loan portfolios. For example, banks that are more optimistic about the external funding conditions (i.e. those that are more surprised) may ex-ante hold riskier loan portfolios prior to the sudden stop. To shed light on these points, we change the dependent variable in our baseline specification (see equation (1) or column (8) of Table 2) with (i) an indicator variable, If b , that is equal to 1 if firm f that is granted a loan in September 2008 by bank b is large (i.e. it has total assets above the median), and 0 otherwise; and similar indicator variables for firm age, debt-to-equity ratio, short-term debt-to-total debt ratio, and export status; or (ii) an indicator variable, D f b , that takes a value 1 if firm f that has granted a loan in September 2008 by bank b has defaulted on a loan at bank b in the past 12 or 24 months, and 0 otherwise; or (iii) an indicator variable, F f b , that takes a value 1 if firm f that has granted a loan in September 2008 by bank b has defaulted on a loan at bank b in the next 12, 24, or 36 months. The latter two analyses, besides providing additional insights and serving as a robustness check, enables us to exploit the universe of firms.19 19 Firm characteristics (e.g. size, indebtedness, etc.) are available for a subset of the universe of firms (the CBRT’s 22
  26. Table 11 presents the first set of results . Columns (1) to (6) provide the first set of results. Columns (7) to (12) further consider within bank-type variations (by including bank-type fixed effects) for robustness. Results suggest weak evidence for whether a firm borrowing from a bank with shorter maturity or higher premium on its foreign funding is riskier (in the sense of being smaller, younger, more indebted, more short-term indebted, or non-exporter). When we further consider within bank-type variations (columns (7) to (12)), the results are in general stronger statistically. A firm that is granted a loan by a shorter maturity bank is more likely to be larger and short-term indebted. Moreover, a firm that is granted a loan by a higher premium bank is more likely to be smaller and younger. For the surprise dimension, there is strong and robust evidence that a firm that is granted a loan by a more optimistic bank about the external funding conditions is more likely to be smaller, younger, more indebted, more short-term indebted, or non-exporter. Table 12 presents the second and third sets of results. Columns (1) and (2) show that a firm that has borrowed in September 2008 from a bank with shorter maturity on its foreign funding is less likely to be a ‘risky’ firm (as given by the positive and significant estimated coefficient for the interaction of bank foreign funding and maturity). On the other hand, banks with higher premium on their foreign funding appear to have granted credit to riskier firms. Moreover, more optimistic banks prior to the sudden stop appear to have extended credit to riskier firms. We obtain similar results when we control for bank types (columns (6) and (7)). Moreover, firms that were granted a loan in September 2008 by a bank with higher premium on its foreign funding, or by a bank that were more optimistic about the external funding conditions prior to the sudden stop, are more likely to default on a loan at the bank in the near future (columns (8) to (10)). Note also that our baseline results (Table 3) show that a bank with a higher reliance on global liquidity reduces its credit supply more strongly, if the bank is more liquid or larger. We have argued previously that such banks may hold riskier loan portfolios. Indeed, we find supportive evidence in this regard (Table 12 (continued)). In particular, we report the coefficient estimates for bank capital adequacy, liquidity ratio and size (in interaction with bank foreign funding). Table 12 (continued) shows that a firm that is granted a loan from a highly exposed and more liquid or larger bank is more likely to have recently defaulted on a loan at the bank (columns 6 and 7). Such Company Accounts Database). Note also that a finer specification would be to include all firms that have applied for a loan (regardless of whether they have been accepted or rejected). Since we do not have data on loan applications, we cannot explore this question. Since firms that are granted loans are more likely to be less risky (compared to firms that are rejected, which we do not observe), our risk results may be a lower bound. 23
  27. firms are also more likely to default in the near future (columns 8 and 9). In sum, our findings lend support to the view that the reduction in credit supply by banks that rely more on short-term foreign funding is mainly related to their difficulty in rolling over their foreign funding during the sudden stop rather than such banks working with riskier firms. For the premium, we find supportive evidence for both views (roll-over risks and riskier loan portfolio). For the surprise dimension, the results are in general stronger and provide robust evidence that banks that were more optimistic about external funding conditions are more likely to be working with riskier firms prior to the sudden stop. 4.6. Time Dimension Another margin of adjustment for banks during the sudden stop period is to decide when and at what strength to reduce credit supply to a given firm, or for example, when to terminate an existing relationship with a firm. In our previous analyses, we use the period of September 2008 – that corresponds to the Lehman Brothers collapse– to September 2009 –that by and large coincides with the lowest growth in aggregate domestic credit after September 2008–. In this section, we relax our assumption on the time period. In particular, we keep September 2008 as our reference month, start with the period September 2008 - January 2009, and study successively larger time periods. This analysis, besides serving as a robustness check, helps trace how the sudden stop hit the domestic credit supply cross-sectionally over time. In particular, we change the dependent variable in our baseline specification with ∆Yb f a,“t+1”−pre , where pre corresponds to September 2008 (the reference month), and “t + 1” corresponds to January 2009, February 2009, ..., or August 2010. We end at August 2010 –that corresponds to one year after our baseline end period– in order to avoid interfering with other macroeconomic developments or with the period of frequent use of macroprudential policies in Turkey after 2010 (Kara, 2016).20 Y denotes the (log) outstanding credit balance of firm f at bank b for loan type a (as in the baseline), the associated (log) credit limit (as we studied in Section 4.2), or the new lending or termination margins (as we studied in Section 4.4). Figure 7 presents the results. The figure suggests that our results are qualitatively robust to using different end periods. Moreover, the sudden stop continues to affect credit supply conditions 20 For instance, for horizons beyond mid 2010, banks with different degrees of exposure to the sudden stop in September 2008 might be affected differentially from such policy actions. 24
  28. beyond September 2009 (at both internal and external margins), with its effect losing its pace after March 2010. Furthermore, being more surprised about the sudden stop matters for a bank’s credit supply decisions in a hump shaped manner (at both internal and external margins). A more surprised bank reduces its credit supply the most (of a similar type of loan to a given firm) by October 2009, and is much less likely to establish a new lending relationship with a firm in October 2009. After October 2009, these effects fade away gradually. For the termination, on the other hand, the highest effect of being surprised is observed in September 2009, our baseline end-period, after which it exhibits a relatively flat pattern. 5. Conclusion We show that taking into account the multidimensional aspect of banks’ exposure to a sudden stop of capital flows is important to fully understand the impact of a sudden stop on the domestic credit supply. We show that banks that ex-ante rely more on shorter maturity foreign funding or pay a higher premium reduce their supply of credit more strongly after the sudden stop. Moreover, banks’ greater optimism about the external funding conditions is followed by a sharper reduction in credit supply if these expectations are not realized and external funding conditions got reversed. We then show evidence for why banks with ex-ante shorter maturity or higher premium on their foreign funding, or banks with ex-ante more optimistic expectations, pave way to more amplified credit supply dynamics after the sudden stop. Our results also show that the reduction in credit supply is felt significantly differently across different firms. Smaller, younger, indebted, or non-exporter firms experience a stronger drop in credit supply. Moreover, these effects are further amplified by banks that rely more on shortterm foreign funding or by banks that are more surprised about the reversal in external funding conditions. We then show that the reduction in bank credit supply is binding, as firms cannot easily substitute the loss in bank credit with other forms of debt, and eventually they reduce their investment and employment more. Our results underline the importance of designing policies that encourage banks to rely more on core funding, seek longer maturity funding as well as encourage banks to be more prudent in lending during buoyant times. 25
  29. References Agarwal , S., B. W. Ambrose, and C. Liu (2006). Credit Lines and Credit Utilization. Journal of Money, Credit, and Banking 38(1), 1–22. Allen, F. and D. Gale (2007). Understanding Financial Crises. Oxford: Oxford Unviersity Press. Berger, A. and G. Udell (1995). Relationship Lending and Lines of Credit in Small Firm Finance. Journal of Business 68(3), 351–382. Calvo, G. (1998). Capital Flows and Capital-Market Crises: The Simple Economics of Sudden Stops. Journal of Applied Economics 1(1), 35–54. Calvo, G., A. Izquierdo, and L. Mejia (2008). Systemic Sudden Stops: The Relevance of Balance Sheet Effects and Financial Integration. NBER Working Paper No.14026. Cetorelli, N. and L. S. Goldberg (2012). Banking Globalization and Monetary Transmission. Journal of Finance 67(5), 1811–1843. di Giovanni, J., S. Kalemli-Ozcan, M. F. Ulu, and Y. S. Baskaya (2018). International Spillovers and Local Credit Cycles. NBER Working Paper No. 23149. Diamond, D. and R. Rajan (2006). Money in a Theory of Banking. American Economic Review 96(1), 30–53. Eichengreen, B. and P. Gupta (2016). Managing Sudden Stops. World Bank Policy Research Working Paper, No.7639. Fendoglu, S., E. Gulsen, and J.-L. Peydro (2018). Global Liquidity Cycles and the Effectiveness of Domestic Monetary Policy: Any Constraint on the Interest Rate Transmission? Working Paper. Ioannidou, V., S. Ongena, and J.-L. Peydro (2015). Monetary Policy, Risk-Taking, and Pricing: Evidence from a Quasi-Natural Experiment. Review of Finance 19(1), 95–144. Iyer, R., J.-L. Peydro, S. da Rocha-Lopes, and A. Schoar (2014). Interbank Liquidity Crunch and the Firm Credit Crunch: Evidence from the 2007-2009 Crisis. Review of Financial Studies 27, 347–372. Jimenez, G., S. Ongena, J. Peydro, and J. Saurina (2012). Credit Supply and Monetary Policy: Identifying the Bank Balance-Sheet Channel with Loan Applications. American Economic Review 102, 2301–2326. Kanatas, G. (1987). Commercial Paper, Bank Reserve Requirements, and the Information Role of Loan Commitments. Journal of Banking and Finance 11, 425–448. Kara, H. (2016). Turkey’s Experience with Macroprudential Policy. BIS Working Paper, No. 86q. Khwaja, A. I. and A. Mian (2008). Tracing the Impact of Bank Liquidty Shocks: Evidence from an Emerging Market. American Economic Review 98(4), 1413–1442. Mendoza, E. (2010). Sudden Stops, Financial Crises, and Leverage. American Economic Review 100(5), 1941–1966. Mendoza, E. and M. Terrones (2012). An Anatomy of Credit Booms and their Demise. Journal Economia Chilena (The Chilean Economy) 15(2), 4–32. Morais, B., J.-L. Peydro, J. Roldan-Pena, and C. Ruiz-Ortega (2018). The International Bank Lending Channel of Monetary Policy Rates and Quantitative Easing: Credit Supply, Reach-for-yield, and Real Effects. Journal of Finance, forthcoming. Ongena, S., J. Peydro, and N. van Horen (2015). Shocks Abroad, Pain at Home? Bank-Firm-Level Evidence on the International Transmission of Financial Shocks. IMF Economic Review 63(4), 698–750. Schnabl, P. (2012). International Transmission of Bank Liquidity Shocks: Evidence from an Emerging Market. Journal of Finance 67, 897–932. Segura, A. and J. Suarez (2017). How Excessive is Banks’ Maturity Transformation? Review of Financial Studies 30(10), 3538–3580. Stein, J. (2012). Monetary Policy as Financial-Stability Regulation. Quarterly Journal of Economics 127, 57–95. Temesvary, J., S. Ongena, and A. Owen (2018). A Global Lending Channel Unplugged: Does U.S. Monetary Policy Affect Cross-border and Affiliate Lending by Global U.S. Banks? Journal of International Economics 112(C), 50–69. 26
  30. 27 Independent Variables Firm Log Assets Firm Log Age Firm Debt-to-Equity Firm STDebt / Debt Firm Export / Sales Dependent Variables (Firm Level) Change in Bank Credit Change in Trade Credit Change in Total Credit Investment Change in Employment Number of banking relationship per firm Bank-Firm Level Control Strength of Bank-Firm Relationship Capital Adequacy Ratio Liquidity Ratio Size ROA NPL Ratio Surprise Independent Variables Bank-Level Variables Foreign Funding … Volume … Maturity … Cross-Border Interest Rate Premium Termination New Lending Dependent Variables Credit Growth Change in Credit Limit Variables 0.267 0.208 1.88 0.414 0.222 0.158 -0.002 -0.224 Mean 0.175 0.163 1.257 0.363 0 0 0 -0.016 Median 0.225 0.176 2.129 1.074 0.416 0.365 0.609 1.304 SD 0.149 0.082 0.417 0.087 0 0 -0.001 -0.616 25% The ratio of overseas sales to total sales The ratio of short-term liabiltiies to total liabilities The ratio of total liabilities to total equity Log of firm age (in years) Log of firm total assets Log change in total number of employees Log change in fixed capital Log change in total trade credit Log change in total bank+trade credit Log change in total bank credit Share of loan amount from bank b to firm f in firm f 's total bank loans in September 2008 Number of banks a firm is working with prior to sudden stop Log difference between the expected end-of-the year USD/TRY exchange rate (as of Sep'08) minus the realized exchange rate (positive values mean that the realized depreciation is higher than the expected) Non-performing loans (with an overdue past 90 days) to total credit Pre-tax net profit to total assets Natural logarithm of total assets 9.714 2.654 3.254 0.788 0.165 0.717 0.52 0.353 0.344 -0.025 0.354 4.449 15.193 1.97 0.802 0.16 9.626 2.773 1.633 0.916 0.002 -0.014 0.008 0.011 0.012 0 0.277 4 14.99 1.953 0.528 0.172 1.498 0.76 9.166 0.266 0.284 2.946 2.1 1.433 1.199 0.445 0.29 2.668 2.199 1.747 1.003 0.076 8.734 2.303 0.652 0.641 0 -0.315 -0.351 -0.232 -0.075 -0.129 0.105 2 13.501 1.296 0.032 0.149 Liquid assets (cash + receivables from the central bank + interbank money market + reverse receivables) 0.201 to total assets 0.332 repo 0.272 0.189 Total Equity to Total Risk-Weighted Assets Weighted average interest rate premium on non-core FX liabilities (over the closest-maturity LIBOR, for each currency (USD, EUR, GBP)) Weighted average maturity of non-core FX Liabilities Total outstanding volume of non-core FX liabilities Non-core Foreign Currency (FX) Liabilities =1 if a bank establishes a new lending relationship with a firm that it was not working with prior to sudden stop; 0 otherwise =1 if a bank terminates an existing relationship with a firm after the sudden stop; 0 otherwise Log change in the credit limit of bank b to firm f for credit lines of loan type a Log change in total outstanding credit provided by bank b to firm f in loan type a Definition Table 1: Summary Statistics 10.607 3.135 3.673 1 0.199 0.419 0.523 0.355 0.25 0.1 0.56 6 0.269 0.411 17.017 2.376 1.274 0.211 0.252 2.667 1.156 0 0 0.077 0.092 75% 9,857 9,857 9,857 9,857 9,857 9,003 9,749 9,857 9,851 9,557 470,816 470,816 31 31 31 31 31 31 31 31 31 438,872 548,453 312,955 470,816 N
  31. 28 -8 .0 -9.1 Yes Yes Yes No No No Yes Yes 470,816 0.265 (0.047) (0.045) No Yes Yes No No No Yes No 538,034 0.288 -0.913*** -0.805*** (2) -4.2 Yes Yes Yes No No (1) & (2) Yes Yes 470,816 0.266 -2.1 -4.2 -0.6 Yes Yes Yes No No (1) & (2) Yes Yes 470,816 0.266 (0.048) (0.046) Yes Yes Yes No No (1) & (2) Yes Yes 470,816 0.266 -0.061 (0.023) -0.211*** (0.022) (0.118) -2.945*** (5) 0.417*** (0.066) -1.382*** (4) 0.423*** (0.117) -2.964*** (3) -10.5 -17.9 Yes Yes Yes Yes Yes (1) & (2) Yes Yes 470,816 0.271 (0.081) -1.795*** (0.045) 1.045*** (1.618) 31.871*** (6) (1.644) (0.959) -6.6 -10.5 -9.0 -9.2 Yes Yes Yes Yes Yes Yes No Yes No Yes (1) & (2) & (3) (1) & (2) & (3) Yes Yes Yes Yes 470,816 470,816 0.267 0.271 -14.897*** (0.118) -0.907*** (0.042) 1.053*** (1.742) 32.374*** (8) -10.679*** (0.155) 0.239 (7) Notes: The dependent variable is the log change in outstanding credit provided by bank b to firm f for loan type a. For detailed definitions and summary statistics of the variables used in the estimations, see Table 1. "Yes" indicates that corresponding variables or fixed effects are included. "No" indicates that corresponding fixed effects are not included. Standard errors are clustered at firm level, and are given in parentheses. *** significant at 1%, ** significant at 5%, and * significant at 10%. … if time to maturity was shorter by 1 year … if cross-border interest premium was higher by 1% points … if the bank is more surprised (p75-p25) Change in the the outcome variable (in percentage terms) for a bank with 10% higher ex-ante Foreign Funding Number of banking relationship per firm>1 Bank Controls Strength of Bank-Firm Relationship Bank Foreign Funding * Bank Controls Bank Foreign Funding * Strength of Bank-Firm R. Maturity (1), Premium (2), Surprise (3) Firm FE Loan-type FE Observations R-squared … * Surprise … * Cross-Border Interest Rate Premium ... * Time to Maturity Bank Foreign Funding (1) Table 2: Baseline Results
  32. ... * Time to Maturity … * Strength of Bank-Firm Relationship … * NPL … * ROA … * Size … * Liquidity … * Capital Adequacy Bank Foreign Funding (0.093) -1.862*** (0.082) -1.367*** (6) (1.335) 18.537*** (7) Capital 18.2 Yes Yes (1) & (2) Yes Yes 470,816 0.266 Liquidity -24.2 Yes Yes (1) & (2) Yes Yes 470,816 0.266 Size -31.6 Yes Yes (1) & (2) Yes Yes 470,816 0.267 ROA 0.3 Yes Yes (1) & (2) Yes Yes 470,816 0.266 NPL 9.6 Yes Yes (1) & (2) Yes Yes 470,816 0.266 Strength 0.0 Yes Yes (1) & (2) Yes Yes 470,816 0.266 Yes Yes (1) & (2) Yes Yes 470,816 0.268 0.243 (0.190) -0.016 (0.183) (0.134) (0.078) (0.125) -1.043*** 0.771*** (0.102) (0.066) 0.260** 0.031 (0.043) (0.795) -1.065*** (0.873) -0.900*** -14.384*** -10.990*** (1.298) (0.206) -1.431*** (5) (0.941) (0.710) 13.482*** (4) 7.663*** (0.203) (3) 15.150*** 1.479*** (0.201) (2) -4.307*** (1) Table 3: Baseline Results: Bank Characteristics -10.5 -17.9 Yes Yes (1) & (2) Yes Yes 470,816 0.271 (0.118) -10.5 -9.0 -9.2 Yes Yes (1) & (2) & (3) Yes Yes 470,816 0.271 (1.644) -14.897*** (0.081) (0.042) -0.907*** -1.795*** 1.053*** (0.045) (0.171) 0.020 (0.154) -0.000 (0.133) -0.428*** (0.089) -1.856*** (1.022) -21.201*** (1.444) 7.550*** (1.742) 32.374*** (9)a 1.045*** (0.171) 0.011 (0.161) 0.281* (0.133) -0.408*** (0.086) -1.923*** (0.904) -25.443*** (1.582) 8.656*** (1.618) 31.871*** (8)a a For ease of comparison, column (8) replicates the column (6) of Table 2, and column (9) replicates the column (8) of Table 2. Notes: The dependent variable is the log change in outstanding credit provided by bank b to firm f for loan type a. In all columns, the sample is restricted to firms that have multiple banking relationships. "Yes" indicates that corresponding variables or fixed effects are included. Standard errors are clustered at firm level, and are given in parentheses. *** significant at 1%, ** significant at 5%, and * significant at 10%. … with high vs. low interaction term (p75-p25) … if time to maturity was shorter by 1 year … if cross-border interest premium was higher by 1% points … if the bank is more surprised (p75-p25) Change in the the outcome variable (in percentage terms) for a bank with 10% higher ex-ante Foreign Funding Bank Controls Strength of Bank-Firm Relationship Maturity (1), Premium (2), Surprise (3) Firm FE Loan-type FE Observations R-squared … * Surprise … * Cross-Border Interest Rate Premium 29
  33. Short-term Debt / Total Debt Debt-to-Equity Ratio Firm Log Age β1- β2 H0: β1= β2 vs. H1: β1≠ β2 -26.73*** Reject H0 13.85*** Reject H0 4.72 -4.97*** Reject H0 -2.25 Yes 9,857 0.089 (0.067) 0.035 (0.114) 0.104 (0.001) -0.004*** (0.022) -0.090*** (0.014) -0.164*** (0.363) 0.802** (0.288) -0.665** (3) -2.35*** Reject H0 -0.65 Yes 9,851 0.067 (0.056) 0.033 (0.056) 0.139** (0.002) 0.000 (0.018) -0.126*** (0.009) -0.054*** (0.322) 0.505 (0.261) -0.193 (4) Dependent Variable ∆ Total Debt ∆ Fixed Assets -0.04 Not reject H0 -0.58 Yes 9,557 0.065 (0.023) -0.013 (0.022) -0.047** (0.001) -0.000 (0.006) -0.040*** (0.004) -0.002 (0.098) -0.165* (0.087) -0.178** (5) ∆ Employment Notes: The dependent variable is percentage change in bank credit, trade credit, total debt, fixed assets or employment. "Bank Dependent" is a dummy variable that is equal to 1 if the ratio of bank credit to total debt for a firm is higher than the 2-digit sector median. Each column includes 2-digit sector fixed effects. Standard errors are clustered at 2-digit sector level, and presented in parentheses. *** Significant at 1%, ** significant at 5%, and * significant at 10%. (Percentage change in the outcome variable if the firm … … was bank dependent as opposed to being non-dependent) H0: Being bank dependent does not matter for a firm -11.13 Percentage change in the outcome variable for a bank dependent firm with higher exposure to foreign funding (p75-p25) Yes 9,749 0.086 (0.105) (0.112) Yes 9,003 0.096 0.051 0.075 -0.144 (0.117) (0.152) (0.002) (0.003) 0.403*** -0.004** (0.051) -0.005* -0.059 (0.037) -0.222*** (0.025) (0.025) (0.500) -0.245*** (0.811) -0.138*** -2.681*** (0.511) 4.578*** 1.385*** (0.653) (2) (1) -3.272*** ∆ Trade Credit ∆ Bank Credit NACE 2 Sector Fixed Effects Observations R-squared Export/Sales β1 β2 Bank Non-Dependent * Firm Exposure to Banks' Foreign Funding Firm Log Assets β1 Coeff. Table 4: Real Effects Bank Dependent * Firm Exposure to Banks' Foreign Funding 30
  34. 31 (0.067) -11.2 -12.2 No Yes Yes Yes Yes Yes No No No No No No Yes Yes No Yes 439,133 377,392 0.312 0.290 (0.063) -1.125*** -1.226*** (2) -10.1 3.2 -16.4 28.0 Yes Yes Yes No No (1) & (2) Yes Yes 377,392 0.294 (0.111) (0.085) Yes Yes Yes No No (1) & (2) Yes Yes 377,392 0.291 2.799*** 0.320*** (0.049) (0.305) (0.039) Yes Yes Yes No No (1) & (2) Yes Yes 377,392 0.293 (6) -18.4 0.7 Yes Yes Yes Yes Yes (1) & (2) Yes Yes 377,392 0.297 (0.299) 0.069 (0.219) 1.837*** (3.043) -10.954*** 14.909*** (5) 1.643*** (0.129) -2.021*** (4) 1.010*** (0.235) -6.641*** (3) (11.932) (6.890) -17.5 -8.5 -12.8 -28.0 Yes Yes Yes Yes Yes Yes No Yes No Yes (1) & (2) & (3) (1) & (2) & (3) Yes Yes No Yes 377,392 377,392 0.294 0.297 -45.231*** (0.235) -1.281*** (0.120) 0.845*** (5.210) 27.889*** (8) -28.302*** (1.357) 4.746*** (7) Notes: The dependent variable is the log change in outstanding credit provided by a domestic bank b to firm f for loan type a. "Yes" indicates that corresponding variables or fixed effects are included. "No" indicates that corresponding fixed effects are not included. Standard errors are clustered at firm level, and are given in parentheses. *** significant at 1%, ** significant at 5%, and * significant at 10%. … if time to maturity was shorter by 1 year … if cross-border interest premium was higher by 1% points … if the bank is more surprised (p75-p25) Change in the the outcome variable (in percentage terms) for a domestic bank with 10% higher ex-ante Foreign Funding Number of banking relationship per firm>1 Bank Controls Strength of Bank-Firm Relationship Bank Foreign Funding * Bank Controls Bank Foreign Funding * Strength of Bank-Firm R. Maturity (1), Premium (2), Surprise (3) Firm FE Loan-type FE Observations R-squared … * Surprise … * Cross-Border Interest Rate Premium ... * Time to Maturity Bank Foreign Funding (1) Table 5: Domestic Banks
  35. 32 -7 .6 Yes Yes No Yes Yes 312,955 0.308 (0.035) -5.5 -8.9 -0.9 -1.9 -14.1 Yes Yes (1) & (2) & (3) Yes Yes 312,955 0.320 (0.128) Yes Yes (1) & (2) Yes Yes 312,955 0.318 -0.101 (0.128) -0.274** (0.202) (0.192) (0.148) -0.729*** -0.110 (0.105) (0.231) -0.915*** (0.149) -0.261** -1.143*** (1.123) -1.299*** -14.890*** (0.925) (2.212) -17.919*** 2.730 (2.750) -0.918 (2.700) -22.825*** -0.198 (0.159) (0.090) (0.247) (0.112) -0.897*** 0.090 (3.423) 25.778*** (3) 0.548*** (2.467) -0.767*** 23.692*** (2) Capital 7.7 Yes Yes (1) & (2) Yes Yes 312,955 0.309 (0.846) 6.416*** (0.161) -2.511*** (4) Liquidity -50.4 Yes Yes (1) & (2) Yes Yes 312,955 0.316 (1.048) -22.794*** (0.233) 3.820*** (5) Size -30.6 Yes Yes (1) & (2) Yes Yes 312,955 0.312 (0.047) -0.870*** (0.799) 13.299*** (6) ROA -6.8 Yes Yes (1) & (2) Yes Yes 312,955 0.309 (0.079) -0.631*** (0.154) -0.118 (7) NPL 18.5 Yes Yes (1) & (2) Yes Yes 312,955 0.311 (0.143) 1.486*** (0.107) -2.407*** (8) Strength -2.6 Yes Yes (1) & (2) Yes Yes 312,955 0.309 (0.131) -0.584*** (0.063) -1.135*** (9) Yes Yes (1) & (2) Yes Yes 312,955 0.316 (0.132) -0.280** (0.173) 0.107 (0.112) -0.386*** (0.098) -0.268*** (1.158) -20.391*** (1.517) 2.131 (1.642) 8.081*** (10) Notes: The dependent variable is the log change in credit limit committed by bank b to firm f for loan type a. In all columns, the sample is restricted to firms that have multiple banking relationships. "Yes" indicates that corresponding variables or fixed effects are included. Standard errors are clustered at firm level, and are given in parentheses. *** significant at 1%, ** significant at 5%, and * significant at 10%. … with high vs. low interaction term (p75-p25) Change in the the outcome variable (in percentage terms) for a bank with 10% higher ex-ante Foreign Funding … if time to maturity was shorter by 1 year … if cross-border interest premium was higher by 1% points … if the bank is more surprised (p75-p25) Bank Controls Strength of Bank-Firm Relationship Maturity (1), Premium (2), Surprise (3) Firm FE Loan-type FE Observations R-squared … * Strength of Bank-Firm Relationship … * NPL … * ROA … * Size … * Liquidity … * Capital Adequacy … * Surprise … * Cross-Border Interest Rate Premium ... * Time to Maturity Bank Foreign Funding (1) Table 6: Do Banks with higher reliance on global liquidity reduce their credit limits?
  36. 33 State 1 .6 Yes Yes Yes Yes 470,816 0.266 Private -36.0 Yes Yes Yes Yes 470,816 0.267 Foreign 16.7 Yes Yes Yes Yes 470,816 0.267 Development 19.2 Yes Yes Yes Yes 470,816 0.270 (0.244) Yes Yes Yes Yes 470,816 0.267 (0.213) (0.036) -0.523** (0.026) 0.606*** -0.756*** -0.347*** (0.025) (0.019) State -23.3 Yes Yes Yes Yes 470,816 0.265 (0.014) 0.357*** (0.019) 0.417*** 0.486*** (0.371) (0.130) -2.336*** (0.060) 0.570*** (6) 0.170*** (0.314) (0.157) 1.768*** (0.184) 1.915*** 0.004 (0.192) (0.155) 1.665*** -7.390*** -3.601*** (0.197) (0.080) -2.456*** (0.184) (0.103) -1.953*** -3.108*** (0.074) (0.085) 0.164 -1.099*** -0.722*** (5) Dependent Variable Private -6.9 Yes Yes Yes Yes 470,816 0.262 (0.012) -0.024* (0.095) -0.699*** (0.046) -0.557*** Foreign 16.6 Yes Yes Yes Yes 470,816 0.261 (0.018) -0.262*** (0.131) 1.664*** (0.077) -1.523*** Development 34.3 Yes Yes Yes Yes 470,816 0.262 (0.144) -0.949*** (0.227) 3.431*** (0.054) -1.753*** Change in Credit Limit (Sep'08 - Sep'09) (7) (8) (9) Yes Yes Yes Yes 470,816 0.268 (0.176) -1.900*** (0.026) -0.928*** (0.017) -0.259*** (0.274) 3.150*** (0.111) 2.290*** (0.125) -2.835*** (0.139) -2.914*** (10) Notes: The dependent variable is the log change in outstanding credit (for columns (1) to (5)) or log change in credit limit (for columns (6) to (10)) provided by bank b to firm f for loan type a. For detailed definitions and summary statistics of the variables used in the estimations, see Table 1. "Yes" indicates that corresponding variables or fixed effects are included. "No" indicates that corresponding fixed effects are not included. "--" indicates that the respective fixed effect is inapplicable or already included in the wider set of fixed effects or variables. Standard errors are clustered at firm level, and are given in parentheses. *** significant at 1%, ** significant at 5%, and * significant at 10%. Change in the outcome variable (in percentage terms) for a bank with 10% higher ex-ante Foreign Funding … if the bank is State, Private, Foreign or Development Bank Bank Controls Strength of Bank-Firm Relationship Firm FE Loan Type FE Observations R-squared Development Bank Foreign Bank Private Bank State Bank … * Development Bank … * Foreign Bank … * Private Bank … * State Bank Bank Foreign Funding (1) Credit Growth (Sep'08 - Sep'09) (2) (3) (4) Table 7: Do Bank Types Matter?
  37. 34 (0.032) (0.042) (0.033) (0.030) -0.434*** (0.029) -0.321*** 0.031 (0.018) (0.017) -0.003 -0.494*** (0.221) -0.522*** -2.562*** (0.194) (0.293) (0.292) -3.804*** -0.844*** -0.219 (0.415) -5.152*** (0.017) (0.009) 0.103*** (0.009) -0.220*** 0.129*** (0.324) 8.680*** 0.124*** (0.314) (3) (5) (6) (7) (8) (9) (0.220) 2.246*** (0.046) (0.181) -1.215*** (0.045) (0.009) -0.273*** (0.153) (0.024) -0.027 (0.049) (0.028) 0.073*** (0.013) -0.381*** (0.172) -2.430*** (0.276) -0.271 (0.248) (0.022) (0.033) 0.049** -0.433*** (0.021) -1.264*** -0.523*** 3.686*** -0.775*** -0.859*** 6.289*** (4) -4.7 -1.2 -2.2 -1.3 1.0 -3.2 Capital 2.7 Liquidity -2.7 Size -9.5 ROA -0.2 NPL 0.6 Yes Yes Yes Yes Yes Yes Yes Yes Yes No (1) & (2) (1) & (2) & (3) (1) & (2) (1) & (2) (1) & (2) (1) & (2) (1) & (2) (1) & (2) Yes Yes Yes Yes Yes Yes Yes Yes Yes 548,453 548,453 548,453 548,453 548,453 548,453 548,453 548,453 548,453 0.384 0.392 0.392 0.389 0.389 0.390 0.389 0.389 0.391 (0.011) -0.477*** 8.565*** (2) Notes: The dependent variable is an indicator variable, called "New Lending", that takes a value 1 if a bank establishes a new lending relationship with a firm after the sudden stop (September 2009) that it has not been working with previously (September 2008). "Yes" indicates that corresponding variables or fixed effects are included. Standard errors are clustered at firm level, and are given in parentheses. *** Significant at 1%, ** significant at 5%, and * significant at 10%. … with high vs. low interaction term (p75-p25) Prob. of establishing a new lending relationship for a bank with 10% higher ex-ante Foreign Funding … if time to maturity was shorter by 1 year … if cross-border interest premium was higher by 1% points … if the bank is more surprised (p75-p25) Bank Controls Maturity (1), Premium (2), Surprise (3) Firm FE Observations R-squared … * NPL … * ROA … * Size … * Liquidity … * Capital Adequacy … * Surprise … * Cross-Border Interest Rate Premium ... * Time to Maturity Bank Foreign Funding (1) Table 8: External Margin: New Lending
  38. 35 -4 .9 Yes No Yes 438,872 0.411 (0.014) (0.037) 0.5 9.5 (5) (6) (7) (8) (9) (10) (0.443) 8.343*** (0.087) (0.185) 3.116*** (0.053) (0.013) 0.017 (0.209) (0.031) 0.336*** (0.064) (0.025) -0.162*** (0.023) (0.208) 0.054 (0.091) (0.278) -0.175 (0.036) -0.003 (0.033) -0.001 (0.022) 0.176*** (0.238) 2.526*** (0.524) 9.698*** (0.381) -2.130*** -1.278*** -0.772*** -1.141*** -0.386*** -0.516*** -5.849*** (4) 1.0 5.1 6.6 Capital 10.0 Liquidity 6.9 Size 0.6 ROA 3.6 NPL -2.0 Strength 0.2 Yes Yes Yes Yes Yes Yes Yes Yes (1) & (2) & (3) (1) & (2) (1) & (2) (1) & (2) (1) & (2) (1) & (2) (1) & (2) (1) & (2) Yes Yes Yes Yes Yes Yes Yes Yes 438,872 438,872 438,872 438,872 438,872 438,872 438,872 438,872 0.418 0.413 0.412 0.411 0.412 0.411 0.411 0.414 (0.443) Yes (1) & (2) Yes 438,872 0.417 8.534*** (0.397) (0.045) (0.051) 6.381*** 0.134*** -0.096* (0.040) (0.035) (0.025) 0.425*** 0.246*** (0.024) (0.333) 0.446*** (0.314) 0.493*** 4.233*** (0.488) (0.514) 5.576*** 4.242*** 4.619*** (0.547) 10.617*** (0.028) (0.016) 0.513*** (0.016) 0.952*** -0.102*** (0.420) -15.492*** (3) -0.054*** (0.403) -0.493*** -13.834*** (2) Notes: The dependent variable is an indicator variable, called "Termination", that takes a value 1 if a bank terminates an existing lending relationship with a firm after the sudden stop (September 2009). In all columns, the sample is restricted to firms that ex-ante was working with multiple banks. "Yes" indicates that corresponding variables or fixed effects are included.Standard errors are clustered at firm level, and are given in parentheses. *** Significant at 1%, ** significant at 5%, and * significant at 10%. … with high vs. low interaction term (p75-p25) Prob. of terminating an existing relationship for a bank with 10% higher ex-ante Foreign Funding … if time to maturity was shorter by 1 year … if cross-border interest premium was higher by 1% points … if the bank is more surprised (p75-p25) Bank Controls Maturity (1), Premium (2), Surprise (3) Firm FE Observations R-squared … * Strength of Bank-Firm Relationship … * NPL … * ROA … * Size … * Liquidity … * Capital Adequacy … * Surprise … * Cross-Border Interest Rate Premium ... * Time to Maturity Bank Foreign Funding (1) Table 9: External Margin: Termination
  39. 36 State -11 .7 Yes -Yes 548,453 0.390 Private -6.0 Yes -Yes 548,453 0.389 Foreign -9.0 Yes -Yes 548,453 0.389 Development 0.6 Yes -Yes 548,453 0.391 (0.033) Yes -Yes 548,453 0.390 (0.031) (0.007) -0.344*** (0.005) -0.278*** -0.149*** -0.006 (0.006) (0.004) State 14.1 Yes Yes Yes 438,872 0.416 (0.006) -0.061*** (0.005) -0.040*** -0.105*** (0.054) (0.060) 2.025*** (0.027) -0.625*** (6) 0.123*** (0.055) (0.035) 0.136** 1.143*** (0.036) (0.037) -0.471*** (0.033) -0.081** -1.083*** 0.162*** (0.043) (0.018) -1.080*** (5) (0.046) (0.018) -0.822*** (4) -1.303*** (0.015) (0.017) (3) -0.746*** -0.768*** -0.433*** (2) Dependent Variable Private 2.2 Yes Yes Yes 438,872 0.413 (0.006) -0.114*** (0.046) 0.562*** (0.019) -0.343*** (7) Foreign -18.9 Yes Yes Yes 438,872 0.416 (0.006) 0.258*** (0.047) -2.020*** (0.023) 0.129*** (8) Yes Yes Yes 438,872 0.413 (0.045) 0.505*** (0.078) -0.047 (0.023) -0.757*** (9) Development -7.9 Termination Yes Yes Yes 438,872 0.419 (0.047) 0.274*** (0.010) 0.387*** (0.008) 0.130*** (0.079) -0.376*** (0.047) -2.042*** (0.054) -0.118** (0.056) 1.932*** (10) Notes: The dependent variable is either "New Lending" (for columns 1 to 5) or "Termination" (for columns 6 to 10). "Yes" indicates that corresponding variables or fixed effects are included. "No" indicates that corresponding fixed effects are not included. "--" indicates that the respective fixed effect is inapplicable or already included in the wider set of fixed effects or variables. Standard errors are clustered at firm level, and are given in parentheses. *** Significant at 1%, ** significant at 5%, and * significant at 10%. The outcome variable (in percentage terms) for a bank with 10% higher ex-ante Foreign Funding … if the bank is State, Private, Foreign or Development Bank Bank Controls Strength of Bank-Firm Relationship Firm FE Observations R-squared Development Bank Foreign Bank Private Bank State Bank … * Development Bank … * Foreign Bank … * Private Bank … * State Bank Bank Foreign Funding (1) New Lending Table 10: External Margin: Do Bank Types Matter?
  40. 37 Yes Yes Yes Yes No (1) & (2) & (3) 55,531 0.004 (0.108) 0.191* (0.013) -0.002 (0.005) 0.007 (0.951) -1.466 (0.075) 0.075 (0.019) -0.006 (0.767) -0.609 (3) Yes Yes Yes Yes Yes (1) & (2) & (3) 55,531 0.009 (0.111) Yes Yes Yes Yes Yes (1) & (2) & (3) 55,531 0.022 0.360*** (0.108) (0.015) 0.529*** -0.034** (0.014) (0.005) (0.005) -0.060*** -0.019*** (0.987) -0.018*** -2.849*** (0.964) (0.082) -3.593*** 0.170** (0.080) (0.027) (0.025) 0.273*** 0.081*** 0.103*** 1.570* (0.931) 1.685* (8) (0.886) (7) Yes Yes Yes Yes Yes (1) & (2) & (3) 55,531 0.004 (0.112) 0.261** (0.015) -0.019 (0.005) 0.005 (0.994) -2.198** (0.083) 0.167** (0.027) 0.024 (0.943) -0.073 (9) Smaller (Assets) Smaller (Emp.) Younger Yes Yes Yes Yes No (1) & (2) & (3) 55,531 0.008 (0.107) (0.104) Yes Yes Yes Yes No (1) & (2) & (3) 55,531 0.021 0.214** 0.346*** -0.002 (0.012) -0.018 (0.005) (0.005) (0.012) -0.013*** -0.009** -1.199 (0.946) -1.408 (0.922) -0.018 (0.074) 0.026 (0.019) (0.018) (0.072) -0.000 -0.000 0.242 (0.742) 0.239 (0.651) (2) Yes Yes Yes Yes No (1) & (2) & (3) 55,531 0.006 (0.110) -0.247** (0.013) -0.023* (0.005) -0.001 (0.963) 0.184 (0.076) 0.162** (0.020) 0.094*** (0.778) -0.739 (5) Yes Yes Yes Yes No (1) & (2) & (3) 55,531 0.010 (0.108) 0.262** (0.013) 0.023* (0.005) -0.007 (0.951) -0.990 (0.075) -0.113 (0.019) 0.021 (0.767) -1.879** (6) Yes Yes Yes Yes Yes (1) & (2) & (3) 55,531 0.010 (0.113) -0.252** (0.015) -0.021 (0.005) 0.000 (1.007) 2.599*** (0.085) -0.024 (0.028) 0.012 (0.947) 1.621* (10) Yes Yes Yes Yes Yes (1) & (2) & (3) 55,531 0.006 (0.113) -0.385*** (0.015) 0.023 (0.005) 0.001 (1.006) 1.888* (0.085) -0.069 (0.027) 0.057** (0.953) -0.818 (11) Yes Yes Yes Yes Yes (1) & (2) & (3) 55,531 0.010 (0.112) 0.311*** (0.015) -0.016 (0.005) -0.004 (0.994) -2.016** (0.083) 0.054 (0.027) -0.002 (0.943) -3.474*** (12) Less Indebted Less ST. Indebted Non-Exporter Yes Yes Yes Yes No (1) & (2) & (3) 55,531 0.010 (0.109) -0.369*** (0.013) 0.004 (0.005) 0.002 (0.957) 3.622*** (0.075) -0.159** (0.020) -0.031 (0.765) 0.496 (4) Notes: The dependent variable in column (1) is an indicator variable that is equal to 1 if the firm that is granted a loan in September 2008 is smaller than the median, and is equal to 0 if greater than or equal to the median. For other columns, the indicator variables are defined similarly. Estimates are based on linear probability model (ordinary least squares). "Yes" indicates that corresponding variables or fixed effects are included. Standard errors are heteroskedasticity robust, and are given in parentheses. *** Significant at 1%, ** significant at 5%, and * significant at 10%. Bank Controls Strength of Bank-Firm Relationship Bank Foreign Funding x Bank Controls Bank Foreign Funding x Strength of Bank-Firm R. Bank Type F.E. Maturity (1), Premium (2), Surprise (3) Observations R-squared Surprise Cross-Border Interest Rate Premium Time to Maturity … * Surprise … * Cross-Border Interest Rate Premium ... * Time to Maturity Bank Foreign Funding Bank Controls Strength of Bank-Firm Relationship Bank Foreign Funding x Bank Controls Bank Foreign Funding x Strength of Bank-Firm R. Bank Type F.E. Maturity (1), Premium (2), Surprise (3) Observations R-squared Surprise Cross-Border Interest Rate Premium Time to Maturity … * Surprise … * Cross-Border Interest Rate Premium ... * Time to Maturity Bank Foreign Funding (1) Dependent Variable: Probability that the firm that is granted a loan in September in 2008 is Smaller (Assets) Smaller (Emp.) Younger Less Indebted Less ST. Indebted Non-Exporter Table 11: Further Discussions on Maturity, Price and Surprise Dimensions
  41. 38 Yes Yes Yes Yes Yes (1) & (2) & (3) 438,872 0.485 (0.070) Yes Yes Yes Yes Yes (1) & (2) & (3) 438,872 0.487 (0.048) (0.011) 1.479*** 1.022*** (0.007) (0.002) -0.200*** -0.113*** (0.002) (0.703) -0.035*** -0.061*** (0.475) (0.064) -15.468*** -12.263*** (0.044) (0.013) 1.071*** 0.662*** (0.010) (0.377) (0.307) 0.035*** (7) -8.220*** (6) -1.659*** 0.171*** Yes Yes Yes Yes No (1) & (2) & (3) 438,872 0.483 (0.557) Yes Yes Yes Yes No (1) & (2) & (3) 438,872 0.485 (0.401) (0.005) 2.971*** 2.567*** (0.005) (0.002) -0.060*** -0.001 (0.002) (0.423) -0.045*** -0.066*** (0.360) (0.030) -9.715*** -7.472*** (0.029) (0.015) 0.362*** (0.011) 0.077*** 0.083*** (0.297) (0.237) 0.188*** -2.984*** 2.281*** Yes Yes Yes Yes Yes (1) & (2) & (3) 438,872 0.566 (0.048) 0.736*** (0.007) -0.134*** (0.002) -0.071*** (0.467) -11.286*** (0.042) 0.636*** (0.011) 0.127*** (0.333) (8) -5.435*** Yes Yes Yes Yes No (1) & (2) & (3) 438,872 0.564 (0.496) 3.916*** (0.006) 0.018*** (0.002) -0.082*** (0.458) -4.729*** (0.038) -0.139*** (0.013) 0.200*** (0.302) 0.888*** Yes Yes Yes Yes Yes (1) & (2) & (3) 438,872 0.595 (0.052) 0.786*** (0.008) -0.107*** (0.002) -0.078*** (0.519) -12.564*** (0.048) 0.546*** (0.012) 0.200*** (0.354) (9) -3.722*** Yes Yes Yes Yes No (1) & (2) & (3) 438,872 0.593 (0.468) 5.032*** (0.006) 0.033*** (0.002) -0.087*** (0.452) -6.244*** (0.037) -0.177*** (0.013) 0.265*** (0.297) 2.288*** Yes Yes Yes Yes Yes (1) & (2) & (3) 438,872 0.594 (0.052) 0.863*** (0.008) -0.087*** (0.002) -0.084*** (0.514) -13.198*** (0.048) 0.482*** (0.012) 0.257*** (0.362) (10) -1.456*** Yes Yes Yes Yes No (1) & (2) & (3) 438,872 0.593 (0.453) 5.362*** (0.006) 0.047*** (0.002) -0.091*** (0.461) -7.055*** (0.038) -0.222*** (0.012) 0.298*** (0.295) 3.985*** Notes: The dependent variable is the probability of a firm that has granted a loan by bank b in September 2008 has defaulted on at least one loan at bank b during (1) previous 12 months; (2) previous 24 months; (3) next 12 months; (4) next 24 months; or (5) next 36 months. Columns (6) to (10) additionally include bank-type fixed effects. Estimates are based on linear probability model (ordinary least squares). "Yes" indicates that corresponding variables or fixed effects are included. Standard errors are clustered at firm level, and are given in parentheses. *** Significant at 1%, ** significant at 5%, and * significant at 10%. Bank Controls Strength of Bank-Firm Relationship Bank Foreign Funding x Bank Controls Bank Foreign Funding x Strength of Bank-Firm R. Bank Type F.E. Maturity (1), Premium (2), Surprise (3) Observations R-squared Surprise Cross-Border Interest Rate Premium Time to Maturity … * Surprise … * Cross-Border Interest Rate Premium ... * Time to Maturity Bank Foreign Funding Bank Controls Strength of Bank-Firm Relationship Bank Foreign Funding x Bank Controls Bank Foreign Funding x Strength of Bank-Firm R. Bank Type F.E. Maturity (1), Premium (2), Surprise (3) Observations R-squared Surprise Cross-Border Interest Rate Premium Time to Maturity … * Surprise … * Cross-Border Interest Rate Premium ... * Time to Maturity Bank Foreign Funding Dependent Variable: Probability that firm f that is granted a loan by bank b in September in 2008 has defaulted on a loan at bank b In the Past In the Future 12 months 24 months 12 months 24 months 36 months (1) (2) (3) (4) (5) Table 12: Further Discussions on Maturity, Price and Surprise Dimensions (The Universe)
  42. 39 Yes Yes Yes Yes Yes (1) & (2) & (3) 438,872 0.485 (0.025) Yes Yes Yes Yes Yes (1) & (2) & (3) 438,872 0.487 (0.019) (0.348) 0.362*** -0.019 (0.249) (0.496) 12.722*** 7.152*** (0.370) (0.377) (0.307) 3.712*** (7) -8.220*** (6) -1.659*** 3.375*** Yes Yes Yes Yes No (1) & (2) & (3) 438,872 0.483 (0.017) Yes Yes Yes Yes No (1) & (2) & (3) 438,872 0.485 (0.015) (0.372) 0.040** -0.261*** (0.262) (0.557) 10.068*** (0.401) 5.178*** 2.971*** (0.297) (0.237) 2.567*** -2.984*** 2.281*** Yes Yes Yes Yes Yes (1) & (2) & (3) 438,872 0.566 (0.020) 0.130*** (0.261) 8.502*** (0.401) 5.008*** (0.333) (8) -5.435*** Yes Yes Yes Yes No (1) & (2) & (3) 438,872 0.564 (0.019) -0.258*** (0.302) 5.242*** (0.496) 3.916*** (0.302) 0.888*** Yes Yes Yes Yes Yes (1) & (2) & (3) 438,872 0.595 (0.022) 0.040* (0.259) 6.538*** (0.410) 6.369*** (0.354) (9) -3.722*** Yes Yes Yes Yes No (1) & (2) & (3) 438,872 0.593 (0.018) -0.329*** (0.285) 3.423*** (0.468) 5.032*** (0.297) 2.288*** Yes Yes Yes Yes Yes (1) & (2) & (3) 438,872 0.594 (0.022) -0.077*** (0.259) 4.988*** (0.415) 6.762*** (0.362) (10) -1.456*** Yes Yes Yes Yes No (1) & (2) & (3) 438,872 0.593 (0.018) -0.411*** (0.277) 2.193*** (0.453) 5.362*** (0.295) 3.985*** Notes: The dependent variable is the probability of a firm that has granted a loan by bank b in September 2008 has defaulted on at least one loan at bank b during (1) previous 12 months; (2) previous 24 months; (3) next 12 months; (4) next 24 months; or (5) next 36 months. Columns (6) to (10) additionally include bank-type fixed effects. Estimates are based on linear probability model (ordinary least squares). "Yes" indicates that corresponding variables or fixed effects are included. Standard errors are clustered at firm level, and are given in parentheses. *** Significant at 1%, ** significant at 5%, and * significant at 10%. Bank Controls Strength of Bank-Firm Relationship Bank Foreign Funding x Bank Controls Bank Foreign Funding x Strength of Bank-Firm R. Bank Type F.E. Maturity (1), Premium (2), Surprise (3) Observations R-squared … * Size … * Liquidity ... * Capital Adequacy Bank Foreign Funding Bank Controls Strength of Bank-Firm Relationship Bank Foreign Funding x Bank Controls Bank Foreign Funding x Strength of Bank-Firm R. Bank Type F.E. Maturity (1), Premium (2), Surprise (3) Observations R-squared … * Size … * Liquidity ... * Capital Adequacy Bank Foreign Funding Dependent Variable: Probability that firm f that is granted a loan by bank b in September in 2008 has defaulted on a loan at bank b In the Past In the Future 12 months 24 months 12 months 24 months 36 months (1) (2) (3) (4) (5) Table 12: Further Discussions on Maturity, Price and Surprise Dimensions (The Universe) (Continued)
  43. Figures 0 2008m1 Log Change in Bank Credit between September 2008 and September 2009 -1 - .5 0 .5 1 -.2 Annual Growth in Aggregate Bank Credit .1 .2 .3 0 .2 .4 Annual Growth in Foreign Wholesale Funding .4 Figure 1: Domestic Credit vs. Bank Foreign Wholesale Funding 2008m7 2009m1 Time 2009m7 0 2010m1 .1 .2 .3 .4 .5 Foreign Wholesale Funding in September 2008 (a) (b) Note. The figure on the left (A) plots annual growth in total bank credit (solid line, left axis) and the annual growth in foreign wholesale funding (dashed line, right axis). Vertical solid lines correspond to September 2008 and September 2009. The figure on the right (B) plots log change in total bank credit from September 2008 to September 2008 against banks’ foreign wholesale funding (in proportion to their assets) in September 2008. The circles are weighted by bank size. Source. Central Bank of the Republic of Turkey and Banking Regulation and Supervision Agency of Turkey. 40
  44. Figure 2 : Banks’ Ex-ante Exposure to the Sudden Stop: Additional Dimensions 0 .1 .2 .3 .4 .5 Foreign Wholesale Funding in September 2008 1 .5 -1 -.5 0 .5 -1 -.5 0 .5 0 -.5 -1 Log Change in Bank Credit between September 2008 and September 2009 Degree of Surprise 1 Premium 1 Maturity 0 .1 .2 .3 .4 .5 Foreign Wholesale Funding in September 2008 0 .1 .2 .3 .4 .5 Foreign Wholesale Funding in September 2008 Note. The figure presents three additional dimensions on the ex-ante exposure to the sudden stop (besides the level of foreign wholesale funding). The blue lines are the benchmark (replicating Figure 1(B)). The ‘Maturity’ dimension (the left panel) suggests that banks that have shorter time to maturity on their foreign wholesale funding in September 2008 reduce their supply of credit more strongly during September 2008-September 2009 (dotted red line). Moreover, the effect is stronger for domestic banks (dashed black line). Similarly, the ‘Premium’ dimension (the mid panel) suggests that banks, particularly the domestic banks, reduce their supply of credit more strongly, if they ex-ante were paying a high cross-border premium (dashed black line). In Maturity and Premium dimensions, we regress time-to-maturity and the interest premium on foreign wholesale funding on bank characteristics (foreign wholesale funding-to-total assets, capital, liquidity, size, return-on-assets, and non-performing loans ratio), and take the residuals (see Section 2 for detialed definitions). We then define a dummy variable that is equal to 1 for a bank if the corresponding estimated residual is lower (higher) than the industry median for the maturity (premium) dimension. The ‘Surprise’ dimension (the right panel) suggests that banks that were ex-ante expecting a more valued domestic currency in December 2008 reduce their credit more strongly in the aftermath. Dotted red line: All banks; Dashed black line: Domestic banks. 41
  45. Log (Assets) Log(Employment) Log(Age) % Change in Credit Supply by higher Foreign W. Funding bank % Change in Credit Supply by higher Foreign W. Funding bank 1. 2. 8. 95 .40 .56 .71 .83 .00 .14 .30 .53 2 2 2 2 3 3 3 3 77 .40 .81 .09 .55 .88 .25 .70 .32 3 3 4 4 4 5 5 6 34 .90 .31 .64 .96 .30 .68 .19 .05 8 9 9 9 10 10 11 12 1. 2. 8. 95 .40 .56 .71 .83 .00 .14 .30 .53 2 2 2 2 3 3 3 3 77 .40 .81 .09 .55 .88 .25 .70 .32 3 3 4 4 4 5 5 6 34 .90 .31 .64 .96 .30 .68 .19 .05 8 9 9 9 10 10 11 12 95 .40 .56 .71 .83 .00 .14 .30 .53 2 2 2 2 3 3 3 3 1. 77 .40 .81 .09 .55 .88 .25 .70 .32 3 3 4 4 4 5 5 6 2. 34 .90 .31 .64 .96 .30 .68 .19 .05 8 9 9 9 10 10 11 12 8. 95 40 56 71 83 00 14 30 53 1. 2. 2. 2. 2. 3. 3. 3. 3. 77 40 81 09 55 88 25 70 32 2. 3. 3. 4. 4. 4. 5. 5. 6. 34 90 31 64 96 30 68 19 05 8. 8. 9. 9. 9. 10. 10. 11. 12. Figure 3: Reduction in credit supply borne differently across different firms? Notes. Horizontal axes correspond to each decile for each firm characteristic. “ ” denote the estimates for firms less than the decile ( βˆ1 in equation 2), and “◦” denote the estimates for firms higher than the decile ( βˆ2 in equation 2). For example, regarding the most upper left graph, firms with Log(Assets) lower than the first decile experience around 8.5% reduction in credit supply (shown with ), whereas firms with Log(Assets) higher than the first decile experience about 3% reduction in credit supply (shown with ◦). Statistically significant estimates are given with bordered symbols (at 0.05 level). For instance, firms with Log(Assets) higher than the 9th decile do not experience a statistically significant change in credit supply. % Change in Credit Supply by higher Foreign W. Funding bank -10 -8 -6 -4 -2 0 -7 -6 -5 -4 -3 -2 2 0 -2 -4 -6 ... shorter time-to-maturity bank ... shorter time-to-maturity bank ... shorter time-to-maturity bank -2 -3 -4 -5 -6 -6.5 -6 -5.5 -5 -4.5 -4 -5.1 -5 -4.9-4.8-4.7-4.6 ... higher premium bank ... higher premium bank ... higher premium bank 2 0 -2 -4 -6 .5 0 -1.5 -1 -.5 0 -1 -2 -3 -4 ... more surprised bank ... more surprised bank ... more surprised bank -8 -12 -11 -10 -9 -7 -8 -11 -10 -9 -6 -14 -12 -10 -8 42
  46. Debt /Equity S.T.Debt/Debt Export/Sales % Change in Credit Supply by higher Foreign W. Funding bank % Change in Credit Supply by higher Foreign W. Funding bank 0. 0. 0. 00 .00 .00 .00 .02 .07 .18 .37 .67 0 0 0 0 0 0 0 0 42 .59 .71 .80 .88 .95 .99 .00 .00 0 0 0 0 0 0 1 1 36 .75 .13 .53 .00 .59 .38 .80 .84 0 1 1 2 2 3 4 7 0. 0. 0. 00 .00 .00 .00 .02 .07 .18 .37 .67 0 0 0 0 0 0 0 0 42 .59 .71 .80 .88 .95 .99 .00 .00 0 0 0 0 0 0 1 1 36 .75 .13 .53 .00 .59 .38 .80 .84 0 1 1 2 2 3 4 7 00 .00 .00 .00 .02 .07 .18 .37 .67 0 0 0 0 0 0 0 0 0. 42 .59 .71 .80 .88 .95 .99 .00 .00 0 0 0 0 0 0 1 1 0. 36 .75 .13 .53 .00 .59 .38 .80 .84 0 1 1 2 2 3 4 7 0. 00 00 00 00 02 07 18 37 67 0. 0. 0. 0. 0. 0. 0. 0. 0. 42 59 71 80 88 95 99 00 00 0. 0. 0. 0. 0. 0. 0. 1. 1. 36 75 13 53 00 59 38 80 84 0. 0. 1. 1. 2. 2. 3. 4. 7. Figure 3: Reduction in credit supply borne differently across different firms? (Continued) Notes. Horizontal axes correspond to each decile for each firm characteristic. “ ” denote the estimates for firms less than the decile ( βˆ1 in equation 2), and “◦” denote the estimates for firms higher than the decile ( βˆ2 in equation 2). For example, regarding the most upper left graph, firms with Debt/Equity ratio lower than the first decile experience around 1.5% reduction in credit supply (shown with ), whereas firms with Debt/Equity ratio higher than the first decile experience about 3.5% reduction in credit supply (shown with ◦). Statistically significant estimates are given with bordered symbols (at 0.05 level). For instance, firms with Debt/Equity ratio lower than the first decile do not experience a statistically significant reduction in credit supply. % Change in Credit Supply by higher Foreign W. Funding bank -4 -3.5 -3 -2.5 -2 -1.5 4 2 0 -6 -4 -2 0 -1 -2 -3 -4 ... shorter time-to-maturity bank ... shorter time-to-maturity bank ... shorter time-to-maturity bank -5 -4.8-4.6-4.4-4.2 -4 -5 -4 -3 -2 -1 0 -5.2 -5 -4.8-4.6-4.4-4.2 ... higher premium bank ... higher premium bank ... higher premium bank 4 2 0 -2 -3 -2 -1 0 1 2 1 0 -1 -2 -3 ... more surprised bank ... more surprised bank ... more surprised bank -7 -8 -11 -10 -9 -10 -8 -6 -4 -2 0 -11 -10.5 -10 -9.5 -9 43
  47. Log (Assets) Log(Employment) Log(Age) % Change in Credit Limit by higher Foreign W. Funding bank % Change in Credit Limit by higher Foreign W. Funding bank % Change in Credit Limit by higher Foreign W. Funding bank 95 .40 .56 .71 .83 .00 .14 .30 .53 2 2 2 2 3 3 3 3 1. 77 .40 .81 .09 .55 .88 .25 .70 .32 3 3 4 4 4 5 5 6 2. 34 .90 .31 .64 .96 .30 .68 .19 .05 8 9 9 9 10 10 11 12 8. 1. 2. 8. 95 77 34 2. 3. 8. 40 40 90 56 2. 81 3. 31 9. 71 83 00 14 30 53 2. 2. 3. 3. 3. 3. 09 55 88 25 70 32 4. 4. 4. 5. 5. 6. 64 96 30 68 19 05 9. 9. 10. 10. 11. 12. 95 1. 77 2. 34 8. 40 2. 40 3. 90 8. 56 2. 81 3. 31 9. 71 2. 09 4. 64 9. 83 2. 55 4. 96 9. 00 3. 88 4. 0 .3 10 8 14 3. 25 5. .6 10 30 3. 70 9 .1 5. 11 53 3. 32 6. 5 .0 12 95 40 56 71 83 00 14 30 53 1. 2. 2. 2. 2. 3. 3. 3. 3. 77 40 81 09 55 88 25 70 32 2. 3. 3. 4. 4. 4. 5. 5. 6. 34 90 31 64 96 30 68 19 05 8. 8. 9. 9. 9. 10. 10. 11. 12. Notes. Horizontal axes correspond to each decile for each firm characteristic. “ ” denote the estimates for firms less than the decile ( βˆ1 in equation 2), and “◦” denote the estimates for firms higher than the decile ( βˆ2 in equation 2). Statistically significant estimates are given with bordered symbols (at 0.05 level). -15 -14 -13 -12 -11 -15 -14 -13 -12 -11 -15 -14 -13 -12 -11 Figure 4: Reduction in credit limits borne differently across different firms? ... shorter time-to-maturity bank ... shorter time-to-maturity bank ... shorter time-to-maturity bank 2 1 0 -1 -2 4 2 0 -2 -4 1 0 -1 -2 -3 ... higher premium bank ... higher premium bank ... higher premium bank -3 -4 -5 -6 -7 -2 -3 -4 -5 -6 -6 -5.5 -5 -4.5 -4 ... more surprised bank ... more surprised bank ... more surprised bank -5 -6 -7 -8 -9 -8 -7.5 -7 -6.5 -6 -8 -7 -6 -5 -4 44
  48. Debt /Equity S.T.Debt/Debt 00 .00 .00 .00 .02 .07 .18 .37 .67 0 0 0 0 0 0 0 0 0. 42 .59 .71 .80 .88 .95 .99 .00 .00 0 0 0 0 0 0 1 1 0. 36 .75 .13 .53 .00 .59 .38 .80 .84 0 1 1 2 2 3 4 7 0. 0. 0. 0. 00 42 36 0. 0. 0. 00 59 75 00 0. 71 0. 13 1. 00 02 07 18 37 67 0. 0. 0. 0. 0. 0. 80 88 95 99 00 00 0. 0. 0. 0. 1. 1. 53 00 59 38 80 84 1. 2. 2. 3. 4. 7. 00 0. 42 0. 36 0. 00 0. 59 0. 75 0. 00 0. 71 0. 13 1. 00 0. 80 0. 53 1. 02 0. 88 0. 00 2. 07 0. 95 0. 59 2. 18 0. 99 0. 38 3. 37 0. 00 1. 80 4. 67 0. 00 1. 84 7. 00 00 00 00 02 07 18 37 67 0. 0. 0. 0. 0. 0. 0. 0. 0. 42 59 71 80 88 95 99 00 00 0. 0. 0. 0. 0. 0. 0. 1. 1. 36 75 13 53 00 59 38 80 84 0. 0. 1. 1. 2. 2. 3. 4. 7. Figure 4: Reduction in credit limits borne differently across different firms? (Continued) Notes. Horizontal axes correspond to each decile for each firm characteristic. “ ” denote the estimates for firms less than the decile ( βˆ1 in equation 2), and “◦” denote the estimates for firms higher than the decile ( βˆ2 in equation 2). Statistically significant estimates are given with bordered symbols (at 0.05 level). Export/Sales % Change in Credit Limit by higher Foreign W. Funding bank % Change in Credit Limit by higher Foreign W. Funding bank % Change in Credit Limit by higher Foreign W. Funding bank -6 -14 -12 -10 -8 0 -5 -10 -15 -14 -13 -12 -11 -10 ... shorter time-to-maturity bank ... shorter time-to-maturity bank ... shorter time-to-maturity bank 10 5 0 -5 4 2 0 -2 -4 3 2 1 -2 -1 0 ... higher premium bank ... higher premium bank ... higher premium bank 2 0 -2 -4 -6 0 -2 -4 -6 0 -2 -4 -6 -8 ... more surprised bank ... more surprised bank ... more surprised bank -2 -4 -6 -8 0 -2 -4 -6 -8 -5 -6 -7 -8 -9 45
  49. Log (Assets) Log(Employment) Log(Age) 1. 2. 8. 79 64 20 30 2. 22 3. 76 8. 48 2. 69 3. 18 9. 2. 3. 9. 71 93 52 77 2. 38 4. 84 9. 94 2. 78 09 3. 14 5. 3. 5. 26 56 50 3. 21 6. 8 4 5 8 .1 0.5 1.0 1.8 1 1 1 4. 10 79 1. 64 2. 20 8. 30 2. 22 3. 76 8. 48 71 77 94 09 26 50 2. 2. 2. 2. 3. 3. 3. 69 93 38 78 14 56 21 3. 3. 4. 4. 5. 5. 6. 18 52 84 18 54 05 88 9. 9. 9. 10. 10. 11. 11. 79 30 48 71 77 94 09 26 50 1. 2. 2. 2. 2. 2. 3. 3. 3. 64 22 69 93 38 78 14 56 21 2. 3. 3. 3. 4. 4. 5. 5. 6. 20 76 18 52 84 18 54 05 88 8. 8. 9. 9. 9. 10. 10. 11. 11. Notes. Horizontal axes correspond to each decile for each firm characteristic. “ ” denote the estimates for firms less than the decile ( βˆ1 in equation 2), and “◦” denote the estimates for firms higher than the decile ( βˆ2 in equation 2). Statistically significant estimates are given with bordered symbols (at 0.05 level). 79 .30 .48 .71 .77 .94 .09 .26 .50 2 2 2 2 2 3 3 3 1. 64 .22 .69 .93 .38 .78 .14 .56 .21 3 3 3 4 4 5 5 6 2. 20 .76 .18 .52 .84 .18 .54 .05 .88 8 9 9 9 10 10 11 11 8. Figure 5: Probability of a bank establishing a new lending relationship with a firm differs across different firms? Prob(New Lending) by higher Foreign W. Funding bank Prob(New Lending) by higher Foreign W. Funding bank Prob(New Lending) by higher Foreign W. Funding bank -4 -5 -6 -7 -4 -5 -6 -7 -8 -6.4-6.2 -6 -5.8-5.6-5.4 ... shorter time-to-maturity bank ... shorter time-to-maturity bank ... shorter time-to-maturity bank -.8 -.6 -.4 -1.2 -1 -.8 -.6 -.4 -1.2 -1 -1.2-1.1 -1 -.9 -.8 ... higher premium bank ... higher premium bank ... higher premium bank 4.5 4 3.5 3 2.5 5 4.5 4 3.5 3 3.2 3.4 3.6 3.8 4 4.2 ... more surprised bank ... more surprised bank ... more surprised bank -4.5 -4 -3.5 -3 -2.5 -3.5 -4 -4.5 -4.1 -4 -3.9-3.8-3.7-3.6 46
  50. Debt /Equity S.T.Debt/Debt Export/Sales 00 .00 .00 .00 .01 .05 .15 .33 .64 0 0 0 0 0 0 0 0 0. 41 .59 .71 .81 .89 .96 .00 .00 .00 0 0 0 0 0 1 1 1 0. 33 .70 .08 .48 .95 .55 .36 .80 .92 0 1 1 1 2 3 4 7 0. 0. 0. 0. 00 41 33 00 0. 59 0. 70 0. 00 0. 71 0. 08 1. 0. 0. 1. 00 81 48 01 0. 89 0. 95 1. 05 0. 96 0. 55 2. 15 0. 00 1. 36 3. 0. 1. 4. 33 00 80 64 0. 00 1. 92 7. 00 0. 41 0. 33 0. 00 0. 59 0. 70 0. 00 00 01 05 15 33 64 0. 0. 0. 0. 0. 0. 0. 71 81 89 96 00 00 00 0. 0. 0. 0. 1. 1. 1. 08 48 95 55 36 80 92 1. 1. 1. 2. 3. 4. 7. 00 00 00 00 01 05 15 33 64 0. 0. 0. 0. 0. 0. 0. 0. 0. 41 59 71 81 89 96 00 00 00 0. 0. 0. 0. 0. 0. 1. 1. 1. 33 70 08 48 95 55 36 80 92 0. 0. 1. 1. 1. 2. 3. 4. 7. Notes. Horizontal axes correspond to each decile for each firm characteristic. “ ” denote the estimates for firms less than the decile ( βˆ1 in equation 2), and “◦” denote the estimates for firms higher than the decile ( βˆ2 in equation 2). Statistically significant estimates are given with bordered symbols (at 0.05 level). Prob(New Lending) by higher Foreign W. Funding bank Prob(New Lending) by higher Foreign W. Funding bank Prob(New Lending) by higher Foreign W. Funding bank Figure 5: Probability of a bank establishing a new lending relationship with a firm differs across different firms? (Continued) -7.5 -7 -6.5 -6 -5.5 0 -2 -4 -6 -8 -7 -6.5 -6 -5.5 -5 -4.5 ... shorter time-to-maturity bank ... shorter time-to-maturity bank ... shorter time-to-maturity bank -.6 -.8 -1 -1.2 0 -.5 -1 -1.5 -.4 -.6 -.8 -1 ... higher premium bank ... higher premium bank ... higher premium bank 6 5 4 3 2 4 3 2 1 0 4.5 4 3.5 3 ... more surprised bank ... more surprised bank ... more surprised bank -4.4 -4.2 -4 -3.8 -3.6 0 -1 -2 -3 -4 -4.3 -4.2 -4.1 -4 -3.9 47
  51. Log (Assets) Log(Employment) Log(Age) 1. 2. 8. 79 64 20 30 2. 22 3. 76 8. 48 2. 69 3. 18 9. 2. 3. 9. 71 93 52 77 2. 38 4. 84 9. 94 2. 78 09 3. 14 5. 3. 5. 26 56 50 3. 21 6. 8 4 5 8 .1 0.5 1.0 1.8 1 1 1 4. 10 79 1. 64 2. 20 8. 30 2. 22 3. 76 8. 48 71 77 94 09 26 50 2. 2. 2. 2. 3. 3. 3. 69 93 38 78 14 56 21 3. 3. 4. 4. 5. 5. 6. 18 52 84 18 54 05 88 9. 9. 9. 10. 10. 11. 11. 79 30 48 71 77 94 09 26 50 1. 2. 2. 2. 2. 2. 3. 3. 3. 64 22 69 93 38 78 14 56 21 2. 3. 3. 3. 4. 4. 5. 5. 6. 20 76 18 52 84 18 54 05 88 8. 8. 9. 9. 9. 10. 10. 11. 11. Notes. Horizontal axes correspond to each decile for each firm characteristic. “ ” denote the estimates for firms less than the decile ( βˆ1 in equation 2), and “◦” denote the estimates for firms higher than the decile ( βˆ2 in equation 2). Statistically significant estimates are given with bordered symbols (at 0.05 level). 79 .30 .48 .71 .77 .94 .09 .26 .50 2 2 2 2 2 3 3 3 1. 64 .22 .69 .93 .38 .78 .14 .56 .21 3 3 3 4 4 5 5 6 2. 20 .76 .18 .52 .84 .18 .54 .05 .88 8 9 9 9 10 10 11 11 8. Figure 6: Probability of a bank terminating an existing relationship with a firm differs across different firms? Prob(Termination) by higher Foreign W. Funding bank Prob(Termination) by higher Foreign W. Funding bank Prob(Termination) by higher Foreign W. Funding bank -3 -3.5 -4 -4.5 -4 -3.5 -3 -2.5 -2 -3.8 -3.6 -3.4 -3.2 -3 ... shorter time-to-maturity bank ... shorter time-to-maturity bank ... shorter time-to-maturity bank 1 1.5 2 2.5 .5 0 3 2.5 2 1.5 1 1.7 1.8 1.9 2 2.1 2.2 ... higher premium bank ... higher premium bank ... higher premium bank 4 3.5 3 2.5 2.6 2.8 3 3.2 3.4 4 3.5 3 2.5 ... more surprised bank ... more surprised bank ... more surprised bank 4.5 4 3.5 3 2.5 4 3.5 3 2.5 3.5 3.6 3.7 3.8 3.9 48
  52. 0 . 0. 0. 00 41 33 00 0. 59 0. 70 0. 00 0. 71 0. 08 1. 0. 0. 1. 00 81 48 01 0. 89 0. 95 1. 05 0. 96 0. 55 2. 15 0. 00 1. 36 3. 0. 1. 4. 33 00 80 64 0. 00 1. 92 7. 00 0. 41 0. 33 0. 00 0. 59 0. 70 0. 00 00 01 05 15 33 64 0. 0. 0. 0. 0. 0. 0. 71 81 89 96 00 00 00 0. 0. 0. 0. 1. 1. 1. 08 48 95 55 36 80 92 1. 1. 1. 2. 3. 4. 7. 00 00 00 00 01 05 15 33 64 0. 0. 0. 0. 0. 0. 0. 0. 0. 41 59 71 81 89 96 00 00 00 0. 0. 0. 0. 0. 0. 1. 1. 1. 33 70 08 48 95 55 36 80 92 0. 0. 1. 1. 1. 2. 3. 4. 7. Notes. Horizontal axes correspond to each decile for each firm characteristic. “ ” denote the estimates for firms less than the decile ( βˆ1 in equation 2), and “◦” denote the estimates for firms higher than the decile ( βˆ2 in equation 2). Statistically significant estimates are given with bordered symbols (at 0.05 level). 00 .00 .00 .00 .01 .05 .15 .33 .64 0 0 0 0 0 0 0 0 0. 41 .59 .71 .81 .89 .96 .00 .00 .00 0 0 0 0 0 1 1 1 0. 33 .70 .08 .48 .95 .55 .36 .80 .92 0 1 1 1 2 3 4 7 0. Figure 6: Probability of a bank terminating an existing relationship with a firm differs across different firms? (Continued) Debt/Equity S.T.Debt/Debt Export/Sales Prob(Termination) by higher Foreign W. Funding bank Prob(Termination) by higher Foreign W. Funding bank Prob(Termination) by higher Foreign W. Funding bank -3.8-3.6-3.4-3.2 -3 -2.8 -5 -4 -3 -2 -1 0 -4 -3.5 -3 -2.5 -2 ... shorter time-to-maturity bank ... shorter time-to-maturity bank ... shorter time-to-maturity bank 2.1 2.2 2 1.8 1.9 1 1.5 2 .5 0 2.2 2 1.8 1.6 ... higher premium bank ... higher premium bank ... higher premium bank 5.5 5 4.5 4 3.5 4 3 2 1 0 4.5 4 3.5 3 ... more surprised bank ... more surprised bank ... more surprised bank 3.4 3.6 3.8 4 4.2 4.4 4 3 2 1 0 3.23.43.63.8 4 4.2 49
  53. ... % Change in Credit Supply from September 2008 to t 0 -5 -10 -15 1 9m 0 20 -20 0 20 0m 1 20 3 5 0m 1 20 7 m 10 20 1 9m 0 20 3 9m 0 20 5 9m 0 20 7 9 1 3 5 7 11 m m m m m m 9m 09 09 10 10 10 10 0 0 0 0 0 0 0 2 2 2 2 2 2 20 ... more surprised bank 1 0m ... more surprised bank 1 20 ... higher premium bank 2 11 m 9 00 ... higher premium bank 9 9m ... shorter time-to-maturity bank 0 20 ... shorter time-to-maturity bank 7 9m 0 20 by a bank with 10% higher foreign funding 5 9m 0 20 by a bank with 10% higher foreign funding 3 9m % Change in Credit Limit from September 2008 to t Figure 7: Time Dimension: Change in Credit Supply Conditions by Months 10 0 -10 -20 -30 -40 50
  54. ... Probability of Establishing a New Lending Relationship with a firm at t 2 0 -2 -4 -6 1 9m 0 20 -8 0 20 0m 1 20 3 5 0m 1 20 7 m 10 20 1 9m 0 20 3 9m 0 20 5 9m 0 20 7 9 1 3 5 7 11 m m m m m m 9m 09 09 10 10 10 10 0 0 0 0 0 0 0 2 2 2 2 2 2 20 ... more surprised bank 1 0m ... more surprised bank 1 20 ... higher premium bank 2 11 m 9 00 ... higher premium bank 9 9m ... shorter time-to-maturity bank 0 20 ... shorter time-to-maturity bank 7 9m 0 20 by a bank with 10% higher foreign funding 5 9m 0 20 by a bank with 10% higher foreign funding 3 9m Probability of Terminating an Existing Relationship with a firm at t Figure 7: Time Dimension: Change in Credit Supply Conditions by Months (Continued) 10 5 0 -5 51
  55. Appendix A . CBRT Survey of Exchange Rate Expectations The Central Bank of the Republic of Turkey (CBRT) collects, in the first two weeks of each month, data on exchange rate expectations of financial institutions and large non-financial corporations. The question states “What does your institution expect the USD-Turkish lira exchange rate to be at the end of current month, at the end of the current year, and after 12 months?” We use bank-level expectations about the end-of-the-year exchange rate. For 19 banks which constitute a total of 83% of total credit outstanding in Turkey in September 2008, we have endof-the-year expectations measured in September 2008. For the remaining 8 banks, we observe end-of-the-year figures generally one or two months before or after September 2008. We followed a conservative route, and take a linear trend from the latest available expectation before September 2008 (usually August 2008) to the first available expectation afterwards (usually October or November 2008), and pick the September value. For the remaining 4 banks (covering about 5% of total credit outstanding), we take the median expectation. Given our conservative stand, the effect of ‘Surprise’ is biased towards zero, or in other words, our results are a lower bound for the true effect. Figure A.1 provides the histogram of the end-of-the-year expectations. Over 90% of banks were expecting a more valued domestic currency for the end of 2008 compared to the realized value, confirming that the reversal in external funding conditions is largely unexpected. Moreover, there is a substantial heterogeneity among banks in their expectations, with an interquartile range of 1.25 to 1.34 for the expected end-of-the-year USD/TRY exchange rate (whereas the realized exchange rate in December 2008 was 1.54). 52
  56. Figure A .1: Histogram of Bank-Level Expectations 0 Number of banks 5 10 Realized end-of-the-year USD/TRY Exchange Rate 1.2 1.3 1.4 1.5 1.6 End-of-the-year USD/TRY Exchange Rate Expectations 53 1.7
  57. Central Bank of the Republic of Turkey Recent Working Papers The complete list of Working Paper series can be found at Bank ’s website (http://www.tcmb.gov.tr) Dependence of “Fragile Five" and “Troubled Ten" Emerging Markets' Financial System to US Monetary Policy and Monetary Policy Uncertainty (Meltem Gülenay Chadwick Working Paper No. 18/17, October 2018) Measuring Financial Systemic Stress for Turkey: A Search for the Best Composite Indicator (Meltem Gülenay Chadwick, Hüseyin Öztürk Working Paper No. 18/16, October 2018) VAT Treatment of the Financial Services: Implications for the Real Economy (Fatih Yılmaz, İsmail Baydur Working Paper No. 18/15, October 2018) Net External Position, Financial Development, and Banking Crisis (Aytül Ganioğlu Working Paper No. 18/14, October 2018) Transportation Mode Choice and International Fragmentation of Production: Evidence from a Developing Country (Hülya Saygılı, Kemal Türkcan Working Paper No. 18/13, October 2018) Türkiye’de Reel Kur Hareketlerinin İhracat Üzerindeki Asimetrik Etkileri (Selçuk Gül Working Paper No. 18/12, Ekim 2018) The Impact of Monetary Policy Stance, Financial Conditions, and the GFC on Investment-Cash Flow Sensitivity (Selçuk Gül, Hüseyin Taştan Working Paper No. 18/11, July 2018) Inflation Dynamics in Turkey from a Bayesian Perspective (Fethi Öğünç, Mustafa Utku Özmen, Çağrı Sarıkaya Working Paper No. 18/10, July 2018) The Effect of Fed’s Future Policy Expectations on Country Shares in Emerging Market Portfolio Flows (Zelal Aktaş, Yasemin Erduman, Neslihan Kaya Ekşi Working Paper No. 18/09, March 2018) Evolution of the University Wage Premium in Turkey: 2004-2015 (Okan Eren Working Paper No. 18/08, March 2018) Multivariate Filter for Estimating Potential Output and Output Gap in Turkey (Selen Andıç Working Paper No. 18/07, February 2018) Quantifying Uncertainty and Identifying its Impacts on the Turkish Economy (Evren Erdoğan Coşar, Saygın Şahinöz Working Paper No. 18/06, February 2018) Forecasting Industrial Production and Inflation in Turkey with Factor Models (Mahmut Günay Working Paper No. 18/05, February 2018) Türkiye Ekonomisi için Güncellenmiş Doğrudan Çıktı Açığı Göstergesi (Evren Erdoğan Coşar Çalışma Tebliği No. 18/04, Şubat 2018) Türkiye İçin İthalat Talep Fonksiyonu (Olcay Yücel Çulha, Okan Eren, Ferya Öğünç Çalışma Tebliği No. 18/03, Şubat 2018) Export Behavior of Turkish Manufacturing Firms Under Crises (Aslıhan Atabek Demirhan, Hakan Ercan Working Paper No. 18/02, January 2018) Foreign Currency Borrowing, Exports and Firm Performance: Evidence from a Currency Crisis (Spiros Bougheas, Hosung Lim, Simona Mateut, Paul Mizen, Cihan Yalçın Working Paper No. 18/01, January 2018) The Empirical Content of Season-of-Birth Effects: An Investigation with Turkish Data (Huzeyfe Torun, Semih Tümen Working Paper No. 17/21, December 2017) Do Subsidized Export Loans Increase Exports? (Yusuf Emre Akgündüz, Süleyman Hilmi Kal, Huzeyfe Torun Working Paper No. 17/20, December 2017)