Capital Shocks, Renegotiation and
the United States Syndicate Loan Market
To my dear family:
Roman, Marcela, Patrik and Matej
for the love and support that made me the person I am;
1 Introduction
The US syndicated loan market is the primary source of funding for large businesses in America (Berk & DeMarzo, 2011). Around 90% of these loans are renegotiated before their original maturity date; the most cited reasons relate to the emergence of new information about the firm, its business or the economy (Roberts & Sufi, 2009). At first glance, this may seem puzzling – what benefit is there to renegotiate loans this frequently?
Dessen argued that banks can gain significant real control over management by imposing overly tight covenants at the origination of the loan, and then relaxing these throughout the lifetime of the loan, conditioned on desirable management behavior. Therefore, he argued that renegotiation is not solely used as a means for resuscitating a distressed business, but also as a monitoring tool that keeps borrowers in continuous check by the lenders (2005).
Aside from the large number of renegotiations, the US syndicated loan market faces additional specificity – demand pressures resulting from its syndicated character. Ivashina and Scharfstein argued that participation of multiple diverse entities in these syndicates expands the credit supply; but also it increases the market’s volatility due to shock- and sentiment-driven changes in the demand for loan participation (2010). These shocks restrict lead bank’s ability to originate further loans by forcing it to accept a larger stake in the deal (Ivashina & Scharfstein, 2010).
This thesis hypothesizes that the exposure to capital shock driven changes in credit supply is not limited to the loan origination; instead it expects the capital losses to affect loans through their lifetime – particularly during their renegotiation. Therefore, it is investigated whether capital shocks
to US syndicates affect loan renegotiation decision. If they do, then it is of interest whether these shocks affect non-distressed borrowers differently from those that are distressed. That is whether
there is difference in the shock’s effect on business resuscitating and business monitoring renegotiations.
The renegotiations in the US syndicated loan market were probit-modeled based on bank’s capital losses and access to funding, firm’s financial and business health as well as the deal details of each individual loan. The dataset, populating this model, spans from the early 2001 until the late 2013 and contains almost 10 000 separate loans extended to circa 2,300 large US corporations. The model was estimated by two separate techniques – the simple maximum likelihood estimator and the
Wooldridge’s method. The former is taken to be a base approach, due to its prevalence among other bachelor theses. However, its assumptions were not found to be supported by the data, thus the latter was offered as a more suitable approach.
Given the collected sample and the models applied a disparity between the distressed and non-distressed borrowers was observed. Firms in good financial health were shown to be significantly affected by the loan losses, while those with financial difficulties were not shown to face the same dynamics. Therefore, it is implied that there is a real risk of the renegotiation as a monitoring device to be backfiring on both, the lenders and the borrowers. That is, the initial loan conditions may be set supraoptimally tight given their thus far overlooked sensitivity to capital shocks.
The paper is split in six sections. First, the literature review presents stylized facts behind loan
renegotiation and syndication. Second, this paper’s methodology is discussed on theoretical grounds. Third, the results are presented and followed by the fourth – the concluding section. Finally two appendices, the first Technical and the second Data-oriented provide additional information on the model, techniques and the data.
2 Literature review
The US capital markets have demonstrated considerable cyclicality over the years, which some argued could stem from shocks to borrower’s collateral and the resulting inability to raise capital in the periods of economic hardship (Bernanke & Gertler, 1989). At the same time, shocks to bank capital may be affecting supply of lendable funds and thus deepening each credit cycle (Bernanke,
during the 2008 crisis; Ivashina and Schafterstein found that reductions in bank capital had adverse effects on lending (2010)while Duchin, Ozbas and Sensoy showed that firms with more collateral were better equipped to withstand the economic hardship of the financial meltdown (2010). In the past twenty-five years syndicated debt has emerged as an alternative to the original banking structure. It allowed lead banks, those that originate the debt contract, to share a portion of the borrower exposure with syndicate members against a fee. This development enabled banks to enhance their diversification by introducing within-syndicate risk sharing. This is, particularly, valuable if syndicate participants finance themselves in imperfect capital markets that become too costly in the event of an adverse capital shocks (Froot & Stein, 1998). However, participants need to weight this benefit against the cost of reduced monitoring and screening incentives (Ivashina, 2009). On the macro level, the introduction of syndicated lending allowed for, previously abstaining, non-bank entities such as pension, mutual and hedge funds to take part in the corporate loan lending, leading to an expansion of credit supply and increased volatility in the market (Ivashina &
Scharfstein, 2010).
Loan renegotiation is a process whereby existing loan is amended to suit changed borrower-lender situation. This paper distinguishes between renegotiations that are needed for borrower’s survival and those that are needed for its further development. The latter is a case of covenants becoming obsolete or overly restrictive as a result of changed borrower situation.
Loan renegotiations aimed at resuscitating business were in past shown to depend on agency costs stemming from information asymmetries between borrowers and lenders, debt features, bank characteristics, regulatory circumstances and the bank-firm business relationship (HassabElnaby, 2006). However, they were not shown to depend on syndication, bank structure or negative economic circumstances (HassabElnaby, 2006).
Furthermore, these findings were supported by Roberts in a recent study of renegotiations as a tool of dynamic ex-post loan agreement completion. He confirmed that borrower conditions are a
significant determinant of loan renegotiation, and argued this to be evidence of banks exercising their power over borrowers (2015). His findings support Dessen’s results that renegotiation is, indeed, used as monitoring tool that grants borrowers the de-facto rather than pro-forma power On the other hand, Ivashina and Scharfstein found syndication to be affecting both, the credit supply and its volatility in the US capital markets (2010). Furthermore, Ivashina and Scharfstein argued that credit supply during the 2008 financial crisis was significantly affected by capital shocks to banks (2009).
Implications of these findings are in contrast with Roberts’ and Dessen’s observations, if the supply of renegotiations is seen as a form of capital supply. This disparity has been partially addressed by Goldewski who studied the determinants of multiple loan renegotiations. He found bank financial health and economic conditions, contrary to Roberts’ findings affect probability of renegotiation (2014).
However, neither Goldewski, Roberts nor Dessen explicitly accounted for banks having potentially fundamentally different approach to distressed and non-distressed borrowers. Their samples included variables that were aimed at controlling for direct effects of firm’s financial health on the renegotiation outcome. However, they did not allow for fundamental differences between banks’ approach to the two groups. That is, underlying their results is an assumption of identical
probabilistic distribution across borrower financial health conditions.
3 Methodology
This section covers in turn this paper’s theoretical model, its specification and the statistical tests applied in the analysis. The estimation techniques and assumption tests are dealt with in the Technical Appendix, while the data is described in the Data Appendix.
3.1 The Probit Model
Given the binary character of the renegotiation outcome – it either happened or not – the entire class of, in parameters, linear models was judged unsuitable for the needs of this paper. Instead, nonlinear limited outcome logit and probit models were suggested as a more suitable alternative. These models are in many respects similar and for the purposes of this thesis they are hardly discriminable. However, probit, along with tobit and poisson are compatible with Wooldridge’s method (2005), which yields it more versatile and handy for panel data analysis. This holds particularly true for cases when heteroskedasticity is of concern. Thus the model is:
( | ⃗)
In detail, the model consists of the left-hand side renegotiation dummy which is probabilistically depending on the loan and security explanatory variables ( ); bank-, firm and macro-level
controls as well as time and industry fixed effect controls. It is, further, allowed to have a base-line intercept . The details of these broad categories are addressed in the following subsection.
This model heavily relies on the assumption that the probability of renegotiation given the right hand side variables is normally distributed. However, if the underlying distribution is not normal, the model and its estimation methods are not applicable and the results are of no use. This assumption was based on the central limit theory and a large sample size. Nevertheless it remains the most important assumption of this paper.
3.2 The Model Specification
In this section theoretical justifications of the selected variables are presented. Their detailed working definitions are further elaborated on in the Data Appendix.
3.2.1 Dependent and Explanatory Variables
In line with the central research question, the probit model relates a set of explanatory variables and controls to a renegotiation binary variable. The latter has the value of 1 if a renegotiation on a given loan took place in the given year and 0 otherwise. Thus the predicted probability relates to a positive outcome – renegotiation agreement. Due to the scope and scale of this research the dummy does not further discriminate among different renegotiations, these are considered to be potential future research areas.
The explanatory variables are to model banks’ loan loss capital shocks, which are taken to mean both, the total losses on syndicate member’s loans as well as losses on different categories of loans, such as agricultural, bank, credit card, industrial and other loans. Furthermore, these are
complemented with losses on bank’s treasury and government securities. These items are further defined in the data appendix.
The loan losses are broadly split into five loan categories: losses on agricultural, bank, credit card, industrial and other loans; and two categories of securities: treasury and government securities. This distinction was dictated by data availability rather than theory – the variables that are used in this paper optimize the variable span against time span.
3.2.2 Control Variables
The most prevalent causes of loan renegotiation are, for both the borrower and lender, an access to novel information, changes in credit conditions, emergence of new investment opportunities, changes in collateral’s value and macroeconomic developments (Roberts & Sufi, 2009). Ivashina added to this list bank’s access to funding (Ivashina & Scharfstein, 2009) as well as different loan characteristics such as the inclusion of covenants (Ivashina, 2009) or the performance pricing grid (Ivashina & Sun, 2010). Thus, the controls are split in three subsections, the bank-, firm- and economy- level controls and fixed effects.
On the bank’s side explanatory variables cover bank’s access to funding and the details of the renegotiation deal.
Access to Funding:
Banks, to different degree, face short term constraints on funding that limit their ability to renegotiate, particularly in times of economic hardship (Ivashina & Scharfstein, 2009). Thus lower exposure to potential short-term liquidity constraints is expected to be positively influencing bank’s decision to renegotiate.
Covenants:
Covenants increase lender’s security by restricting borrower’s actions that could potentially undermine its recovery value to the former (Berk & DeMarzo, 2011).
Markup:
Markup represents the compensation the syndicate receives for entering into a renegotiated contract. Thus, it is going to positively affect their willingness to do so.
Deal Amount:
Larger deals associate with larger exposure to the syndicate banks and thus greater contribution to their risk exposure. Thus the deal amount is expected to negatively affect syndicate’s willingness to renegotiate.
Loan type:
Credit lines and term loans are fundamentally different products that both, the bank and the customer approach differently (Berk & DeMarzo, 2011). Thus the model takes this difference into account by identifying term loans.
Pricing mechanism:
Performance pricing grid has shown to have an effect on credit borrower’s availability and its price (Ivashina & Sun, 2010; Ivashina, 2009; Roberts & Sufi, 2009).
Furthermore, firm-level variables focus on the probability of default, and on the loss given default, in line with industry-wide risk management’s practice.
Probability of default:
Z-score:
Altman’s z-score predicts the probability of firm’s bankruptcy in the coming two years (1968). It was shown to have a significant effect on syndicate’s decision-making on loan
renegotiation in a sample of European banks (Godlewski, 2014). Furthermore, it affects credit availability in general (Godlewski, 2014; Ivashina & Sun, 2010; Ivashina, 2009; Roberts & Sufi, 2009).
Leverage:
Leverage relates to the incentive structure firm’s management and shareholders face with respect to their short-term decisions under threat of bankruptcy. Also, it is to account for potential negative effects of debt overhang. Overall, it affects negatively bank’s willingness to extend credit to a potential borrower (Godlewski, 2014; Ivashina & Sun, 2010; Ivashina, 2009; Roberts & Sufi, 2009).
Tobin’s Q:
Tobin’s Q proxies market’s perception of firm’s ability to create value for its investors (Brainard & Tobin, 1968).Thus it is a measure of the businesses long term outlooks and it is expected to have a positive effect on firm’s decision.
Return on Assets:
ROA is a historic, backward-looking, measure of the same ability to create value for investors that Tobin’s Q estimates going forward. ROA complements the former in shedding more light on the borrower’s position in firms’ life cycle. Thus, it has a positive effect on firm’s decision (Ivashina & Sun, 2010; Ivashina, 2009; Roberts & Sufi, 2009).
Loss given default:
Total Assets:
Total Assets in the event of bankruptcy improve chances of bank recovering part of its capital. Thus they have positive effect on bank’s decision (Godlewski, 2014; Ivashina & Sun, 2010; Ivashina, 2009; Roberts & Sufi, 2009).
Asset Tangibility:
Due to the tricky valuation of intangible assets, and their complicated re-sale, asset tangibility is expected to negatively affect loan renegotiation. This was confirmed by past research (Roberts & Sufi, 2009).
Industry:
Durchin, Ozbas and Sensoy found prevalent and economically significant cross-industry differences in reliance on external finance (2010). To control for these differences, industry-fixed effects based on Fama’s and French’s industry classification (Fama & French, 1997) was used. The remaining text refers to this variable as the ‘industry fixed effects’.
Economic conditions:
Stock Market:
The general economic conditions in which renegotiation take place affects both the lenders as well as the borrower. The most of borrower’s assets are declining in value and its business
is very likely to suffer. The lenders suffer from negative shocks to their loan portfolios and low capital availability. Consequently, periods of economic turmoil are negatively associated with borrowers’ chances to renegotiate (Godlewski, 2014; Ivashina & Sun, 2010; Roberts & Sufi, 2009).. The stock market was significantly correlated with time fixed effects and the remaining control variables. Its inclusion was causing non-concavity of the Wooldridge-estimated probit models, and thus it was omitted in every model specification presented in section 4.
3.3 Statistical Testing
This paper aims at answering whether capital shocks to syndicate affect renegotiation, and if so then whether this effect differs between borrowers of different financial health. Such set of questions is too broad and diverse for a single hypothesis. It has to be broken down into testable claims. First, the question of a general effect is broken down between the total loan losses and loan losses by industry. That is, first, the item I (simple MLE) and the item II (Wooldridge’s method) investigate whether the total syndicate capital loan losses have an effect on the renegotiation probability; then potential diversification present in bank’s portfolio is filtered away by breaking down its loan
portfolio into different loan types. Isolating the investigated effects is done by testing a set of diverse model specifications that are then tested with the standard t-tests and the more sophisticated multiple coefficient Wald f-tests.
This preliminary analysis is then, in item III, expanded to account for any fundamental differences between the treatments the distressed and non-distressed borrowers receive. Effectively, it is done by splitting the sample in three subsamples, based on the 1.89 and 3.00 Altman’s z-score thresholds (Altman’s z-score<1.89 is distressed, >3 is non-distressed, between is the gray area). This approach is preferred over an inclusion of slope- and intercept-shifting interaction-term dummies since the dummies would push up already potentially high multicollinearity, restricting the applicability of
coefficient tests. Moreover, the split allows for a more holistic and fundamental differences between the three subsamples, thus it provides more information and is easier to interpret.
Finally, due to the severe and unprecedented character of the 2008 financial meltdown, it is possible that it has introduced additional breaks in the data that are not accounted for by simple time fixed effects. That is, it is plausible that during the financial meltdown not only the base probability of renegotiation (the intercept) changed, but also syndicate’s decision and sensitivity to different shocks. If this is indeed the case, one should expect the identity of distribution (part of assumption 1, see technical appendix) to be broken. Therefore, item IV repeats the analysis of item III on two subsamples –pre January 2007 and post January 2007 – in order to control for this potential complication.
In the course of these tests, the simple t-tests and the more sophisticated Wald f-tests were used. The t-tests are limited to testing hypotheses of simple one coefficient type:
Hypotheses pair 1:
These tests follow standard student-t distributions. Therefore they rely on the correctness of the coefficients’ standard errors. Should these be inflated by multicollinearity, the test proportionally to the inflation loses power.
The Wald F-tests are more versatile, they allow for linear combinations of tested coefficients. Such as simple tests of joint significance:
Hypotheses pair 2:
Or for testing equality of two coefficients:
Hypotheses pair 3:
In effect a combination of multiple Wald f-tests allows for singling out of multiple compound coefficient effects. This is particularly useful if there is a reason to believe that standard errors were inflated by multicollinearity.
They belong to the class of restricted model tests. That is they are, in a finite sample, equivalent to the likelihood ratio or Lagrange multiplier tests. Given this equivalence, Wald F-tests were chosen out of STATA programming convenience.
Wald tests’ statistic for multiple parameters test is:
( ̂ ) * (̂) +
( ̂ )
where the denotes the tested matrix of parameters; the ̂ is a matrix of their estimates and ̂ is an estimate of their covariance matrix. Wald f-test’s statistic is Chi-squared distributed.
4 Results
This section is split into four items, which in turn address the question of whether capital shocks to a syndicate affect their decision to renegotiate borrower’s loans, and if they do, then whether there is difference between loans to distressed and non-distressed borrowers in the response to these shocks. The first item contains an application of the simple probit estimation technique and the reasons why continuing the Wooldridge’s method is preferred. The second item applies the
alternative, Wooldridge’s method. It first constructs the model, and then it expands the model across losses in different loan categories such as agricultural loans, loans to banks, credit card loans,
industrial loans or the other loans. Then it splits these loans into the nonperforming and nonaccrual types. The third item investigates sensitivity of their findings to different creditor financial health conditions. That is, it aims to establish whether borrowers of all financial health conditions face the same consequences of syndicate’s capital shock. The fourth item controls for potential fundamental differences in the banking industry between the pre- and post-2007 periods. Lastly, all is summarized in the final subsection.
4.1 Item I: Simple ML Probit Estimation
Specifications 1 to 3 of table 1 test the set of control variables. Due to significant multicollinearity among them, the most appropriate test is the Wald F-test on the entire model specification. If its p-value is significant, the specification as such collectively explains the probability of renegotiation. All control variable specifications tested significant.
Specifications 4 to 7 of the table 1 test whether total loan losses and or losses on treasury and government securities affect probability of renegotiation. First, statistically significant negative coefficients on any one of the three variables in any of the columns 4 through 7 would lead to rejection of the individual null hypotheses that the given capital shock has individually no effect on the probability of renegotiation. The alternative that the shock has negative effect would be then accepted. However, none of the three coefficients appearing in the columns 4 to 7 tested significant;
therefore, the null was not rejected. Second, this is supported by the set of Wald F-tests reported at the bottom of the table. Particularly, the two under specification 7 are of interest. On their basis, one fails to reject both, the null that all coefficients on explanatory variables are collectively equal to zero and that government and treasury securities coefficients are jointly equal to zero. Therefore, the alternative hypothesis that total loans would be affecting loan renegotiation cannot be accepted. This can be both, either due to this model’s unsuitability for the purposes of modeling renegotiation or due to diversification on the side of banks. The former reason is a subject of the following table; the latter is discussed in the next item.
Table 1: Total Loan Loss Probit Specifications
The assumption tests of the total loans probit model in specification 7 in table 1 are presented in table 2. The normality of renegotiation probability, the assumptions one, two, three and four are addressed in succession. Each includes a short description of its null as well as the p-value of the observed statistic. Statistically significant p-values (marked with *) lead to rejection of the null hypothesis and acceptance of the alternative.
(1) (2) (3) (4) (5) (6) (7) 1.994 2.454 2.14 -1.378 -1.492 -1.522 0.497 0.703 0.758 -0.477 -0.493 -0.494 -3.182 -2.841
year fixed effect (12) yes yes yes yes yes yes yes
industry fixed effect (11) yes yes yes yes yes
0.289** 0.569*** 0.568*** 0.565*** 0.597*** 0.595*** 0.597*** -0.107 -0.139 -0.139 -0.14 -0.146 -0.146 -0.146 10.652*** 10.184*** 10.362*** 10.712*** 10.268*** 10.634*** 10.532*** -1.79 -2.179 -2.196 -2.217 -2.362 -2.371 -2.368 -1072.775 -19200 -19800 -17600 -9483.629 -8417.297 -13600 -2808.547 -23206.23 -23235.9 -26578.92 -25499.38 -31478.8 -32048.56 0 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0 0 0 0 0 0 0 0.061 0.081 0.075 0.072 0.071 0.07 0.07 -0.043 -0.05 -0.051 -0.051 -0.052 -0.052 -0.052 -0.336*** -0.306*** -0.296*** -0.295*** -0.306*** -0.306*** -0.307*** -0.039 -0.045 -0.045 -0.045 -0.047 -0.047 -0.047 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0 0 0 0 0 0 0 0.135** -0.123 -0.113 -0.114 -0.138* -0.140* -0.144* -0.049 -0.064 -0.065 -0.065 -0.067 -0.067 -0.067 0.035 0.016 0.029 -0.059 -0.041 -0.038 -0.115 -0.117 -0.118 -0.121 -0.121 -0.122 0 0.001 0 0.005 0.004 0.004 -0.007 -0.007 -0.008 -0.008 -0.008 -0.008 0.164* 0.021 0.019 0.019 0.019 0.017 -0.074 -0.1 -0.1 -0.102 -0.102 -0.102 0.159 0.244 0.234 0.241 0.23 0.232 -0.212 -0.218 -0.218 -0.224 -0.224 -0.224 -0.506 -0.524 -0.518 -0.593 -0.577 -0.581 -0.303 -0.307 -0.308 -0.315 -0.316 -0.316 0.06 0.080* 0.082* 0.085** 0.087** 0.088** -0.032 -0.032 -0.032 -0.033 -0.033 -0.033 0.391*** 0.422*** 0.419*** 0.459*** 0.456*** 0.459*** -0.113 -0.119 -0.12 -0.122 -0.122 -0.122 -0.001*** -0.001*** -0.001*** -0.001*** -0.001*** -0.001*** 0 0 0 0 0 0 -0.002 -0.013 -0.014 -0.01 -0.011 -0.011 -0.013 -0.014 -0.014 -0.015 -0.015 -0.015 -1.775*** -1.826*** -1.857*** -1.922*** -1.624*** -1.718*** -1.623*** -0.034 -0.457 -0.465 -0.468 -0.478 -0.482 -0.49 pseudo R squared 0.0452 0.0573 0.0626 0.0626 0.0611 0.0616 0.0618 N 19853 16697 16697 16662 15959 15954 15954 F-TESTS: (2) 3.80 (2) 3.54 0.1495 0.1702 (1) 2.09 (1) 1.09 (2) 3.80 (3) 5.06 0.1479 0.2966 0.1495 0.1678 (20) 334.47 (29) 353.78 (41) 386.86 (41) 386.64 (41) 361.40 (42) 363.84 (43) 365.08 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0000*** Legend: * p<0.05, ** p<0.01, *** p<0.001
dependent variable: renegotiation dummy (1 if it happened)
Wa ld F -test s probit
government securities shocks total loan capital shocks tresury securities shocks
C o ntr o l V ar ia bl es & F ix ed E ff ec ts ** * C o ntr o l V ar ia bl es & F ix ed E ff ec ts ** * C o ntr o l V ar ia bl es & F ix ed E ff ec ts Ex pl an ato ry V ar . Constant
bank secutities (tresury&gov) markup (basis points) covenant (1 if included) ln of borrower's ln of borrower's squared tangibility
roa
mean bank time deposit ratio mean bank leverage deal amount normalised bank deal amount normalised firm pricing grid (1 if yes) term loan (1 if yes)
all independent variables the entire model profitability tobins_q leverage (borrower) loans (borrower) zscore
Table 2:
The overall conclusion of the assumption tests is that while the underlying probit model is applicable, the simple ML estimation technique is not. There is significant statistical evidence in favor of non-independence among model variables, existence of omitted variables and error non-normality. Therefore, an alternative approach was suggested – the Wooldridge method (see technical appendix).
4.2 Item II: Wooldridge ML Probit Estimation
Table 3, just as table 1 tests the control variables (column 1 to 3) and the total loan losses with losses on securities (4 to 7). Just as before, the controls are due to multicollinearity best tested with Wald tests, which reject the null that they would have collectively no effect.
The three explanatory variables (total loan losses, losses on government securities and losses on treasury securities) in columns 4 to 7 are, first t-tested for significance of their individual coefficients. Statistically significant negative coefficients would lead to acceptance of this papers’ alternative hypothesis that loan losses negatively affect syndicate’s decision to renegotiate. But, none of them tests significant. Furthermore, the Wald F-tests don’t produce statistically significant statistics, neither. Therefore, it cannot be rejected that the total loan losses and losses on both types of securities, both individually and in conjunction don’t explain loan renegotiation.
This result is the same in table 1. Interestingly, the numerical results are barely different. The only change is the slightly lower Wald F-statistic of the entire model for all of the specifications. This result is to be expected if the bank diversification is the chief driver of the results in table 1 and 3.
Therefore, the individual loan types are tested next.
Assumption Statistical Test Null Hypothesis Observed Statistic p-value Decision Significance probit model applicability Pearson Goodness-of-Fit models fit Pearson chi2(15897) = 15430.30 0.9959 model fit not rejected at 5% Assumption 1: iid Wooldridge Test for Autocorrelation no first-order autocorrelation F(1, 2723) = 75364.873 0.0000*** fnull rejected, first order autocorrelation accepted at 0.1% Assumption 2: no multicollinearity Variance Inflation Factor Analysis Mean VIF = 141.82 multicollinearity suspected
no ommited variables z(_hat) = -0.50 0.614
z(_hatsq) = -2.88 0.004 ** at 1% error normality Chi2(2) = 10.1550 0.0062** at 1% Legend: * p<0.05, ** p<0.01, *** p<0.001
Assumption 4: no heteroscedasticity and autocorrelation
Lagrange Multiplier Test for the Normality of the Residuals
null rejected, existance of ommited variables accepted
Assumption 3: no ommited variables Link Test For Model Specification
Table 3:
Table 4 tests individual types of loan losses by industry. Specification 8 is the base specification that contains only the total loan losses, it serves as a comparison for the reader.
Specifications 9 to 13 of table 4 test, whether loans to banks, agricultural loans, credit card loans, industrial or other loans, in separation explain renegotiation. Statistically significant negative coefficients are in conflict with the null hypothesis and lead to acceptance of this paper’s alternative that the effect is negative. Loans to banks, agricultural loans and other loans test insignificant thus
(1) (2) (3) (4) (5) (6) (7) 1.994 2.453 2.14 -1.378 -1.492 -1.522 0.497 0.703 0.758 -0.477 -0.493 -0.494 -3.182 -2.84
year fixed effect (12) yes yes yes yes yes yes yes
industry fixed effect (11) yes yes yes yes yes
0.560*** 0.569*** 0.568*** 0.565*** 0.597*** 0.595*** 0.597*** -0.125 -0.139 -0.139 -0.14 -0.146 -0.146 -0.146 9.818*** 10.184*** 10.362*** 10.713*** 10.268*** 10.633*** 10.532*** -1.954 -2.179 -2.196 -2.217 -2.362 -2.371 -2.368 -6501.99 -19200 -19800 -17600 -9483.743 -8417.662 -13600 -18623.41 -23205.95 -23236.07 -26579.11 -25499.14 -31478.53 -32048.29 0.000* 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0 0 0 0 0 0 0 0.092* 0.081 0.075 0.072 0.071 0.07 0.07 -0.045 -0.05 -0.051 -0.051 -0.052 -0.052 -0.052 -0.343*** -0.306*** -0.296*** -0.295*** -0.306*** -0.306*** -0.307*** -0.04 -0.045 -0.045 -0.045 -0.047 -0.047 -0.047 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0 0 0 0 0 0 0 -0.163** -0.123 -0.113 -0.114 -0.138* -0.140* -0.144* -0.058 -0.064 -0.065 -0.065 -0.067 -0.067 -0.067 0.035 0.016 0.029 -0.059 -0.041 -0.038 -0.115 -0.117 -0.118 -0.121 -0.121 -0.122 0 0.001 0 0.005 0.004 0.004 -0.007 -0.007 -0.008 -0.008 -0.008 -0.008 0.164* 0.021 0.019 0.019 0.019 0.017 -0.074 -0.1 -0.1 -0.102 -0.102 -0.102 0.159 0.244 0.234 0.241 0.23 0.232 -0.212 -0.218 -0.218 -0.224 -0.224 -0.224 -0.506 -0.524 -0.518 -0.593 -0.577 -0.581 -0.303 -0.307 -0.308 -0.315 -0.316 -0.316 0.06 0.080* 0.082* 0.085** 0.087** 0.088** -0.032 -0.032 -0.032 -0.033 -0.033 -0.033 0.391*** 0.422*** 0.419*** 0.459*** 0.456*** 0.459*** -0.113 -0.119 -0.12 -0.122 -0.122 -0.122 -0.001*** -0.001*** -0.001*** -0.001*** -0.001*** -0.001*** 0 0 0 0 0 0 -0.002 -0.013 -0.014 -0.01 -0.011 -0.011 -0.013 -0.014 -0.014 -0.015 -0.015 -0.015 -1.418*** -1.826*** -1.857*** -1.922*** -1.624*** -1.718*** -1.623*** -0.086 -0.457 -0.465 -0.468 -0.478 -0.482 -0.49 N 19853 16697 16697 16662 15959 15954 15954 F-TESTS: (2) 3.80 (2) 3.54 0.1495 0.1702 (1) 2.09 (1) 1.09 (2) 3.80 (3) 5.06 0.1479 0.2966 0.1495 0.1678 (20) 295.58 (29) 307.88 (40) 334.92 (41) 335.87 (41) 314.78 (42) 316.49 (43) 317.48 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0000*** Legend: * p<0.05, ** p<0.01, *** p<0.001 Wa ld F -test s
tresury securities shocks government securities shocks total loan capital shocks
Ex pl an ato ry V ar . C o ntr o l V ar ia bl es & F ix ed E ff ec ts ** * C o ntr o l V ar ia bl es & F ix ed E ff ec ts ** * C o ntr o l V ar ia bl es & F ix ed E ff ec ts
dependent variable: renegotiation dummy (1 if it happened)
xtprobit
the entire model Constant
bank secutities (tresury&gov) all independent variables profitability
tobins_q leverage (borrower) loans (borrower) zscore
markup (basis points) covenant (1 if included) ln of borrower's ln of borrower's squared tangibility
roa
mean bank time deposit ratio mean bank leverage deal amount normalised bank deal amount normalised firm pricing grid (1 if yes) term loan (1 if yes)
their nulls are not rejected. Industrial loans and credit card loans test significant and negative, in line with the theoretical expectations. However, these specifications are only preliminary results, as they are prone to omitted variable biases resulting from not including all potentially significant
explanatory variables. Thus, they should be seen as tool for the setup of an appropriate set of Wald F-tests in specification 14.
The specification 14 of the table 3 tests for individual and coupled significance of the different type loan loss coefficients when tested together. Thus, it aims to eliminate any omitted variable bias that could be present in and driving the results of the specifications 9 to 13. Negative and significant coefficients on industrial loans lead to acceptance of the alternative hypothesis that their individual effect is, as hypothesized, negative. It is noteworthy that credit card loans’ individual coefficient is no longer significant in this specification. Therefore its significance in the individual specification 11 was likely driven by an omitted variable bias. The Wald F-tests confirm this - the restricted model test on all variables but industrial loan losses tests insignificant, thus the null that the loans to banks, agricultural loans, credit card loans and other loans have no effect cannot be rejected. Also, all Wald f-tests that include industrial loans test significant, that is, due to industrial loans’ significance the null that all tested variables are insignificant is rejected.
In summary, based on these statistics industrial loans are considered to be the only variable that has an effect on the renegotiation decision. Credit card loans seemed to have a potential effect, based on their individual test, but this was shown to be driven by an omitted variable bias.
These variables were the totals of nonaccrual and nonperforming loans; it is therefore possible that their performance was driven by one of the two rather than both jointly. The next table presents tests aimed at their breakdown.
Table 4:
Table 5 presents the tests of relative importance of the two main components of total industrial losses – the nonperforming industrial loans (past due 30 to 89 days) and the nonaccrual (past due more than 89 days). Specification 15 replicates specification 12 in table 4, for reader’s comparison.
(8) (9) (10) (11) (12) (13) (14) -1.247 -2.337 -0.157 -1.992 -3.3 -3.492 -0.81 -0.291 -0.758 -0.983 -5.264* -4.151 -2.401 -2.2 -9.982** -10.896** -3.126 -3.35 1.038 -1.342 -2.438 -2.821
year fixed effect yes yes yes yes yes yes yes
industry fixed effect yes yes yes yes yes yes yes
0.568*** 0.565*** 0.589*** 0.627*** 0.566*** 0.568*** 0.631*** -0.139 -0.139 -0.14 -0.142 -0.14 -0.139 -0.142 10.340*** 10.523*** 10.185*** 10.253*** 10.192*** 10.368*** 10.012*** -2.196 -2.212 -2.202 -2.26 -2.212 -2.197 -2.283 -21500 -12900 -19900 -19100 -28400 -19000 -30700 -23861.46 -25236.97 -23220.28 -25556.32 -25627.24 -23119.22 -29893.38 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0 0 0 0 0 0 0 0.074 0.074 0.074 0.074 0.074 0.074 0.073 -0.051 -0.051 -0.051 -0.051 -0.051 -0.051 -0.051 -0.297*** -0.295*** -0.297*** -0.300*** -0.296*** -0.295*** -0.301*** -0.045 -0.045 -0.045 -0.046 -0.045 -0.045 -0.046 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0 0 0 0 0 0 0 -0.114 -0.113 -0.115 -0.114 -0.123 -0.112 -0.127 -0.065 -0.065 -0.065 -0.065 -0.065 -0.065 -0.066 0.018 0.007 0.013 -0.003 0.021 0.015 -0.006 -0.117 -0.118 -0.117 -0.119 -0.117 -0.117 -0.119 0.001 0.001 0.001 0.002 0.001 0.001 0.003 -0.007 -0.007 -0.007 -0.008 -0.007 -0.007 -0.008 0.02 0.02 0.019 0.024 0.006 0.02 0.01 -0.1 -0.1 -0.1 -0.1 -0.1 -0.1 -0.101 0.244 0.254 0.246 0.246 0.247 0.244 0.248 -0.218 -0.218 -0.218 -0.22 -0.219 -0.218 -0.221 -0.523 -0.532 -0.537 -0.636* -0.512 -0.524 -0.619* -0.307 -0.307 -0.308 -0.311 -0.308 -0.307 -0.313 0.081* 0.080* 0.081* 0.081* 0.087** 0.080* 0.088** -0.032 -0.032 -0.032 -0.033 -0.032 -0.032 -0.033 0.422*** 0.427*** 0.421*** 0.430*** 0.428*** 0.421*** 0.444*** -0.119 -0.12 -0.12 -0.12 -0.12 -0.12 -0.121 -0.001*** -0.001*** -0.001*** -0.001*** -0.001*** -0.001*** -0.001*** 0 0 0 0 0 0 0 -0.013 -0.013 -0.013 -0.009 -0.016 -0.013 -0.012 -0.014 -0.014 -0.014 -0.014 -0.014 -0.014 -0.015 -1.826*** -1.827*** -1.807*** -1.575*** -1.554** -1.873*** -1.197* -0.469 -0.466 -0.467 -0.478 -0.475 -0.466 -0.492 N 16697 16654 16637 16390 16697 16697 16372 F-TESTS: (2) 15.94 0.0003*** (3) 0.61 0.8951 (4) 4.25 0.3738 (1) 0.28 (1) 0.00 (1) 1.14 (1) 4.81 (1) 10.20 (1) 0.18 (5) 16.95 0.5938 0.9621 0.2856 0.0283* 0.0014** 0.6704 0.0046** (41) 335.28 (41) 335.03 (41) 335.79 (41) 339.62 (41) 342.37 (41) 334.92 (45) 348.70 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0000*** Legend: * p<0.05, ** p<0.01, *** p<0.001 Ex p la n ato ry V ar ia b les C o n tr o l V ar ia b les & F ix ed E ff ec ts ** * C o n tr o l V ar ia b les & F ix ed E ff ec ts ** * C o n tr o l V ar ia b les & F ix ed E ff ec ts Wa ld F -test s
other loan total shocks
dependent variable: renegotiation dummy (1 if it happened)
xtprobit
total loan capital shocks loans to banks total shocks agricultural loans total shocks credit card loans total shocks industrial loans total shocks
credit card and industrial loans
all independent variables the entire model banks, agriculture and other loans
banks, agriculture, other and credit card loans markup (basis points) covenant (1 if included) ln of borrower's ln of borrower's squared tangibility
roa
mean bank time deposit ratio mean bank leverage deal amount normalised bank deal amount normalised firm pricing grid (1 if yes) term loan (1 if yes)
Constant profitability tobins_q leverage (borrower) loans (borrower) zscore
Specifications 16 and 17 test individual significance of the two components of total industrial loan losses. Again, coefficient tests are applicable but potentially prone to omitted variable bias.
Nonperforming loans in specification 16 test insignificant and thus the null that they do not have an effect is not rejected. Nonaccrual loans in specification 17 tests significant and negative conflicting the null and leading to acceptance of this paper’s alternative hypothesis of a negative causal effect. Specification 18 puts the results of 16 and 17 into a broader perspective. Both loan loss components retain their significance t-statistics which would suggest that only nonaccrual industrial loans affect the renegotiation decision. However, Wald f-test on the nonaccrual and nonperforming coefficient equality fails to reject the hypothesis that they have the same effect. This is in conflict with the findings of specifications 16 and 17 and in conflict with the individual coefficient tests of the
specification 18. Due to this disparity and keeping in mind the potential multicollinearity that may be inflating coefficient’s standard deviation and thus deflating their t-statistics, the joint explanatory power of nonaccrual and nonperforming loans is not rejected.
Table 5: (15) (16) (17) (18) -9.982** -3.126 -14.194 -13.429 -9.932 -10.123 -9.813** -9.550** -3.361 -3.337
year fixed effect yes yes yes yes
industry fixed effect yes yes yes yes
0.566*** 0.572*** 0.564*** 0.568*** -0.14 -0.14 -0.139 -0.14 10.192*** 10.142*** 10.354*** 10.136*** -2.212 -2.198 -2.213 -2.216 -28400 -20900 -27200 -28500 -25627.24 -23540.52 -25325.71 -25627.23 0.002*** 0.002*** 0.002*** 0.002*** 0 0 0 0 0.074 0.073 0.075 0.074 -0.051 -0.051 -0.051 -0.051 -0.296*** -0.297*** -0.296*** -0.296*** -0.045 -0.045 -0.045 -0.045 0.001*** 0.001*** 0.001*** 0.001*** 0 0 0 0 -0.123 -0.115 -0.121 -0.123 -0.065 -0.065 -0.065 -0.065 0.021 0.018 0.02 0.022 -0.117 -0.117 -0.117 -0.117 0.001 0.001 0.001 0.001 -0.007 -0.007 -0.007 -0.007 0.006 0.018 0.008 0.006 -0.1 -0.1 -0.1 -0.1 0.247 0.232 0.255 0.245 -0.219 -0.218 -0.219 -0.22 -0.512 -0.518 -0.515 -0.511 -0.308 -0.307 -0.308 -0.308 0.087** 0.080* 0.087** 0.086** -0.032 -0.032 -0.032 -0.032 0.428*** 0.425*** 0.425*** 0.428*** -0.12 -0.12 -0.12 -0.12 -0.001*** -0.001*** -0.001*** -0.001*** 0 0 0 0 -0.016 -0.013 -0.016 -0.016 -0.014 -0.014 -0.014 -0.014 -1.554** -1.757*** -1.630*** -1.540** -0.475 -0.47 -0.472 -0.476 N 16697 16697 16697 16697 F-TESTS: (1) 0.13 0.7198 (1) 10.20 (1) 2.04 (1) 8.53 (2) 10.34 0.0014** 0.1530 0.0035** 0.0057** (41) 342.37 (41) 336.89 (41) 340.69 (42) 342.68 0.0000*** 0.0000*** 0.0000*** 0.0000*** Legend: * p<0.05, ** p<0.01, *** p<0.001 C o n tr o l V ar ia b les & F ix ed E ff ec ts ** * C o n tr o l V ar ia b les & F ix ed E ff ec ts ** * C o n tr o l V ar ia b les & F ix ed E ff ec ts Wa ld F -test s
industrial loans total shocks industrial nonperforming loan shocks (past due 30-89 days) industrial nonaccrual loan shocks (past due 90+ days)
xtprobit dependent variable: renegotiation dummy
Ex p la n ato ry V ar . ln of borrower's squared tangibility roa
mean bank time deposit ratio mean bank leverage
deal amount normalised bank deal amount normalised firm pricing grid (1 if yes) term loan (1 if yes)
all independent variables the entire model nonperforming = nonaccrual Constant profitability tobins_q leverage (borrower) loans (borrower) zscore
markup (basis points) covenant (1 if included) ln of borrower's
Table 6 repeats tests of table 5 at the total loan level, to check for the previous tests’ sensitivity to the order in which they are performed. Specification 19 is the same as specification 8 of table 3, and, again, it is used as the base comparison for the reader.
Specifications 20 and 21 test the individual significance of the two components – the nonaccrual and nonperforming total loans. Statistically significant and negative coefficients would be in conflict with the null and would result in an acceptance of this paper’s alternative hypothesis. Neither of them tests significant.
Specification 22 combines them, similarly to the specification 18 of the previous table. The same tests are run as before. Coefficients are insignificant; there is insufficient statistical evidence to conclude that they differ from zero. Wald f-test on joint significance is, also, not in conflict with the null that there is no effect. Finally, Wald f-test on coefficient equality fails to reject null that
nonaccrual and nonperforming loans contribute equally to the total account of loan losses.
Therefore, it is concluded that the results presented in table 3 were not distorted by the use of sum of the two components of total loan losses.
Table 6: (19) (20) (21) (22) -1.247 -2.337 -7.367 0.738 -8.047 -3.478 -0.955 -8.518 -2.899 -9.711
year fixed effect yes yes yes yes
industry fixed effect yes yes yes yes
0.568*** 0.568*** 0.568*** 0.568*** -0.139 -0.139 -0.139 -0.139 10.340*** 10.402*** 10.340*** 10.426*** -2.196 -2.195 -2.197 -2.198 -21500 -23300 -20700 -23100 -23861.46 -24033.2 -23628.2 -23936.79 0.002*** 0.002*** 0.002*** 0.002*** 0 0 0 0 0.074 0.074 0.075 0.074 -0.051 -0.051 -0.051 -0.051 -0.297*** -0.297*** -0.296*** -0.297*** -0.045 -0.045 -0.045 -0.045 0.001*** 0.001*** 0.001*** 0.001*** 0 0 0 0 -0.114 -0.116 -0.113 -0.116 -0.065 -0.065 -0.065 -0.065 0.018 0.017 0.018 0.016 -0.117 -0.117 -0.117 -0.117 0.001 0.001 0.001 0.001 -0.007 -0.007 -0.007 -0.007 0.02 0.021 0.02 0.021 -0.1 -0.1 -0.1 -0.1 0.244 0.243 0.244 0.242 -0.218 -0.218 -0.218 -0.218 -0.523 -0.525 -0.523 -0.526 -0.307 -0.307 -0.307 -0.307 0.081* 0.081* 0.080* 0.081* -0.032 -0.032 -0.032 -0.032 0.422*** 0.425*** 0.422*** 0.425*** -0.119 -0.12 -0.119 -0.12 -0.001*** -0.001*** -0.001*** -0.001*** 0 0 0 0 -0.013 -0.013 -0.013 -0.013 -0.014 -0.014 -0.014 -0.014 -1.826*** -1.765*** -1.845*** -1.760*** -0.469 -0.476 -0.466 -0.476 N 16697 16697 16697 16697 F-TESTS: (1) 0.59 0.4406 (1) 0.28 (2) 0.84 (2) 0.11 (2) 0.88 0.5938 0.3599 0.7417 0.6438 (41) 335.28 (41) 335.71 (41) 335.10 (42) 335.69 0.0000*** 0.0000*** 0.0000*** 0.0000*** Legend: * p<0.05, ** p<0.01, *** p<0.001 Ex p la n ato ry V ar . C o n tr o l V ar ia b les & F ix ed E ff ec ts ** * C o n tr o l V ar ia b les & F ix ed E ff ec ts ** * C o n tr o l V ar ia b les & F ix ed E ff ec ts Wa ld F -test s
total loan capital shocks nonperforming loan capital shocks (past due 30-89 days) nonaccrual loan capital shocks (past due 90+ days)
xtprobit dependent variable: renegotiation dummy
ln of borrower's squared tangibility
roa
mean bank time deposit ratio mean bank leverage
deal amount normalised bank deal amount normalised firm pricing grid (1 if yes) term loan (1 if yes)
all independent variables the entire model Constant nonperforming = nonaccrual profitability tobins_q leverage (borrower) loans (borrower) zscore
markup (basis points) covenant (1 if included) ln of borrower's
This concludes item II. Here, the total industrial loan losses were shown to be determining loan renegotiation, while the other loan types were not. These results hold regardless of whether the sum of nonaccrual and nonperforming or only nonperforming loan losses are used to model loan capital shocks. These results are based on the total sample, that is they hold regardless of borrower type (distressed or not) and time period (pre- post-crisis). The next section tests whether these results hold if one systematically distinguishes between distressed and non-distressed borrowers.
4.3 Item III: Distressed Firm Analysis
The previous items I and II did not account for fundamental differences in syndicate’s approach to distressed and non-distressed corporations, this item makes up for that by splitting the sample into three subsamples. It then re-runs the tests of section II on them. Additional details behind this decision can be found in the statistical testing subsection 3.3.
Table 7 tests the same underlying idea as the table 4; however, it accounts for syndicate’s potentially fundamentally different approach to different distress borrower types. Its specifications 23 to 29 correspond to the specifications 8 to 14 of table 4, these are further broken down into the three borrower categories – the distressed borrowers with z-scores below 1.89, the gray area borrowers with scores between 1.89 and 3 and the non-distressed borrowers with z-scores above 3.
Specification 23, similarly to specification 8 presents the total loan losses as a base comparison. Neither of the coefficients under the three borrower distress states is significant, thus the finding that total loan losses don’t have an effect is not rejected.
Specifications 24 to 28 test effects of individual loan loss types broken down by distress status. First, in the group of distressed borrowers, those with z-scores smaller than 1.89, losses on industrial loans and losses on loans to banks test significantly different from zero. Due to the negative coefficient on the former loan type, the alternative hypothesis that losses on industrial loans negatively affect probability of renegotiation is accepted. The latter, the bank loan losses, test with a positive instead
of a negative coefficient. This means that based on this observation the null hypothesis that losses on loans to banks have no effect cannot be rejected in a one-sided test. Nevertheless, this finding should not be taken lightly. It indicates that this dataset contains correlations that were not fully accounted for in this specification. More specifically, high losses on loans to banks are correlated with high volume of renegotiations. While there may be several theoretical explanations of this, the 2008 crisis is taken to be the most likely cause. This additional theoretical expectation is further addressed in the item IV.
In the group of non-distressed borrowers in specifications 24 to 28 only credit card loan shocks test significant. Their coefficient is negative, thus the alternative is again accepted in place of the null. Specification 29 of the table 7 puts all the losses on different types of loans together to once again do away with any potential omitted variable biases. First, under this holistic specification the negative effect of industrial loan losses on distressed borrower renegotiation is confirmed by a student t-test. Furthermore, this observation is supported by the set of Wald f-tests that are significant only if the losses on industrial loans are included. Thus the evidence against the null is considered to be
particularly strong and the alternative hypothesis is accepted. The statistical significance of losses on loans to banks has vanished in this specification, its individual significance is, therefore, considered to be driven by an omitted variable bias.
Second, the negative effect of credit card loans on non-distressed borrowers is confirmed by a significant negative coefficient. Additionally, another statistically significant positive coefficient emerges in the analysis, this time it is on the losses from agricultural loans. Such observation does not end in rejection of the null hypothesis under the one sided test employed here, however it remains a point for further analysis in section IV. These results were further confirmed by observing significant Wald f-tests of joint significance – all specifications including at least one of the two significant variables tested significant while test on the remaining three variables, bank, industrial and other loan losses, was insignificant.
Table 7: xtP ro b it zsc o re< 1 .8 9 b etw ee n zsc o re> 3 zsc o re< 1 .8 9 b etw ee n zsc o re> 3 zsc o re< 1 .8 9 b etw ee n zsc o re> 3 zsc o re< 1 .8 9 b etw ee n zsc o re> 3 zsc o re< 1 .8 9 b etw ee n zsc o re> 3 zsc o re< 1 .8 9 b etw ee n zsc o re> 3 zsc o re< 1 .8 9 b etw ee n zsc o re> 3 -0 .5 2 1 -2 .3 1 3 -5 .5 8 7 -3 .2 3 7 -5 .1 3 8 -5 .2 5 9 9 .0 3 9 * -3 .4 1 -1 1 .2 5 4 6 .1 2 2 -3 .3 8 1 -1 2 .9 3 2 -4 .4 8 3 -7 .0 9 4 -6 .9 4 8 -4 .9 0 3 -7 .3 2 -7 .2 7 4 -2 .7 1 7 -2 .1 9 7 1 .0 4 7 -2 .1 8 6 -2 .2 8 3 4 .8 1 4 * -1 .4 2 1 -2 .1 6 -1 .1 2 7 -1 .4 9 5 -2 .2 7 5 -2 .2 4 2 -1 .8 5 7 -9 .1 6 3 -1 1 .9 9 3 * -1 .4 4 7 -7 .5 9 8 -1 0 .2 5 7 * -2 .1 9 5 -4 .7 7 3 -4 .8 6 8 -1 .6 7 6 -4 .7 1 3 -4 .7 4 5 -1 2 .4 8 5 ** -1 0 .8 5 1 -7 .0 5 -1 5 .1 9 4 ** -8 .4 3 1 -6 .6 3 4 -4 .3 1 -6 .8 4 5 -7 .0 3 5 -4 .7 8 7 -7 .2 5 7 -7 .4 3 8 2 .1 9 7 -0 .3 9 1 -2 .5 6 3 -1 .6 9 3 3 .4 7 6 -8 .4 1 1 -3 .4 4 7 -5 .3 1 4 -5 .3 3 2 -4 .2 1 5 -6 .2 0 3 -5 .8 1 8 yea r fi xed ef fec t yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes in d u st ry f ix ed ef fec t yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes 0 .5 7 2 ** 0 .8 7 4 ** 0 .4 0 7 0 .5 6 6 ** 0 .8 7 5 ** 0 .4 0 1 0 .6 2 3 ** 0 .8 9 7 ** 0 .3 7 6 0 .6 9 5 ** 0 .9 1 9 ** 0 .4 2 4 0 .5 6 4 ** 0 .8 7 1 ** 0 .4 1 0 .5 7 3 ** 0 .8 7 1 ** 0 .4 0 5 0 .7 0 7 ** * 0 .9 5 2 ** 0 .3 9 4 -0 .2 0 7 -0 .2 9 3 -0 .2 6 1 -0 .2 0 8 -0 .2 9 3 -0 .2 6 2 -0 .2 0 8 -0 .2 9 5 -0 .2 6 4 -0 .2 1 3 -0 .2 9 7 -0 .2 6 4 -0 .2 0 7 -0 .2 9 3 -0 .2 6 3 -0 .2 0 8 -0 .2 9 3 -0 .2 6 2 -0 .2 1 4 -0 .2 9 9 -0 .2 6 7 1 1 .4 4 5 ** * 0 .4 3 6 1 6 .4 3 0 ** * 1 1 .4 3 8 ** * 0 .5 4 1 1 7 .1 6 4 ** * 1 1 .0 4 9 ** * -0 .0 8 1 6 .8 7 7 ** * 1 0 .7 8 4 ** 0 .2 0 4 1 6 .9 2 0 ** * 1 1 .5 8 8 ** * -0 .0 9 8 1 6 .4 3 3 ** * 1 1 .4 9 8 ** * 0 .4 6 3 1 6 .6 1 8 ** * 1 0 .4 5 4 ** -0 .7 7 6 1 7 .8 8 2 ** * -3 .3 3 7 -4 .4 0 5 -4 .2 9 8 -3 .3 4 8 -4 .4 1 6 -4 .3 1 6 -3 .3 3 9 -4 .4 1 6 -4 .3 0 4 -3 .4 5 4 -4 .5 6 8 -4 .3 8 1 -3 .3 6 5 -4 .4 3 6 -4 .3 1 4 -3 .3 4 2 -4 .4 1 5 -4 .2 9 8 -3 .4 9 -4 .6 3 8 -4 .4 3 5 -6 5 1 0 0 -3 0 1 4 .7 4 6 9 4 7 .8 1 7 -6 0 0 0 0 3 3 0 3 7 .8 3 3 5 3 1 7 .1 0 2 -6 6 5 0 0 4 2 8 .8 9 3 5 5 3 8 .0 0 2 -7 5 1 0 0 -8 0 2 1 .5 9 5 7 8 9 6 .8 4 3 -8 0 6 0 0 -3 7 2 9 .4 1 5 1 7 3 7 .0 1 -5 9 5 0 0 -2 6 7 6 .4 0 6 4 1 6 9 .3 9 6 -1 .0 5 e+ 0 5 * 6 5 9 6 .2 9 5 4 2 9 6 .2 6 3 -4 0 9 6 3 .6 8 1 -1 2 2 1 5 .1 6 2 -4 2 3 7 6 .6 0 8 -4 1 2 3 9 .0 7 4 -5 0 7 0 6 .6 9 7 -3 9 7 7 1 .4 6 -4 0 0 8 3 .9 5 3 -2 9 8 2 0 .8 0 3 -4 0 2 5 3 .9 8 4 -4 6 5 4 4 .6 4 4 -3 2 2 8 7 .8 8 9 -3 6 2 6 0 .1 5 2 -4 2 2 1 8 .8 5 1 -1 7 0 0 5 .8 2 5 -4 2 7 8 8 .7 5 -3 9 6 6 1 .8 8 4 -1 1 1 4 7 .6 8 9 -4 0 7 5 0 .1 8 6 -4 9 5 1 8 .9 -5 6 6 9 1 .3 8 8 -3 6 9 7 8 .2 9 5 0 .0 0 2 ** * 0 .0 0 2 ** 0 .0 0 4 ** * 0 .0 0 2 ** * 0 .0 0 2 0 .0 0 4 ** * 0 .0 0 2 ** * 0 .0 0 2 * 0 .0 0 4 ** * 0 .0 0 2 ** * 0 .0 0 2 * 0 .0 0 4 ** * 0 .0 0 2 ** * 0 .0 0 2 ** 0 .0 0 4 ** * 0 .0 0 2 ** * 0 .0 0 2 ** 0 .0 0 4 ** * 0 .0 0 2 ** * 0 .0 0 2 0 .0 0 4 ** * 0 -0 .0 0 1 -0 .0 0 1 0 -0 .0 0 1 -0 .0 0 1 0 -0 .0 0 1 -0 .0 0 1 0 -0 .0 0 1 -0 .0 0 1 0 -0 .0 0 1 -0 .0 0 1 0 -0 .0 0 1 -0 .0 0 1 -0 .0 0 1 -0 .0 0 1 -0 .0 0 1 0 .1 3 6 0 .1 1 9 -0 .0 0 2 0 .1 3 3 0 .1 1 8 -0 .0 0 3 0 .1 3 3 0 .1 1 5 0 0 .1 4 4 0 .1 1 8 -0 .0 0 8 0 .1 3 3 0 .1 2 2 0 0 .1 3 6 0 .1 1 8 -0 .0 0 1 0 .1 3 3 0 .1 1 7 -0 .0 1 4 -0 .0 7 7 -0 .1 -0 .0 9 9 -0 .0 7 7 -0 .1 -0 .0 9 9 -0 .0 7 8 -0 .1 -0 .0 9 9 -0 .0 7 8 -0 .1 0 1 -0 .1 -0 .0 7 7 -0 .1 -0 .0 9 9 -0 .0 7 7 -0 .1 -0 .0 9 9 -0 .0 7 9 -0 .1 0 1 -0 .1 -0 .2 6 2 ** * -0 .2 5 8 ** -0 .3 7 3 ** * -0 .2 6 2 ** * -0 .2 5 7 ** -0 .3 7 2 ** * -0 .2 6 0 ** * -0 .2 6 8 ** -0 .3 7 4 ** * -0 .2 5 9 ** * -0 .2 6 9 ** -0 .3 8 2 ** * -0 .2 6 1 ** * -0 .2 5 7 ** -0 .3 6 9 ** * -0 .2 6 0 ** * -0 .2 5 8 ** -0 .3 7 2 ** * -0 .2 6 5 ** * -0 .2 7 0 ** -0 .3 8 6 ** * -0 .0 6 6 -0 .0 9 -0 .0 9 9 -0 .0 6 6 -0 .0 9 -0 .0 9 9 -0 .0 6 6 -0 .0 9 1 -0 .0 9 9 -0 .0 6 7 -0 .0 9 1 -0 .1 -0 .0 6 6 -0 .0 9 -0 .0 9 9 -0 .0 6 6 -0 .0 9 -0 .0 9 9 -0 .0 6 7 -0 .0 9 2 -0 .1 0 1 0 .0 0 1 ** * 0 0 .0 0 1 0 .0 0 1 ** * 0 0 .0 0 1 0 .0 0 1 ** * 0 0 .0 0 1 0 .0 0 1 ** * 0 0 .0 0 1 0 .0 0 1 ** * 0 0 .0 0 1 0 .0 0 1 ** * 0 0 .0 0 1 0 .0 0 1 ** * 0 0 .0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -0 .1 0 8 -0 .1 4 4 -0 .1 5 8 -0 .1 0 2 -0 .1 5 3 -0 .1 4 8 -0 .1 1 3 -0 .1 5 1 -0 .1 4 7 -0 .1 1 6 -0 .1 3 6 -0 .1 4 1 -0 .1 1 4 -0 .1 5 9 -0 .1 5 9 -0 .1 0 6 -0 .1 4 1 -0 .1 5 3 -0 .1 2 3 -0 .1 6 4 -0 .1 3 9 -0 .1 -0 .1 2 8 -0 .1 2 5 -0 .1 -0 .1 2 9 -0 .1 2 5 -0 .1 -0 .1 2 8 -0 .1 2 5 -0 .1 0 1 -0 .1 2 9 -0 .1 2 6 -0 .1 -0 .1 2 8 -0 .1 2 5 -0 .1 -0 .1 2 7 -0 .1 2 5 -0 .1 0 2 -0 .1 3 1 -0 .1 2 7 -0 .2 0 3 0 .3 0 9 0 .3 7 5 -0 .2 2 0 .3 0 2 0 .3 7 3 -0 .1 9 0 .2 9 9 0 .3 6 5 -0 .2 2 8 0 .2 6 9 0 .3 8 -0 .1 6 6 0 .2 8 8 0 .3 5 6 -0 .2 1 0 .3 1 0 .3 6 7 -0 .1 7 4 0 .2 3 7 0 .4 3 4 -0 .1 8 3 -0 .2 7 3 -0 .2 3 -0 .1 8 4 -0 .2 7 4 -0 .2 3 1 -0 .1 8 4 -0 .2 7 4 -0 .2 2 9 -0 .1 8 6 -0 .2 7 6 -0 .2 3 6 -0 .1 8 4 -0 .2 7 5 -0 .2 2 9 -0 .1 8 3 -0 .2 7 3 -0 .2 3 -0 .1 8 8 -0 .2 7 8 -0 .2 3 9 0 .0 1 7 -0 .0 1 6 -0 .0 2 5 0 .0 1 8 -0 .0 1 6 -0 .0 2 5 0 .0 1 7 -0 .0 1 6 -0 .0 2 5 0 .0 1 9 -0 .0 1 4 -0 .0 2 5 0 .0 1 6 -0 .0 1 4 -0 .0 2 4 0 .0 1 8 -0 .0 1 6 -0 .0 2 5 0 .0 1 7 -0 .0 1 1 -0 .0 2 9 -0 .0 1 1 -0 .0 1 8 -0 .0 1 5 -0 .0 1 1 -0 .0 1 8 -0 .0 1 5 -0 .0 1 1 -0 .0 1 8 -0 .0 1 5 -0 .0 1 2 -0 .0 1 8 -0 .0 1 6 -0 .0 1 1 -0 .0 1 8 -0 .0 1 5 -0 .0 1 1 -0 .0 1 8 -0 .0 1 5 -0 .0 1 2 -0 .0 1 8 -0 .0 1 6 0 .0 0 8 0 .1 3 4 -0 .0 7 0 .0 0 9 0 .1 3 1 -0 .0 7 1 0 .0 0 3 0 .1 3 3 -0 .0 6 4 0 .0 2 0 .1 4 7 -0 .0 6 2 -0 .0 0 7 0 .1 2 6 -0 .0 8 5 0 .0 0 9 0 .1 3 4 -0 .0 6 7 -0 .0 0 2 0 .1 3 7 -0 .0 5 3 -0 .1 5 4 -0 .2 1 6 -0 .2 0 5 -0 .1 5 4 -0 .2 1 6 -0 .2 0 5 -0 .1 5 5 -0 .2 1 6 -0 .2 0 5 -0 .1 5 6 -0 .2 1 7 -0 .2 0 7 -0 .1 5 5 -0 .2 1 6 -0 .2 0 6 -0 .1 5 4 -0 .2 1 6 -0 .2 0 5 -0 .1 5 7 -0 .2 1 7 -0 .2 0 8 0 .6 9 4 * -0 .6 4 3 -0 .3 9 9 0 .7 2 0 * -0 .6 3 8 -0 .3 7 7 0 .7 0 0 * -0 .6 4 1 -0 .3 7 3 0 .7 3 7 * -0 .6 7 4 -0 .4 2 2 0 .7 0 9 * -0 .6 2 6 -0 .4 1 2 0 .6 9 2 * -0 .6 4 -0 .3 8 9 0 .7 6 8 * -0 .6 7 2 -0 .3 9 6 -0 .2 9 4 -0 .4 7 -0 .6 6 4 -0 .2 9 6 -0 .4 7 -0 .6 7 1 -0 .2 9 5 -0 .4 7 1 -0 .6 6 9 -0 .3 -0 .4 7 1 -0 .6 7 2 -0 .2 9 6 -0 .4 7 2 -0 .6 6 6 -0 .2 9 4 -0 .4 7 -0 .6 6 6 -0 .3 0 4 -0 .4 7 3 -0 .6 7 6 -0 .7 7 9 -1 .0 1 0 .4 3 4 -0 .8 0 7 -1 .0 6 0 .4 0 8 -0 .7 8 8 -1 .0 0 4 0 .4 2 -0 .9 2 9 * -1 .0 5 5 0 .4 5 6 -0 .7 2 8 -1 .0 6 1 0 .4 6 5 -0 .7 7 9 -1 .0 0 9 0 .4 3 6 -0 .8 6 4 * -1 .1 2 8 0 .3 7 2 -0 .4 2 -0 .8 0 9 -0 .7 5 6 -0 .4 2 1 -0 .8 1 1 -0 .7 5 9 -0 .4 2 1 -0 .8 1 -0 .7 5 9 -0 .4 2 7 -0 .8 1 6 -0 .7 6 4 -0 .4 2 3 -0 .8 1 -0 .7 5 7 -0 .4 1 9 -0 .8 1 -0 .7 5 7 -0 .4 3 4 -0 .8 1 8 -0 .7 6 7 0 .0 1 4 -0 .0 1 1 0 .1 0 7 * 0 .0 1 1 -0 .0 1 0 .1 0 6 * 0 .0 0 4 -0 .0 0 8 0 .1 0 5 * -0 .0 1 3 0 0 .1 0 3 * 0 .0 2 -0 .0 0 1 0 .1 0 8 * 0 .0 1 3 -0 .0 1 2 0 .1 0 6 * -0 .0 1 0 .0 1 2 0 .1 0 7 * -0 .0 9 1 -0 .1 1 6 -0 .0 5 -0 .0 9 1 -0 .1 1 6 -0 .0 5 -0 .0 9 2 -0 .1 1 6 -0 .0 5 -0 .0 9 3 -0 .1 1 6 -0 .0 5 1 -0 .0 9 2 -0 .1 1 6 -0 .0 5 -0 .0 9 1 -0 .1 1 6 -0 .0 5 -0 .0 9 4 -0 .1 1 6 -0 .0 5 1 0 .3 2 0 .9 4 6 ** 0 .1 7 2 0 .3 1 7 0 .9 4 1 ** 0 .1 7 6 0 .3 0 9 0 .9 4 7 ** 0 .1 7 3 0 .3 1 3 0 .9 2 1 ** 0 .1 8 3 0 .3 0 9 0 .9 6 1 ** 0 .1 6 8 0 .3 2 0 .9 4 0 ** 0 .1 6 9 0 .2 9 4 0 .9 3 7 ** 0 .2 3 -0 .2 1 -0 .3 0 2 -0 .3 0 4 -0 .2 1 1 -0 .3 0 2 -0 .3 0 4 -0 .2 1 1 -0 .3 0 2 -0 .3 0 4 -0 .2 1 4 -0 .3 0 3 -0 .3 0 5 -0 .2 1 1 -0 .3 0 3 -0 .3 0 4 -0 .2 1 -0 .3 0 2 -0 .3 0 4 -0 .2 1 5 -0 .3 0 4 -0 .3 0 9 -0 .0 0 0 ** * -0 .0 0 2 ** * -0 .0 0 2 ** * -0 .0 0 0 ** * -0 .0 0 2 ** * -0 .0 0 2 ** * -0 .0 0 0 ** * -0 .0 0 2 ** * -0 .0 0 2 ** * -0 .0 0 0 ** -0 .0 0 2 ** * -0 .0 0 2 ** * -0 .0 0 0 ** * -0 .0 0 2 ** * -0 .0 0 2 ** * -0 .0 0 0 ** -0 .0 0 2 ** * -0 .0 0 2 ** * -0 .0 0 0 ** * -0 .0 0 2 ** * -0 .0 0 2 ** * 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -0 .0 6 2 0 .1 9 8 -0 .0 6 4 ** -0 .0 6 7 0 .1 9 -0 .0 6 3 ** -0 .0 6 7 0 .1 9 7 -0 .0 6 4 ** -0 .0 7 0 .1 9 6 -0 .0 6 3 ** -0 .0 7 6 0 .1 8 9 -0 .0 6 6 ** -0 .0 6 2 0 .1 9 6 -0 .0 6 4 ** -0 .0 9 5 * 0 .1 8 3 -0 .0 6 0 * -0 .0 4 6 -0 .1 2 5 -0 .0 2 4 -0 .0 4 6 -0 .1 2 6 -0 .0 2 4 -0 .0 4 6 -0 .1 2 6 -0 .0 2 3 -0 .0 4 6 -0 .1 2 6 -0 .0 2 4 -0 .0 4 6 -0 .1 2 5 -0 .0 2 4 -0 .0 4 6 -0 .1 2 5 -0 .0 2 4 -0 .0 4 7 -0 .1 2 7 -0 .0 2 4 -1 .1 2 7 -3 .2 9 6 ** -2 .7 4 0 ** -1 .1 5 1 -3 .3 2 7 ** -2 .8 5 5 ** -1 .0 5 -3 .2 0 9 ** -2 .9 3 3 ** * -1 .0 7 2 -2 .8 0 6 * -2 .4 4 3 ** -0 .8 7 5 -2 .9 1 8 * -2 .6 1 2 ** -1 .1 6 5 -3 .3 6 5 ** -2 .8 3 7 ** -0 .6 7 -2 .3 6 8 * -2 .5 4 0 ** -0 .7 5 7 -1 .1 3 4 -0 .8 9 2 -0 .7 5 8 -1 .1 2 7 -0 .8 8 8 -0 .7 5 7 -1 .1 3 1 -0 .8 8 7 -0 .7 7 3 -1 .1 4 8 -0 .9 0 8 -0 .7 6 -1 .1 6 -0 .9 1 8 -0 .7 5 4 -1 .1 2 6 -0 .8 8 7 -0 .7 9 -1 .1 8 9 -0 .9 6 2 N 6769 4334 5594 6744 4325 5585 6727 4324 5586 6589 4283 5518 6769 4334 5594 6769 4334 5594 6584 4278 5510 F-TE ST S: (5 ) 1 6 .2 2 (5 ) 6 .7 0 (5 ) 1 6 .2 2 0 .0 0 6 2 ** 0 .2 4 3 9 0 .0 0 6 2 ** (4 ) 5 .8 2 (4 ) 3 .8 4 (4 ) 1 4 .8 2 0 .2 1 2 7 0 .4 2 7 6 0 .0 0 5 1 ** (3 ) 4 .9 2 (3 ) 1 .2 9 (3 ) 9 .3 0 0 .1 7 7 5 0 .7 3 1 5 0 .0 2 5 5 * (2 ) 1 1 .0 0 (2 ) 5 .0 1 (2 ) 6 .1 8 0 .0 0 4 1 ** 0 .0 8 1 8 0 .0 4 5 5 * (4 ) 1 5 .1 3 (4 ) 2 .9 4 (4 ) 9 .9 9 0 .0 0 4 4 ** 0 .5 6 8 0 0 .0 4 0 7 * (3 ) 1 1 .5 9 (3 ) 2 .0 4 (3 ) 5 .8 5 0 .0 0 8 9 ** 0 .5 6 5 1 0 .1 1 8 9 (2 ) 2 .8 9 (2 ) 3 .6 7 (2 ) 9 .7 5 0 .2 3 5 5 0 .1 5 9 9 0 .0 0 7 6 ** (4 1 ) 1 8 4 .4 7 (4 1 ) 1 1 6 .1 1 (4 1 ) 1 3 6 .5 7 (4 1 ) 1 8 7 .5 2 (4 1 ) 1 1 6 .7 4 (4 1 ) 1 3 7 .3 9 (4 1 ) 1 8 8 .0 2 (4 1 ) 1 1 6 .1 1 (4 1 ) 1 3 6 .1 5 (4 1 ) 1 8 7 .3 5 (4 1 ) 1 1 8 .1 8 (4 1 ) 1 4 1 .0 9 (4 1 ) 1 9 0 .2 2 (4 1 ) 1 1 7 .5 8 (4 1 ) 1 3 5 .9 9 (4 1 ) 1 8 4 .6 8 (4 1 ) 1 1 5 .9 6 (4 1 ) 1 3 5 .7 7 (4 5 ) 1 9 9 .4 2 (4 5 ) 1 2 0 .5 1 (4 5 ) 1 4 8 .5 4 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * 0 .0 0 0 0 ** * Leg en d : * p <0 .0 5 , ** p <0 .0 1 , ** * p <0 .0 0 1 Exp lan ato ry Var iab
les ts fec Ef ixed & F les ab ari ol V ntr * Co ts ** fec Ef ixed & F les ab ari ol V ntr * Co ts ** fec Ef ixed & F les ab ari ol V ntr Co tests F- Wald
o th er lo an to ta l sh o ck s d e p e n d e n t va ri ab le : r e n e go ti at io n d u m m y (1 if it h ap p e n e d ) to ta l l o an c ap ita l sh o ck s lo an s to b an ks to ta l sh o ck s ag ri cu ltu ra l l o an s to ta l sh o ck s cr ed it ca rd lo an s to ta l sh o ck s in d u st ri al lo an s to ta l sh o ck s in d u st ri al , b an k an d o th er cr ed it ca rd a n d a gr ic u ltu ra l th e en ti re m o d el al l i n d ep en d en t va ri ab les cr ed it ca rd , b an k, a gr ic u ltu re an d o th er b an k, a gr ic u ltu re a n d o th er in d u st ri al a n d c red it ca rd in d u st ri al , b an k, a gr ic u ltu re an d o th er lo an s (b o rr o w er ) m ea n b an k ti m e d ep o si t ra ti o m ea n b an k lev er ag e d ea l a m o u n t n o rm al ised b an k d ea l a m o u n t n o rm al ised f ir m p ri ci n g gr id ( 1 if y es) m ar ku p ( b asi s p o in ts ) co ven an t (1 if in cl u d ed ) ln o f b o rr o w er 's ln o f b o rr o w er 's sq u ar ed ter m lo an ( 1 if y es) ta n gi b ili ty ro a p ro fi ta b ili ty to b in s_q lev er ag e ( b o rr o w er ) (2 6 ) (2 8 ) (2 9 ) C o n st an t (2 3 ) (2 7 ) (2 5 ) (2 4 ) zsc o re
In summary, both table 6 and table 7 conclude that losses on industrial loans negatively affect distressed borrowers. Next, table 7 concludes that the gray area borrowers are not affected by any losses. Also, based on findings of table 7, non-distressed borrowers are being negatively affected by credit card losses and positively affected by agricultural loan losses. Finally, the inconsistent
emergence of positive coefficients in specifications 24 to 29 of table 7 raises suspicion.
Given the positive character of the bank loans and agricultural loans coefficients, a likely disturbing factor was the 2008 financial crisis. If banks changed their behavior after 2007, the assumption on identical distribution would be broken rendering the results biased. Therefore, the thus obtained findings are challenged one last time by checking their consistency between the pre and post 2007 periods.
4.4 Item IV: Final Analysis
Item III ended at a confusing note, industrial loan shocks were shown to have negative effect only at distressed borrowers while credit card loan shocks negatively affected only non-distressed
corporations. Moreover, agricultural loan shocks were associated with a higher loan renegotiation probability. It was suggested that there may be unobserved fundamental changes to the industry resulting from the 2007 crisis. These concerns are addressed in this item.
The specification 30 of table 8 tests for the same as the specification 28 of table 7; however it is limited to the pre 2007 period. Contrary to the findings of item III’s table 7, no loan losses tested significantly positive. Instead, agricultural loan losses in the first column (zscore<1.89) tested significantly negative, thus the alternative hypothesis that losses on agricultural loans negatively affect probability of renegotiation was accepted based on this observation. Furthermore, this
conclusion was supported by two of the Wald f-tests that included agricultural loans, but at the same time it was contested by the last test on the joint significance of credit card loan losses and
Next, in the table 8 under specification 30, loans to banks, credit card loans and other loans tested significantly negative for non-distressed borrowers (zsocre>3). Moreover, this finding was supported by all of the Wald f-tests. Consequently, there is strong statistical evidence in support of the
hypothesis that in the pre-crisis period (2001 to 2007) non-distressed borrowers were significantly and negatively affected by syndicate’s losses on loans to other banks, credit card loans as well as other loans. In summary in the pre-2007 period, loan losses were affecting both distressed as well as non-distressed borrowers, the latter, however, faced significantly stronger per loss impact on the renegotiation probability.
The specification 31 of table 8 conducts the same tests for the period beginning in early 2007. Under this specification only one coefficient is significantly different from zero. That is the positive
coefficient on agricultural loans for non-distressed borrowers. This is not in conflict with the one sided hypotheses test on coefficients and therefore the null hypothesis that there is no effect is not rejected. It is, however, noteworthy. It may indicate further unaccounted for problems with this specification. The most likely, from the author’s perspective, is the chaotic state of markets during the financial meltdown. This, however, does not suffice as a scientific explanation and the positive coefficient remains an unanswered question.
Table 8:
xtProbit zscore<1.89 between zscore>3 zscore<1.89 between zscore>3
3.304 9.535 -22.595* 8.044 -22.966 -16.309 -7.128 -11.03 -9.534 -7.739 -17.415 -15.412 -5.756* -3.9 -0.884 -1.04 -1.683 8.482** -2.869 -4.624 -4.167 -1.86 -3.156 -3.21 -0.816 -6.491 -15.233** -4.22 -8.729 2.729 -1.404 -6.06 -5.545 -7.076 -10.682 -13.208 -12.971 -23.098 -9.307 -10.703 1.666 11.823 -7.57 -13.836 -11.497 -6.582 -11.068 -10.634 -6.685 34.025 -55.876*** 3.181 2.016 5.169 -11.629 -18.61 -16.912 -5.168 -7.886 -7.334
year fixed effect yes yes yes yes yes yes
industry fixed effect yes yes yes yes yes yes
-1.917** -2.281 -1.505* 1.082*** 1.983*** 0.826* -0.617 -1.188 -0.647 -0.254 -0.392 -0.332 4.859 -1.292 20.931*** 15.069** -4.311 14.697* -4.951 -6.715 -6.211 -5.573 -7.77 -7.209 -2.06e+05** -69700 -14300 -19300 -8816.637 2.12e+05* -77297.724 -82958.676 -48923.647 -67584.156 -181000 -99080.778 0.003* 0.004 0.004** 0.002** 0.001 0.002 -0.001 -0.002 -0.002 -0.001 -0.002 -0.001 0.342** 0.022 0.25 -0.127 0.198 -0.398* -0.117 -0.15 -0.139 -0.126 -0.155 -0.183 -0.169 -0.117 -0.131 -0.401*** -0.534*** -0.755*** -0.1 -0.141 -0.137 -0.099 -0.154 -0.186 0.001* 0 0 0.001*** 0 0.001* 0 -0.001 -0.001 0 -0.001 -0.001 -0.252* -0.022 -0.298 -5.449 -5.758 -5.108 -0.127 -0.167 -0.156 -14439.454 -13074.782 -12344.81 -0.056 0.144 0.424 0.082 0.389 0.246 -0.289 -0.426 -0.358 -0.297 -0.441 -0.355 0.004 -0.01 -0.026 0.007 -0.012 -0.015 -0.019 -0.028 -0.024 -0.018 -0.028 -0.023 -0.076 0.121 0.025 0.114 0.189 -0.209 -0.248 -0.335 -0.311 -0.222 -0.347 -0.306 1.088* 0.251 1.219 0.297 -1.387* -0.848 -0.464 -1.128 -1.311 -0.425 -0.613 -0.859 -0.525 -1.208 1.471 -0.717 -0.968 -0.796 -0.892 -1.424 -1.226 -0.552 -1.217 -1.102 -0.131 0.218 0.079 0.085 -0.252 0.158* -0.144 -0.2 -0.078 -0.143 -0.187 -0.071 0.301 0.842 0.711 0.438 0.879* -0.153 -0.35 -0.585 -0.545 -0.299 -0.431 -0.405 0 -0.001 -0.003*** -0.001*** -0.002*** -0.002*** 0 -0.001 -0.001 0 -0.001 -0.001 -0.157 -0.038 -0.066 -0.098 0.298 -0.073* -0.083 -0.197 -0.035 -0.063 -0.193 -0.036 -0.326 -1.669 -1.425 -3.410** -4.496* -3.149* -1.165 -1.848 -1.432 -1.291 -1.84 -1.438 N 2186 1547 2118 4398 2731 3392 F-TESTS: (5) 11.18 (5) 5.30 (5) 27.23 (5) 5.51 (5) 2.87 (5) 10.77 0.0479* 0.3809 0.0001*** 0.3566 0.7202 0.0561 (4) 7.14 (4) 4.32 (4) 23.09 (4) 2.05 (4) 2.81 (4) 9.46 0.1289 0.3646 0.0001*** 0.7258 0.5905 0.0506 (3) 6.77 (3) 3.60 (3) 14.21 (3) 1.66 (3) 2.08 (3) 9.45 0.0797 0.3079 0.0026** 0.6460 0.5568 0.0239* (2) 3.28 (2) 3.98 (2) 8.63 (2) 3.46 (2) 0.69 (2) 1.58 0.1943 0.1369 0.0133* 0.1774 0.7065 0.4550 (4) 10.82 (4) 4.42 (4) 17.23 (4) 4.71 (4) 2.08 (4) 10.72 0.0287* 0.3527 0.0017** 0.3188 0.7209 0.0299* (3) 4.08 (3) 4.18 (3) 17.20 (3) 4.42 (3) 1.84 (3) 2.68 0.2531 0.2424 0.0006*** 0.2192 0.6069 0.4438 (2) 4.38 (2) 1.85 (2) 7.55 (2) 0.69 (2) 0.93 (2) 7.04 0.1119 0.3965 0.0230* 0.7098 0.6288 0.0295 (38) 93.62 (38) 10.77 (38) 90.75 (39) 139.93 (39) 93.99 (39) 88.41 0.0000*** 1.0000 0.0000*** 0.0000*** 0.0000*** 0.0000*** Legend: * p<0.05, ** p<0.01, *** p<0.001 loans to banks total
shocks
agricultural loans total shocks
credit card loans total shocks
industrial loans total shocks
other loan total shocks
dependent variable: renegotiation dummy (1 if it happened)
Ex p la n ato ry V ar ia b les C o n tr o l V ar ia b les & F ix ed E ff ec ts ** * C o n tr o l V ar ia b les & F ix ed E ff ec ts ** * C o n tr o l V ar ia b les & F ix ed E ff ec ts Wa ld F -test s (30) (31) industrial, bank, agriculture and other industrial, bank and other
credit card and agricultural the entire model industrial and credit card
2007 and after
Constant
all independent variables credit card, bank, agriculture and other bank, agriculture and other profitability tobins_q leverage (borrower) loans (borrower) zscore before 2007
markup (basis points) covenant (1 if included) ln of borrower's ln of borrower's squared tangibility roa
term loan (1 if yes) mean bank time deposit ratio mean bank leverage deal amount normalised bank deal amount normalised firm pricing grid (1 if yes)
4.5 Results Summary
Item I showed potential difficulties with the simple probit maximum likelihood estimator and suggested using the Wooldridge’s method. Item II applied this method and tested its result’s
sensitivity to splitting the data by the loan types (bank, agricultural, credit card, industrial and other) and by the loss types (nonperforming and nonaccrual). Item III built on these findings and
investigated how these change for the individual subgroups of borrowers (distressed, gray area and non-distressed). It discovered that there may be reason for expecting a structural change in the data. The most likely cause of this was deemed to be the 2008 crisis. Should that be true, the identity of distribution behind the first assumption (i.i.d. assumption) would be broken. Therefore, item IV aimed at controlling for this potential break by splitting the sample in two subsamples.
Based on these tests it was accepted that in the pre-2007 period loan losses affected both distressed and non-distressed borrowers negatively. This effect was in real terms disproportionate, in particular, the non-distressed borrowers were affected substantially more severely than the distressed
borrowers were. On the other hand, in the post-2007 period, there is insufficient statistical evidence to accept the alternative hypothesis. Finally, it was inferred that there was a structural change in the data which was distorting the results in item III.
5 Conclusions
This paper investigated whether capital shocks to US syndicates affect loan renegotiation decision; and if they do, then to what extent do these shocks affect non-distressed borrowers differently from those that face financial difficulties. It was concluded that in the pre-2007 period capital shocks originating in losses on loans to banks, credit card loans and other loans significantly lower probability of renegotiation for non-distressed borrowers; while losses on agricultural losses negatively affect distressed borrowers. Economically, the effect on non-distressed borrowers was much stronger than the effect on distressed firms. This disparity between distressed and
