In Table III I report results from the main regression specification testing the relationship lending hypothesis. The sample is a panel of firm level data collected at quarterly frequencies. The dependent variable is LBit and is part of the identification strategy in this chapter.
All the coefficients in Table III are estimated using OLS with standard errors clustered at quarterly level. The regression equation also includes firms fixed effects to account for compositional effects. Finally the difference-in-differences approach net out all of the fixed differences between firms that are in the treatment group (those who have a close lending relationship with a bank) and firms in the control group. The key variable of interest in this chapter is the interaction between the lending variable and the relationship lending variable. In column (1) the coefficient of the interaction variable is positive and significant at 5% level. According to the estimated coefficient, when credit standards tighten, firms with close bank relationship should be favoured in their loan application compared to other firms. This is evidence of a flight to quality for banks who experience a shock in loanable funds. Conversely, the coefficient for the lending variable is negative and is significant at 1% level, showing that while the lending standards are relevant in explaining why some firms choose to switch from bank loans to bonds, the presence of relationship lending can attenuate the shift from one form of financing to the other.
In column (2) the coefficient of the interaction variable has a negative sign and it is significant at 5% level. The lending growth variable also presents a positive coefficient. I interpret this result as an indication that in normal times, when lending growth is high, having a close relationship with a bank it is not enhancing the firm’s probability to obtain credit, if something it might push away from bank loans and towards market finance some firms who have the possibility to switch to alternative forms of financing in order to avoid to pay the monopoly rent to the relationship bank. This hypothesis is also consistent with the academic literature on relationship lending (see Rajan (1992) and Bolton et al. (2016)). In column (3) the variable of interest is formed by the interaction of the relationship
lending variable and the non-performing loans variable. The point estimate for non per- forming loans is negative - as expected from theoretical literature and empirical evidences provided in the previous chapter. The coefficient of the interacted variable instead is positive and significant at 1% level. The coefficient estimated for the lending variable in column (3) is negative and significant at 1% level. The sign of the coefficient implies that firms tend to resort less to bank loans and more on bond financing when non performing loans in the banking industry increase. However, we are more interested in the interpretation of the co- efficient of the interaction variable which is positive and strongly significant. The sign of the coefficient indicates that if a firm set up a close relationship with a bank, then the reduction in bank loan financing caused by the high level of non performing loans in the economy is less severe. Firms closely connected to a relationship bank still find relatively convenient to apply for loans relative to the costs of issuing bonds even in period of instability for the banking industry.
Overall the presence of relationship banking can improve the economic terms of the loan contract for the firms. The finding can be interesting also from a regulator point of view. The development of relationship banking links presents a trade-off for the firms. On the one hand it reduces competitiveness in the credit market allowing relationship banks to extract monopoly rents from the firm due to the informational advantage they acquired compared to other banks. On the other hand relationship lenders can mitigate the adverse effect of a negative economic shock by providing stable funding to profitable firms. The relationship lending practice can also reduce the problem of liquidity hold-ups that plagued the Eurozone financial system and that would threaten the functioning of the monetary policy lending channel. According to results in column (3), relationship banks tend to grant loans to already known firms with good investment projects as long as they have available funds, despite the high level of non-performing loans in the banking industry. Results in column (4) follow those in column (3). The loan allowances variable is very correlated to the amount of non-performing loans and therefore it is also correlated to the performance of the
banking industry. For the same reason mentioned above, a firm that usually relies heavily on relationship funds will experience a lesser reduction of bank funds availability compared to firms without bank’s ties.
Finally, in column (5) the coefficient for the monetary policy variable is positive and sig- nificant. I interpret the sign of the coefficient as a by-product of central bank interventions to keep yields low during crises, and the shortage of banks’loanable funds which lead to a consequential increase in the cost of bank’s funds relative to the costs in the bond market.
21 However, as theory predicts, the increase in costs (and decline in loanable funds) should
be experienced to a lesser extent from firms linked to banks through a significant lending relationship. The negative coefficient for the interaction variable in the last column is consis- tent with the above mentioned theoretical prediction. Indeed, the firms who usually obtain loans from a relationship banks, are less likely to switch to bonds even in period in which the liquidity injections of the central banks lower the yields of corporate bonds.
Table III. Relationship Lender
Table III reports the results of the main specification for the five lending variables considered: 1) Lending standards, 2) Lending growth, 3)Non-performing loans, 4) Loan allowances and 5) Monetary policy. The dependent variable isLBitwhich is the ratio
of the amount of loans raised by a firm in a given quarter divided by the total amount raised by the firm in that quarter. LBit
equals 0 if a firm issue loans only, equals 1 if the firm issues bonds and is a number between 0 and 1 if a firm issued both forms of finance in a given quarter. RL is the relationship lending variable. The interaction term is the product of the RL variable with the lending variable used in the specification. Each coefficient reported is the result of an OLS regression which includes firm-fixed effects and controls for firms’ characteristics.The estimation period is 2002:Q1 - 2015:Q2, all errors are clustered quarterly. Dependent Variable: LBit (1) (2) (3) (4) (5) Lending variables: (RL*Lending variable) 0.000805** -0.00477** 0.0191*** 0.0190** -0.0240** (0.000315) (0.00215) (0.00665) (0.00717) (0.00903) Lending standards -0.000985*** (0.000340) Lending growth 0.00251* (0.00146) NPL’s -0.0327*** (0.00627) Loan allowances -0.0400*** (0.00816) Monetary policy 0.0142** (0.00617) RL 0.908*** 0.912*** 0.932*** 0.930*** 0.905*** (0.0116) (0.0104) (0.00954) (0.00968) (0.0113) Firm’s characteristics:
Lag log assets -0.0457*** -0.0375*** -0.0301** -0.0323** -0.0360** (0.0140) (0.0138) (0.0138) (0.0139) (0.0139) Lag log ppe -0.0109 -0.0141 -0.0154 -0.0145 -0.0144
(0.0102) (0.00975) (0.0109) (0.0109) (0.0100)
ROA -0.142 -0.102 -0.134 -0.129 -0.0962
(0.130) (0.138) (0.137) (0.137) (0.138) Market to book assets 6.08e-09* 4.21e-09 4.80e-09 4.81e-09 5.96e-09
(3.43e-09) (3.41e-09) (3.61e-09) (3.52e-09) (4.00e-09) Lagged return 0.00187 0.00937 0.00861 0.00897 0.00953 (0.00527) (0.00643) (0.00653) (0.00653) (0.00668) Lag leverage 0.0804** 0.0747** 0.0791** 0.0784** 0.0809** (0.0346) (0.0366) (0.0381) (0.0381) (0.0361) Dividend -0.0111 -0.00782 -0.00654 -0.00698 -0.00842 (0.00940) (0.00889) (0.00896) (0.00895) (0.00902) Time trend -0.00923*** -0.00761*** -0.00591*** -0.00465** -0.00752*** (0.00197) (0.00206) (0.00206) (0.00211) (0.00201) Constant 1.101*** 1.051*** 0.981*** 0.990*** 1.038*** (0.0748) (0.0781) (0.0655) (0.0663) (0.0769) Firm FE Y Y Y Y Y Calendar FE N N N N N Observations 16,784 16,909 16,150 16,150 16,585 R-squared 0.503 0.503 0.508 0.507 0.501
5.
Robustness
In this section I provide results for alternative specifications of the main empirical analy- sis. The objective of this section is to provide additional support to the evidences presented in section 4.