Page 223 - ES 2020-21_Volume-1-2 [28-01-21]

206 Economic Survey 2020-21 Volume 1

otherwise similar, any difference in their propensity to restructuring could be attributed to
their varying susceptibility to exploiting forbearance. With this strategy for identification of
the causal effects, the lending activity of banks over the years 2002-2015 is analyzed. The
years are split into three groups: the pre-forbearance period of 2002-2008, the crisis period of
2009-2011, and the post-crisis period of 2011-2015. Organizing the data at a firm-bank-year
level, the following regression is estimated:
Restructured Loan Amount = α+β_1×Treatment ×Crisis + β × Treatment × PostCrisis
j
ijt
t
t
j
2
+ β × X + δ + νj + ϵ ijt
it
it
3
(1) (2)
VARIABLES Restructured Loan Amount (in INR Million)
Treatment X Crisis (2009-2011) 9.070 10.130
(12.952) (13.675)
Treatment X Post-Crisis (2012-2015) 34.582** 35.952**
(16.735) (17.642)
Bank Controls No Yes

Observations 237,690 237,690
R-squared 0.183 0.183
(Firm, Year) Fixed Effects Yes Yes
Bank Fixed Effects Yes Yes

Table 1: The table shows the difference-in-difference estimate described in the equation
above. Restructured loan amount measures the total amount of loan restructured by the bank
j for the borrower i in the year t. Treatment is an indicator variable that takes the value of 1
for banks that have above-median exposure to crisis-hit borrowers and 0 otherwise. Crisis
(Post-crisis) is another indicator variable that takes the value of 1 for years 2009 – 2011
(2012 – 2015) and 0 otherwise, included in column 2, refers to the bank-specific time-varying
controls, namely the bank’s proportion of government ownership, the proportion of foreign
institutional ownership, and the CEO tenure. refers to (firm, year) fixed effects which ensure
that restructuring activity is compared within a (firm, year).stands for bank fixed effects
to absorb any unobserved variations across banks. ** represents a 5% level of statistical
significance. Standard errors, clustered at the bank level, are reported in the parenthesis.
Source: Mannil, Nishesh, and Tantri (2020).
The coefficients denote the average difference in restructuring amount compared to the
pre-forbearance period (2002-08), between the two types of banks. The coefficients are
statistically indistinguishable from zero during the crisis period (2009-11). This suggests
that the restructuring activity by the two types of banks does not differ during the crisis
phase. Note that, rather than comparing total restructuring, the framework compares
restructuring propensity vis-à-vis a given firm. Hence, there are no signs of distortions
during the crisis. The coefficients, however, become positive in the post-crisis period with
a statistical significance of 5%. This indicates that banks more likely to utilize forbearance
became imprudent in the selection of firms for restructuring after the crisis dissipated.
Thus, the prolonged nature of forbearance seems to have distorted banks’ incentives only
after the end of the crisis.