The Microeconomic Impact of Debt Recovery Tribunals - Pattern of DRT Establishment

Appendix A.1 City of DRT Kolkata Delhi Jaipur Bangalore Ahmedabad Pattern of DRT Establishment Date of est. Apr 27 Jul 5 Aug 30 Nov 30 Dec 21 1994 1994 1994 1994 1994 Chennai Guwahati Nov 4 1996 Jan 7 1997 Patna Jabalpur Mumbai Jan 24 1997 Apr 7 1998 Jul 16 1999 a b Jurisdiction GROUP 1 STATES West Bengal, Andaman & Nicobar Islands Delhi Rajasthan, Himachal Pradesh, Haryana, Punjab, Chandigarh Karnataka, Andhra Pradesh Gujarat, Dadra & Nagar Haveli, Daman & Diu GROUP 2 STATES Tamil Nadu, Kerala, Pondicherrya Assam, Meghalaya, Manipur, Mizoram, Tripura, Arunachal Pradesh, Nagalandb Bihar, Orissa Madhya Pradesh, Uttar Pradesh Maharashtra, Goa The Chennai DRT’s jurisdiction was expanded to include Lakshadweep on Dec 5 1997. The Guwahati DRT’s jurisdiction was expanded to include Sikkim on Dec 5 1997. 34 Figure A.1: Map of India 35 Appendix A.2 Proof of Hypotheses Hypothesis 1 The equilibrium condition (6) is D0 (a2 ) = π 0 (a2 )(R − i2 + φC) ⇒ 1 π 0 (a2 ) = 0 D (a2 ) R − i2 + φC When φ increases, the left hand side of this equation must decrease. Since π 00 (a) < 0 and D00 (a) > 0, this requires that a should increase. Hypothesis 2 The bank’s zero profit condition (7) gives us i= ρ − φC + π(a)φC π(a) Therefore, da da [−(1 − π(a))C + π 0 (a)φC] − π 0 (a) dφ [ρ − (1 − π(a))φC] π(a) dφ di = dφ [π(a)]2 We know that C < 1 ≤ ρ. This gives us di < 0. dφ 36 Appendix A.3 Data Cleaning Database Transfer In September 2000, the bank moved its project loan database from an old database system to a new one. Only loans that were active at the time of migration were transferred to the new system. All loans sanctioned after the date of migration are in the new system. For any active loan the entire repayment schedule is available and hence can be used to reconstruct the history of repayment as described above. My data consist of all loans that existed in the new database at the time of data collection (May 2003), currently active or not. However, the removal of currently inactive loans at the time of database migration causes the following problem due to systematic attrition in the data. The objective of this paper is to examine delinquency, or delays in loan repayment. If a loan is delinquent, the account will remain active for longer since the bank will employ various methods to obtain the payment until the payment is made, or else the loan is written off the books. Therefore at any point in time if we look only at active loans, they are disproportionately likely to be delinquent. Thus the loans transferred to the new system are likely to have disproportionately large number of delinquent loans. However we observe the entire population of loans sanctioned after the database migration, thus these loans have the correct proportion of delinquent loans. This biases the data in favor of finding that delinquency has decreased over time. To remove this problem, I restrict my sample to loans whose last invoice date was scheduled to occur after the date of migration. Barring pre-payment, all of these loans would have to be in the new database regardless of past performance.a Survival The snapshot nature of the data also introduce the issue of survival probabilities. When computing the variable dayslate for invoices that were not repaid by the time of data collection I can only say that repayment is at least x number of days late, but cannot accurately measure the actual number of days late. Since newer loans begin issuing invoices later in the time period, the variable dayslate will tend to show that their payment is less late than for older loans. In all specifications I include the year of sanction as a control. In addition to picking up cohort effects, it controls for the problem that newer sanctions that are delinquent have systematically different dayslate than older delinquent sanctions. a Among the loans sanctioned since the date of migration, very few invoices are pre-paid. This suggests that the absence of prepaid invoices should not bias the results appreciably. 37 .45 .5 .55 .6 .65 Band: +/− 300000 600000 800000 1000000 Amount Overdue after state DRT 1200000 1400000 before state DRT Figure A.2: Probability that invoices are paid within 180 days .12 Change in probability .14 .16 .18 .2 Band: +/− 300000 600000 800000 1000000 Amount Overdue Change in probability 1200000 1400000 Change in probability Figure A.3: Change in probability of repayment after the state DRT was established 38 −.6 −.4 −.2 0 .2 .4 Effect of state DRT on loan size 0 2 4 6 8 10 Decile Figure A.4: Effect of state DRT on loan size Amount overdue 1993 Q 2 Amount overdue 1993 Q 1 1.5e-06 6.7e-08 444469 1.6e+06 Outlate Kernel Density of Amount overdue Figure A.5: Density of loans around Rupees 1 million overdue in 1993:Q1 and 1993:Q2 39 Table 1: Summary statistics about case processing in Bombay High Court versus Mumbai DRTs Number Filed in HC Filed in DRT 18 32 File date Mean Median January 1997 March 1998 August 2002 May 2003 Claim size Mean Median Rs. 820 million Rs. 120 million Rs. 189 million Rs. 126 million Time to summons 393 days (916 days) 136 days (244 days) Time to first hearing 335 days (687) 211 days (295) Time to applicant’s evidence 2628 days (775) 802 days (457) Time to defendant’s evidence 2981 days (794) 696 days (153) Time to arguments starting 2623 days (557) 1095 days (548) 0.67 (0.49) 0.47 (0.51) 27.8 0.0 16.7 38.9 16.7 0.0 18.8 9.4 3.1 28.1 24.4 6.3 44.4 31.3 Interim relief granted (probability) Type of resolution Recovery certificate Consent decree Settled Hearings on Stuck at BIFR In appeal at higher forum (percentages) Finished (percentages) Standard errors in parentheses. Table 2: Descriptive statistics Mean Standard Deviation N Size of loan (millions of Rupees) Interest rate (percent) Year of sanction 24.2 15.31 1989.29 43.7 5.68 2.93 798 798 786 Borrower’s cash flow (millions of Rs.) Borrower’s assets (millions of Rs.) 1859.68 25718.01 591.03 147870.6 15034 15034 State DRT Above State DRT X Above 0.62 0.27 0.14 0.49 0.45 0.35 15034 15034 15034 All paid within 180 days if State DRT = 0 if State DRT = 1 0.65 0.55 0.71 0.48 0.50 0.45 15034 5787 9247 Days late if not paid in 180 days if State DRT = 0 if State DRT = 1 889.17 976.53 804.55 667.05 728.29 589.75 5327 2621 2706 Table 3A: Effect of DRT establishment on repayment behavior of loans sanctioned before June 24th, 1993 –- Baseline Effect (1) All paid within 180 days Bands around Rupees 1 million 150000 100000 85000 (2) (3) (4) Days late if not paid within 180 days Bands around Rupees 1 million 150000 100000 85000 (5) (6) (7) (8) State DRT -0.02 (0.01) -0.40*** (0.04) -0.31*** (0.05) -0.32*** (0.07) 207.01*** (35.65) 542.88* (286.09) 842.14*** (194.80) 842.14*** (194.23) Above -0.15*** (0.01) -1.61** (0.88) -6.52*** (0.49) 0.00 (0.00) 179.17*** (27.94) 261.82 (291.27) -3392.24*** (1633.62) -1020.58 (1189.14) State DRT X Above 0.11*** (0.01) 0.35*** (0.16) 0.25*** (0.10) 0.27* (0.14) -264.89*** (34.30) -324.99 (292.83) -578.62*** (221.83) -578.62*** (221.18) Observations R-squared 15034 0.20 527 0.65 352 0.74 325 0.72 5327 0.26 233 0.70 172 0.76 171 0.75 Controls include year of sanction, age of sanction, borrower’s cash flow, and dummies for borrower’s industry, state where project is located, and quarter when invoice was issued. Columns (2) and (6) also contain a quadratic polynomial, and columns (3), (4), (7) and (8) contain a linear term in the amount overdue. Standard errors are in parentheses. In columns (1) and (5) they are block bootstrapped, and in the remaining columns they are clustered by groups of states that share a DRT. *: significant at 10%; ** significant at 5%; *** significant at 1% Table 3B: Effect of DRT establishment on repayment behavior of loans sanctioned before June 24th, 1993 – Controlling for other unobservables (1) State DRT All paid within 180 days (2) (3) (4) -0.02 (0.01) -0.01 (0.01) -0.03 (0.02) (5) Days late if not paid within 180 days (6) (7) (8) 171.61*** (36.40) 214.89*** (38.29) 200.71*** (36.23) Above -0.14*** (0.02) -0.21*** (0.02) 0.41 (0.32) -0.15*** (0.01) 155.52*** (30.26) 312.79*** (40.55) 2213.73*** (864.72) 181.68*** (27.99) State DRT X Above 0.09*** (0.02) 0.08*** (0.02) 0.06** (0.03) 0.10*** (0.02) -221.09*** (37.81) -173.12*** (42.27) -349.90*** (61.98) -291.13*** (39.06) National DRT X Above 0.08*** (0.02) -232.34*** (82.68) Group 2 state X State DRT 0.01 (0.02) -120.30*** (39.94) Group 2 state X State DRT X Above 0.01 (0.02) 81.97** (44.68) Additional fixed effects state cluster X quarter Observations R-squared 15034 0.21 Above X quarter 15034 0.20 15034 0.20 state cluster X quarter 15034 0.20 5327 0.32 Above X quarter 5327 0.26 5327 0.28 Controls include year of sanction, age of sanction, borrower’s cash flow and dummies for borrower’s industry, state where project is located and quarter. Standard errors are in parentheses are block bootstrapped by clusters of states that share a DRT. * significant at 10%; ** significant at 5%; *** significant at 1% 5327 0.26 Table 4: Fixed effects All paid within 180 days Loan fixed effects Borrower fixed effects (1) (2) (3) (4) (5) (6) Days late if not paid within 180 days Loan fixed effects Borrower fixed effects (7) (8) (9) (10) (11) (12) State DRT 0.01 (0.01) -0.10* (0.05) -0.04** (0.02) 0.02 (0.01) -0.10** (0.05) -0.03* (0.02) 78.29** (35.19) -176.82 (126.95) 80.56* (48.25) 81.46** (32.65) -144.06 (125.86) 89.55* (46.11) State DRT X Above 0.03* (0.02) 0.01 (0.03) 0.05** (0.02) 0.01 (0.01) -0.01 (0.02) 0.04** (0.02) 46.81 (44.30) 115.37* (67.42) -86.05* (49.58) 13.45 (26.47) 30.31 (33.90) -64.14 (44.64) Observations R-squared 15030 0.56 7822 0.58 7208 0.54 15030 0.54 7822 0.56 7208 0.52 5327 0.67 3004 0.69 2323 0.65 5327 0.64 3004 0.67 2323 0.60 In columns (1), (4), (7) and (10), the regression was run over the entire sample. In columns (2), (5), (8) and (11) it was run only for states in group 1, and in the remaining columns only for states in group 2. Controls include year of sanction, age of sanction and dummies for project location in columns (4)-(6) and (10)-(12), and borrower’s cash flow and quarter dummies in all columns. Standard errors are in parentheses and are block bootstrapped by clusters of states that share a DRT. * significant at 10%; ** significant at 5%; *** significant at 1% Table 5: Size of loans sanctioned after DRTs established (1) Post state DRT -0.14 (0.21) Post National DRT Levels (2) -0.13 (0.21) -0.23 (0.24) (4) -0.08 (0.56) 0.64 (0.39) Group 2 X Post state DRT Observations R-squared (3) Borrower fixed effects (5) -0.04 (0.56) 2018 0.43 -0.20 (0.61) 1.99** (0.87) 0.33 (0.23) 2018 0.43 (6) 2018 0.43 0.44 (0.56) 2018 0.70 2018 0.70 2018 0.70 The dependent variable is log(loan size). Controls include year of sanction, firm’s assets, dummies for project type, loan currency and state of project location in all columns, and additionally dummies for borrower’s industry in columns (1)-(3). Standard errors are in parentheses and are bootstrapped by clusters of states that shared a DRT. * significant at 10%; ** significant at 5%; *** significant at 1% Table 6: Interest rates on disbursements occurring after DRT establishment (1) Post state DRT -2.15* (1.19) Post National DRT Act Levels (2) -2.14* (1.18) -2.69** (1.19) (4) Borrower fixed effects (5) (6) -0.94 (1.34) -0.68 (0.72) Group 2 X Post state DRT Observations R-squared (3) -0.94 (1.34) -0.68 (1.28) 1.34** (0.58) 2179 0.34 2179 0.34 -0.75 (1.48) 2179 0.34 -0.44 (1.13) 2179 0.71 2179 0.71 2179 0.71 Other controls include the log of sanction size, type of project, currency, dummies for state of project location and year, dummies for whether the commitment and sanction occurred after DRT establishment, and quadratic trends in timing of commitment and sanction in all columns, and additionally borrower’s assets and dummies for borrower’s industry in columns (1)-(3). Standard errors are in parentheses and are block bootstrapped by clusters of states that share a DRT. * significant at 10%; ** significant at 5%; *** significant at 1% Table A.1: Predicting the pattern of DRT incidence, dependent variable = 1 if state i had a debt recovery tribunal in year t. (1) GDP per capita (2) (3) (4) Cross-sectional OLS (5) (6) (7) Probit (10) Court FE (11) 0.00 (0.00) 0.04 (0.03) -0.00 (0.00) 0.00 (0.01) -10.75** (4.80) 0.08 (0.08) -81.39*** (29.41) -0.01 (0.03) -4.22 (5.85) -0.07 (0.05) 0.09 (0.09) 0.11 (0.10) 0.12 (0.09) 0.11 (0.09) -0.03 (0.07) 0.75 (0.60) 1.39** (0.68) 1.09 (0.68) 0.95 (0.58) -0.54 (0.48) 0.05 (0.09) 0.06 (0.13) 0.06 (0.12) -0.02 (0.13) -0.06 (0.07) 192 0.73 168 0.71 96 0.41a 168 0.79 -0.05 (0.05) GDP growth rate -0.00 (0.00) Credit per capita -0.00 (0.00) Credit growth rate -0.00 (0.00) Cases pending per capita 0.00 (0.00) Judges per capita -9.28** (3.74) State government Congress & allies 0.00 (0.07) 0.09 (0.08) 0.11 (0.08) 0.10 (0.07) Janata & allies Communist party Regional party Center’s ally Observations R-squared (9) (8) 192 0.73 168 0.68 192 0.72 192 0.72 192 0.72 192 0.74 192 0.73 Robust standard errors in parentheses. Year dummies in all columns not reported. Observations correspond to 8 years of data (1993-2000) for 24 states. GDP growth rates are not available for 1993. Union territories are excluded. The court fixed effects regression is run for groups of states which share a high court. * significant at 10%; ** significant at 5%; *** significant at 1%. a: Pseudo R-squared.