Formal Financial Institutions and Informal Finance Experimental Evidence from Village India Isabelle Cohen (Centre for Micro Finance) isabelle.cohen@ifmr.ac.in September 3, 2014, Making Impact Evaluation Matter
Acknowledgements This presentation is based on the following in-progress papers: Binzel, Christine, Erica Field and Rohini Pande. Does the Arrival of a Formal Financial Institution Alter Informal Sharing Arrangements? Experimental Evidence from Village India. 2014. Cohen, Isabelle, Erica Field, Elisa Maffioli and Rohini Pande. The Impact of Formal Finance on the Moneylender Market: Evidence from Rural India. 2014.
Table of Contents 1. Introduction 2. Study Design and Context 3. Access to Formal Banking Services 4. Formal Finance and Village Risk-Sharing Capacity 5. Formal Finance and Moneylenders 6. Conclusion
Table of Contents 1. Introduction 2. Study Design and Context 3. Access to Formal Banking Services 4. Formal Finance and Village Risk-Sharing Capacity 5. Formal Finance and Moneylenders 6. Conclusion
Introduction: Motivation The arrival of formal financial institutions is considered a key step in the path of development (Besely, 1995; Burgess and Pande, 2005) From a policy standpoint, there is currently a push towards financial inclusion in rural India Modi promises access to banking for India s poor Financial Chronicle, August 16, 2014 Microlenders to open 30 million bank accounts by Aug 2015 Livemint.com, August 13, 2014 Banks gear up for financial inclusion drive in 2 states Business Standard, August 22, 2014 However, even without formal institutions, households still have the ability to avail informal financial services, which may be a critical component of their economic well-being (or lack thereof)
Informal Financial Arrangements Village Risk-Sharing Capacity Informal reciprocal sharing arrangements between households In the absence of formal financial systems, such informal borrowing and lending arrangements emerge to address risk and enable investment (e.g., Townsend, 1994; Udry, 1994; Fafchamps and Lund, 2003; Kinnan and Townsend, 2012) Moneylenders Individuals who lend as a (regulated) business activity; considered a major source of credit for rural households Share of market in India has decreased over time; in Tamil Nadu, moneylenders constituted approximately 78% of outstanding cash debt in the 1970s, decreasing to 42% in 1991 (Pradhan, 2013) Since then, however, the share has increased, to 53% as of 2002 (Pradhan, 2013) Lack of reliable data on moneylender activities How does the advent of formal finance affect these informal institutions?
Study Overview Exploit randomized rollout of bank branches by a large rural formal financial institution (FFI) in South India Characterized by physical bank branch and broad range of products offered Loans: Grameen-style JLG loans, jewelry loans, and other types of loans Insurance: Personal accident insurance, term life insurance, and other type of insurance Multiple waves of surveying: baseline (pre-opening), midline (1.5 years post-opening) and endline (3 years post-opening) Use household/individual survey data on financial behavior (loans and gifts/transfers) and social network-based borrowing capacity Borrowing capacity is the maximum the respondent could have borrowed from a social network contact in the event of a (hypothetical) emergency
Key Findings To date: data from initial 8 Phase 1 service area pairs Phase II on-going, includes an additional 35 pairs, baseline survey complete, endline survey to start January 2015 Phase III on-going, including an additional 7 pairs, baseline survey in progress Preliminary findings Significant increase in both outstanding formal loan amount and number of loans Decline in outstanding informal loan amount and incidence of having an outstanding informal loan Reduction in borrowing capacity across within-village contacts Decrease in incidence of outstanding loans from moneylenders
Table of Contents 1. Introduction 2. Study Design and Context 3. Access to Formal Banking Services 4. Formal Finance and Village Risk-Sharing Capacity 5. Formal Finance and Moneylenders 6. Conclusion
Pairwise Matched Design Constructed service areas using GPS-based population survey Average service area has a radius of 3-5 km from the branch location and contains approximately 10,000 households Service areas paired using Edmond s algorithm for global distance minimization In each pair, one service area assigned to treatment and the other to control Pair-fixed effects explain roughly 70% of the variation in several 2001 census village outcomes (e.g. caste composition, number of primary schools, etc.) Use data from household surveys to examine the impact on loans, savings, insurance and transfers One respondent per household is asked to list her social network
Empirical Strategy We estimate intent-to-treat effects using the following regression specification: y ip = α + β 1 T ip + β 2 X ip + γ p + ε ip where T is an indicator for being randomly assigned to financial access. Pair fixed effects, γ p, account for stratification. Regressions are done with and without basic household controls Age and years of education of the head of the household, household size, and land ownership Standard errors are clustered at the service area level, using a wild boostrap method with 2000 replications (Cameron et al., 2008) Additionally, we compute randomization inference p-values for comparison
Table of Contents 1. Introduction 2. Study Design and Context 3. Access to Formal Banking Services 4. Formal Finance and Village Risk-Sharing Capacity 5. Formal Finance and Moneylenders 6. Conclusion
Over 25% of Households Take FFI Loans
Increased Access to Formal Finance Table 1: Formal Financial Access Loan Amount Number of Loans Any Loan Panel A: without controls Treatment 14109.5* (7983.9) 0.072 Panel B: with controls Treatment 16138.6** (6616.9) 0.047 Outstanding Repaid Outstanding Repaid Outstanding Repaid (1) (2) (3) (4) (5) (6) 3809.3*** (1318.8) 0.016 4010.6*** (1213.3) 0.019 0.213* (0.115) 0.059 0.248*** (0.0800) 0.029 0.0855** (0.0335) 0.027 0.0938*** (0.0347) 0.020 0.0264 (0.0504) 0.262 0.0360 (0.0364) 0.151 0.0353* (0.0189) 0.057 0.0393* (0.0205) 0.040 Control Mean 35,101.3 3,514.0 1.152 0.196 0.574 0.160 N 651 663 651 663 666 666 ***p < 0.01, **p < 0.05, *p < 0.10 Notes: Estimated using OLS with wild bootstrap errors clustered at the treatment level in parenthesis (nr. of replications 2000) and randomization inference p-values reported below. All specifications include service area pair fixed effects and a constant. Outstanding refers to all outstanding loans (over Rs. 2000) at the time of the survey while repaid refers to loans (over Rs. 2000) the household had repaid during the 12 months prior to the survey. The loan and the average amounts are in Rs. Controls are age and years of education of the household head, household size, and whether the household owns any land.
Decrease in Informal Loans and Transfers Panel A: without controls Treatment -5845.7** (2963.8) 0.037 Panel B: with controls Treatment -5306.2** (2525.1) 0.037 Table 2: Informal Loans and Transfers Loan Amount Any Loan Outside-Village Transfers Lending to Non-Relatives Outstanding Repaid Outstanding Repaid Sent Received Any Items Nr of Items (1) (2) (3) (4) (5) (6) (7) (8) -10876.2 (9110.8) 0.133-10150.5 (7423.7) 0.126-0.0836** (0.0396) 0.042-0.0755** (0.0369) 0.037-0.0137 (0.0413) 0.305-0.0081 (0.0355) 0.381-0.114*** (0.0343) 0.022-0.107*** (0.0345) 0.032-0.0022 (0.0484) 0.457 0.0063 (0.0324) 0.439-0.0222 (0.0392) 0.265-0.0136 (0.0406) 0.352-0.116* (0.0636) 0.046-0.0914 (0.0581) 0.069 Control Mean 31,787.8 16,504 0.708 0.280 0.373 0.330 0.407 0.875 N 651 663 667 664 659 657 669 670 ***p < 0.01, **p < 0.05, *p < 0.10 Notes: Estimated using OLS with wild bootstrap errors clustered at the treatment level in parenthesis (nr. of replications 2000) and randomization inference p-values reported below. All specifications include service area pair fixed effects and a constant. Outstanding refers to all outstanding loans (over Rs. 2000) at the time of the survey while repaid refers to loans (over Rs. 2000) the household had repaid during the 12 months prior to the survey. Loan amounts are in Rs. Transfers refer to informal transfers /gifts (over Rs. 500) sent to and received from households outside the village during the 4 months prior to the survey. Items lent refers to items the household had lent to other households in the village or let people use during the week prior to the survey. Only items lent to non-relatives are considered. Controls are age and years of education of the household head, household size, and whether the household owns any land.
Table of Contents 1. Introduction 2. Study Design and Context 3. Access to Formal Banking Services 4. Formal Finance and Village Risk-Sharing Capacity 5. Formal Finance and Moneylenders 6. Conclusion
Mechanisms of Impact Formal financial access may expand the giving capacity of resourceconstrained households (Feigenberg et al., 2013) Conversely, formal financial access may reduce the real and perceived need for sharing and may reduce possible punishment options (e.g., Ligon et al. 2000; Foster and Rosenzweig, 2000; Conning and Udry, 2007) Decreases incentive to maintain and formal informal sharing arrangements Potential decrease in trust within the network due to an increase in external options for financial support (substitutes) Historical evidence suggests that traditional informal financial institutions diminish in importance as capital markets develop
Decrease in Household Risk-Sharing Capacity Panel A: without controls Borrowing Capacity Treatment -395.7** (176.4) 0.062 Panel B: with controls Treatment -363.9* (188.0) 0.077 Table 3: Strength of Borrowing Network Total Borrow Capacity Spread of Max Borrowing Max-Min Capacity (across Contacts) Variance Coeff Var (1) (2) (3) (4) (5) (6) -1585.0* (921.8) 0.076-1383.2 (861.8) 0.101-1121.8 (698.9) 0.085-1055.0 (644.2) 0.097-1320.3** (669.3) 0.051-1318.9** (668.6) 0.050-24902704.1** (10840948.1) 0.048-25348611.1** (10387266.7) 0.047 0.0286 (0.0572) 0.285 0.0242 (0.0557) 0.313 Control Mean 1,673.4 7,258.3 4,408.4 3,822.4 34,431,544.0 0.688 N 2,853 665 615 615 548 548 ***p < 0.01, **p < 0.05, *p < 0.10 Notes: Estimated using OLS with wild bootstrap errors clustered at the treatment level in parenthesis (nr. of replications 2000) and randomization inference p-values reported below. All specifications include service area pair fixed effects and a constant. Borrowing capacity refers to the maximum amount the respondent expected to be able to borrow from a network contact. In column (1), results are based on an unbalanced respondent-contact panel depending on the number of contacts listed in the network section. Column (2) reports results for a respondent s total borrowing capacity (i.e. the sum over all contacts). The spread is measured as follows: the single largest amount that a respondent can borrow from one of her contacts (column 3), the absolute difference between the maximum and minimum amounts (column 4), the variance (column 5), and the coefficient of variation (sd/mean, column 6). Controls are age and years of education of the household head, household size, and whether the household owns any land.
Table of Contents 1. Introduction 2. Study Design and Context 3. Access to Formal Banking Services 4. Formal Finance and Village Risk-Sharing Capacity 5. Formal Finance and Moneylenders 6. Conclusion
Mechanisms of Impact Formal financial institution (FFI) and moneylenders may be substitutes or complements Channels of impact in the literature: Crowding Out: Entrance of FFI encourage borrowers to shift from moneylenders to FFI due to lower interest rates Crowding In: Inflexible and frequent repayments required for FFI loans may actually increase borrowing from moneylenders (Mallick, 2012) Scale Dis-economies and/or Cream-Skimming: Entry of FFI may result in the loss of an economy of scale for moneylenders; alternatively, low-risk borrowers may gravitate to the cheaper FFIs, leaving higher-risk borrowers to face a (worse) moneylender market Re-lending: Moneylenders may borrow from formal financial institutions and relend to other householders Collusion: Moneylenders may collude with formal financial institutions to set interest rates (Maitra et al, 2013)
Mechanisms of Impact (cont) Complements Substitutes Incidence of borrowing from moneylenders Interest rate charged by moneylenders Crowding-In Increase Increase Collusion Increase Ambiguous Competition Decrease Decrease Crowding-Out Decrease Ambiguous Cream-Skimming Decrease Increase Scale Diseconomies Decrease Increase
Decrease in Borrowing from Moneylenders Panel A: without controls Treatment -0.119*** (0.0360) Panel B: with controls Treatment -0.114*** (0.0367) Table 4: Informal Moneylender Loans Any Loan Number of Loans Loan Size Outstanding Repaid Outstanding Repaid Outstanding Repaid (1) (2) (3) (4) (5) (6) 0.00581 (0.0147) 0.00964 (0.0170) -0.208*** (0.0719) -0.184** (0.0782) -0.0352 (0.0415) -0.0324 (0.0433) -5530.2 (3532.7) -5059.4 (3561.9) -832.4 (1148.9) -787.7 (1117.1) Control Mean 0.453 0.109 0.852 0.188 17372.4 3046.3 N 623 635 613 634 613 634 ***p < 0.01, **p < 0.05, *p < 0.10 Notes: Estimated using OLS with wild bootstrap errors clustered at the treatment level in parenthesis (nr. of replications 2000) and randomization inference p-values reported below. All specifications include service area pair fixed effects and a constant. Outstanding refers to all outstanding loans (over Rs. 2000) at the time of the survey while repaid refers to loans (over Rs. 2000) the household had repaid during the 12 months prior to the survey. The loan and the average amounts are in Rs. Controls are age and years of education of the household head, household size, and whether the household owns any land.
Table of Contents 1. Introduction 2. Study Design and Context 3. Access to Formal Banking Services 4. Formal Finance and Village Risk-Sharing Capacity 5. Formal Finance and Moneylenders 6. Conclusion
Conclusion How does the process of economic development influence household behavior? When other options become available, households shift away from pre-existing informal finance options This is true for both informal sharing arrangements and regulated moneylender activities Moneylender loans in particularly are likely to be more expensive than formal credit; evidence also suggests that friends and relatives also charge interest on many loans In response to availability of formal credit, respondent households may be less willing to invest in maintaining financial ties
Thank You