Sectoral credit choice in rural India

The Journal of Choice Modelling 14 (2015) 1–16

Contents lists available at ScienceDirect

The Journal of Choice Modelling journal homepage: www.elsevier.com/locate/jocm

Sectoral credit choice in rural India Debdatta Pal a,n, Arnab K. Laha b a b

Indian Institute of Management Raipur, India Indian Institute of Management Ahmedabad, India

a r t i c l e i n f o

abstract

Article history: Received 6 March 2013 Received in revised form 15 January 2015 Accepted 3 March 2015

This article examines the question, what makes a rural household a preferred choice for formal lenders? A sample selected ordered probit model is developed to address this question. While the selection equation models the determinants of access to credit, an ordered probit model is used to determine the factors affecting the choice of credit sources in hierarchical order. Using household data from six Indian states, this study finds corroborative evidence that relatively resource-rich households, even while staying at distant locations, enjoy greater access to formal creditors. It also identifies a new factor, i.e., interlinked credit, as a significant variable influencing access to formal credit. & 2015 Published by Elsevier Ltd.

JEL classification: C35 G21 Q14 Keywords: Rural credit Sample selected ordered probit India

1. Introduction Recognizing the importance of the agrarian economy in India's overall macroeconomic framework, significant policy initiatives have been undertaken since the British colonial period to reduce the existing imperfections in the rural credit market. Essentially, the government's rural credit policy emphasizes two approaches. One approach seeks to augment credit flow to rural sectors – both farm and non-farm – by expanding outlets of formal financial institutions. These institutions include public or private sector scheduled commercial banks1 (SCB), regional rural banks2 (RRB), and cooperative banks.3 The second approach seeks to provide credit at more favorable terms, through rural credit planning, adoption of regionspecific strategies, rationalization of lending procedures, reducing interest rates, and even providing different interest rates for the poor. Furthermore, the government has tried to rein in the operations of informal lenders, namely moneylenders. These legislations require one to obtain a license to run a money-lending business, impose ceilings on interest rates, and n

Correspondence to: Indian Institute of Management Raipur, GEC Campus, Sejbahar, Raipur 492015, India. E-mail address: [email protected] (D. Pal). 1 Scheduled commercial banks (SCB) are comprised of those banks registered under the second schedule of India's central bank, i.e., the Reserve Bank of India's Act, 1934. They include both public and privately owned banks. Their operation is mostly spread over multiple districts and even across states. 2 Regional rural banks (RRB) came into existence in 1975 with the main goal of catering to rural clients. Operation of RRBs is limited to a few districts. RRBs are jointly owned by the Indian government, the respective state government, and a sponsor bank, which is either an SCB or a state cooperative bank. 3 Cooperative banks are financial institutions where owners are the customers. Primary agricultural cooperative societies (PACS) are localized units owned by local people sharing a common interest. PACS are involved in only deposit mobilization and lending activity. In cases where the cooperatives are also engaged in non-financial activities – such as selling of agricultural input, running consumer stores, trading of agricultural output, etc. – they are known as multi-purpose cooperative societies (MPACS). Others are district central cooperative banks (DCCB) and state cooperative banks. http://dx.doi.org/10.1016/j.jocm.2015.03.001 1755-5345/& 2015 Published by Elsevier Ltd.

2

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

Table 1A Categories of households according to credit sources. Source of loan

Category of household

Formal

Semi-formal

Informal

Yes Yes Yes No Yes No No No

Yes Yes No Yes No Yes No No

Yes No Yes Yes No No Yes No

Loan outstanding with all three credit sources Loan outstanding with formal and semi-formal sources Loan outstanding with formal and informal sources Loan outstanding with semi-formal and informal sources Loan outstanding with a formal source only Loan outstanding with a semi-formal source only Loan outstanding with an informal source only Without any credit facility

ensure transparency in operation. In spite of the conventional wisdom that, along with concessional pricing, broadening the formal credit delivery mechanism is capable of reducing rural households' dependence on informal credit channels (see for example Beck et al., 2007), increasing evidence suggests that the formal channel of credit is yet to make a serious dent in the domain of its informal counterpart. The World Bank's Rural Finance Access Survey (Basu, 2006) indicates that only 21 percent of rural borrower households are indebted to formal financial institutions in India. Several explanations regarding the limited credit from the formal financial sector to rural households have surfaced in the literature. Starting with the contribution by Stiglitz and Weiss (1981), a growing body of literature highlights that fear of adverse selection and moral hazard leads to supply-side quantity rationing. Others attribute the poor coverage of rural borrowers to the formal lenders' lack of personal knowledge about the characteristics and activities of the target group, as well as their inability to monitor the loan. Here, we build our analysis based on a number of stylized observations. First, as compared to lenders from the formal sector, the interest rates charged by informal lenders (e.g., moneylenders, commission agents, traders, and input dealers) are usually high (Sarap, 1990; Ghate, 2007; Pal, 2012). Second, formal credit is offered at a subsidized rate, which is often below the market-clearing price (Ghosh et al., 2001). However, despite these two facts, borrowers may actually bear higher transaction costs for formal lending, in terms of travel expenses for the repetitive visits, opportunity cost of wage loss, and even bribes to negotiate a loan (Mahajan and Ramola, 1996; Guirkinger, 2008). The applicant may therefore be transaction cost-rationed from formal financial sources to account for such added expenditure over that of the nominal interest rate (Guirkinger and Boucher, 2008). Third, in a closely-knit agrarian economy, the informal lenders' close proximity to the borrowers can obviate the need for marketable collateral (Boucher and Guirkinger, 2007), while formal sector creditors primarily resort to collateral-backed finance or ask for a suitable third party guaranty to ensure timely repayment (Hoff and Stiglitz, 1990; Mohieldin and Wright, 2000; Barslund and Tarp, 2008). The inability to arrange for either of the collaterals often keeps poor people out of the banking purview, in spite of feasible and promising investment ideas that could turn into profitable initiatives (Basu, 2006). Thus, the benefit of subsidized formal credit remains restricted within the resource-rich rural households. Fourth, even if credit is available from formal sources, it remains restricted for only production activities, neglecting substantial demand for consumption loans in a poor agrarian economy (Fisher and Sriram, 2002, p.40; Yadav et al., 1992). This unequal access to credit opened the door for the evolution of the semi-formal sector, popularly known as microfinance (Hassan, 2008). This includes self-help groups4 (SHGs), linked with banks and cooperatives, as well as private microfinance institutions5 (MFI). Microfinance lenders are clearly distinct from both the formal banking sector and the informal sector, as they extend loans to poor clients primarily based on a group guarantee in which an individual member stands as guarantor for other group members (Besley and Coate, 1995). The principle of joint liability in the form of peer monitoring and peer pressure ensures timely repayment of the loan, thus waiving the need for marketable collateral when negotiating a loan contract (Stiglitz, 1990; Johnston and Morduch, 2008). For a scientific and empirical analysis of credit delivery in rural economies, one needs to undertake a micro-level study to identify the distinguishing characteristics of rural households. Such an analysis would be useful for understanding the reasons a group of borrowers approaches one type of credit institution instead of another. The analysis can also help in restructuring the rural credit policy for better impact. Given this backdrop, the study presented in this paper attempts to identify the factors determining rural households' choice of formal credit sources over both semi-formal and informal credit sources in India. The results suggest that formal lenders favor an agricultural household, possessing a minimum of two hectares of land and capable of offering marketable collateral, even staying at a distant location. Interestingly, we find that 4 Self-help groups (SHG) are groups of about 20 poor people, belonging to common socioeconomic strata. They initially come together to save and use the corpus for internal lending within the group members. Usually after six months of saving, SHGs may approach banks or cooperatives asking for credit facility. 5 Microfinance institutions (MFI) are financial intermediaries between self-help groups and the banks. They form groups of financially challenged people and extend credit facility to those groups after sourcing loan from donors, investors, or commercial banks.

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

3

Table 1B Categories of households according to credit sources. Source of loan

Category of the households

Ranking

Households met their credit demand exclusively from the formal sector Households met part of their credit demand from the formal sector

4 3

Formal Semi-formal Informal Yes Yes Yes Yes No No No No

No Yes No Yes Yes Yes No No

No No Yes Yes No Yes Yes No

Households excluded from the formal sector, but which availed credit facility from semi-formal and/ 2 or informal sector Households solely dependent on the informal sector 1 Households without any credit facility

Table 2 Indicators of branch penetration and rural indebtedness. State

Chhattisgarh Maharashtra West Bengal Andhra Pradesh Tamil Nadu Gujarat All India

Average population per rural financial institution branch as at end March 2009a

Incidence of indebtedness of rural households (in percent) as on June 2002b Institutional

Noninstitutional

All

10,410 2688 6095 11,784 5976 3759 6554

14.4 22.8 12.1 14.9 13.9 14.7 13.4

6.7 7.2 11.0 32.9 21.3 15.8 15.5

19.8 27.5 21.8 42.3 31.3 28.1 26.5

a Authors' own calculation. Average population per rural financial institution branch was calculated as mid-year rural population (source: Centre for Monitoring Indian Economy) divided by cumulative of commercial bank's rural branches (source: Quarterly Statistics on Deposits and Credit of Scheduled Commercial Banks, Reserve Bank of India) and cooperative societies (source: Trend and Progress of Banking in India, Reserve Bank of India). b Household indebtedness in India as of 30.06.02 (NSSO, 2005).

Table 3 Share of agriculture and allied sector in GSDP, 2009–2010 (In million INR. 1 USD ¼46 INR during the period the reference.) Source: Agricultural Statistics at a Glance 2011 (Government of India, 2011) State

GSDP from the agriculture and allied sector

Percent share of agriculture and allied in GSDP

Chhattisgarh Maharashtra West Bengal Andhra Pradesh Tamil Nadu Gujarat

195,417 918,062 937,679 11,18,755 602,981 675,180

17.79 10.19 23.50 23.54 12.73 15.73

Table 4 Area, yield, and production of food grains in sample states. Source: Agricultural Statistics at a Glance 2011 (Government of India, 2011) State

2009–2010 Area (million Hectares)

Chhattisgarh 4.86 Maharashtra 12.11 West Bengal 6.24 Andhra Pradesh 6.67 Tamil Nadu 3.03 Gujarat 3.69 All India 121.33

Percentage to AllIndia

Production (million tons)

Percentage to allIndia

Yield (Kg/ hectare)

Area under irrigation (%)

4.01 9.98 5.14 5.49 2.50 3.04

4.90 12.59 15.74 15.30 7.51 5.76 218.11

2.25 5.77 7.22 7.01 3.44 2.64

1008 1039 2522 2294 2477 1560 1798

27.6 16.8 48.2 63.9 63.1 44.7 48.3

4

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

Table 5A Summary statistics of continuous variables. Variable

Definition of variable

Mean Minimum Maximum n

Village attributes PUCASTE Percentage of upper caste households in the village AGRI Average distance (in km.) from agricultural infrastructures, such as markets, office of extension service providers, agriculture input retailers, and dealers of farm machinery EDU Average distance (in km.) from educational institutions FININS Average distance (in km.) from formal financial institutions, such as commercial banks, cooperatives Household attributes PCLRUMI Per capita large ruminants FAMTOT Number of key social leaders, such as head of village administration, government extension officers, head master of local school, and officials of financial institutions, whom the household is familiar with

53.24 5.15

5 0.83

96 10.33

14 14

2.23 3.76

0 0

6.5 8.33

14 14

0.28 3.95

0 0

3.5 7

600 600

Note: For the village attributes, zero (0) indicates that the facility is available within the village. Table 5B Summary statistics of categorical variables. Variable

Definition of variable

Frequency

Household attributes LANDD1 Household with minimum agricultural land holding of one hectare (Yes¼ 1, No¼ 0) LANDD2 Household with minimum agricultural land holding of two hectares (Yes¼ 1, No¼ 0) IRRIDUM Household whose agricultural land is covered under assured irrigation (Yes¼ 1, No¼ 0) MTDUM Household mortgaged their asset during last two years (Yes¼ 1, No¼ 0) MIGRATEDUM Any member of the household migrated during 2009-10 to find a job outside (Yes ¼ 1, No¼0) LINK Household which has availed a loan tied with at least one credit complementary service, such as insurance, input selling, output marketing, and extension service (Yes¼ 1, No¼0)

Cumulative

1

0

199 (33.17) 69 (11.50) 188 (31.33) 255 (42.50) 130 (21.67)

401 (66.83) 531 (88.50) 412 (68.67) 345 (57.50) 470 (72.33)

600 600 600 600 600

446 (84.95)

79 (15.05)

525 (100)

(100) (100) (100) (100) (100)

Note: Figures in parentheses are percentages.

Table 6A Village attributes across the categories of households. Village attributes (1)

Non-borrowers (2)

Borrower households (3)

Borrower households

n ¼75

n¼ 525

n¼ 74

Households excluded from the formal sector but with loan from the semi-formal and/or informal sector (5) n¼91

45.51 6.1 3.37 3.21

54.35 5.02 2.07 3.84

45.14 5.93 2.68 3.59

47.23 5.81 2.58 3.68

Households solely dependent on informal sector (4)

PUCASTE AGRI EDU FININS

Households met part Households met credit of their credit demand demand exclusively from formal sector (6) from formal sector (7) n¼ 129

n¼231

49.13 4.97 2.33 3.86

63.03 4.45 1.52 3.98

seeking credit linked with complementary services – such as insurance coverage, the purchase of inputs, selling of output, and extension services – improves borrowers' probability of being serviced by formal lenders. The paper is arranged as follows. Section 2 explores the analytical arguments rationalizing the sectoral choice among formal, semi-formal, and informal credit sources. Section 3 describes the sampling framework and data. We present the empirical model in Section 4 and discuss the results in Section 5. Section 6 concludes and highlights the policy implications.

2. Analytical arguments Before empirically examining the determinants of rural households' choice between formal, semi-formal, and informal credit sources, this section describes the analytical arguments involved in such a decision. At any point in time, a representative household may not have any loan outstanding with any of the lenders. In the case where the household has a loan, it can be from the formal, semi-formal, or informal sectors or from a combination of

Table 6B Household attributes across the categories of households (categorical variables). Non-borrower households (2)

Borrower households (3)

Borrower households Households solely dependent on informal sector (4)

Categorical variables

n¼ 75

n¼ 525

n¼74

Households excluded from formal sector but loan from semi-formal and/or informal sector (5) n¼91

Frequency

Frequency

Frequency

Frequency

1 LANDD1 LANDD2 IRRIDUM MTDUM MIGRATEDUM LINK

0

15 (20.00)

1

50 (80.00) 184 (35.05) 2 (2.67) 73 (97.33) 67 (12.76) 33 (44.00) 42 (56.00) 310 (59.05) 14 (18.67) 61 (81.33) 241 (45.90) 16 (21.33) 59 (78.67) 114 (21.71) 0 (00.00) 75 (100.0) 446 (84.95)

Households met part of their credit demand from formal sector (6) n¼ 129

Households met credit demand exclusively from formal sector (7) n¼ 231

Frequency

Frequency

0

1

0

1

0

1

0

1

0

341 (64.95) 458 (87.24) 215 (40.95) 284 (54.10) 411 (78.29) 79 (15.05)

17 (22.97)

67 (77.03)

23 (25.27)

68 (74.73)

53 (41.09)

76 (58.91)

91 (39.39)

140 (60.61)

4 (5.41)

70 (94.59)

8 (8.79)

83 (91.21)

14 (10.85)

115 (89.15)

41 (17.75)

190 (82.25)

30 (40.54)

44 (59.46)

32 (35.16)

59 (64.84)

82 (63.57)

41 (36.43)

166 (71.86)

65 (28.14)

25 (33.78)

49 (66.22)

30 (32.97)

61 (67.03)

62 (48.06)

67 (51.94)

124 (53.68)

107 (46.32)

20 (27.03)

54 (72.97)

12 (13.19)

79 (86.81)

31 (24.03)

98 (75.97)

5 1(22.08)

180 (77.92)

30 (40.54)

44 (59.46)

72 (79.12)

19 (20.88)

125 (96.90)

4 (3.10)

219 (94.81)

12 (5.19)

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

Household attributes (1)

Note: Figures in parentheses are percentages.

5

6

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

Table 6C Household attributes across the categories of households (continuous variables). Household attributes (1)

Non-borrower households (2)

Borrower households (3)

Borrower households Households solely dependent on informal sector (4)

PLCRUMI FAMTOT

n¼ 75

n¼ 525

n ¼74

Households excluded from formal sector but with loan from semi-formal and/or informal sector (5) n¼ 91

0.13 2.86

0.30 4.11

0.14 3.36

0.16 3.80

Households met part of their credit demand from formal sector (6)

Households met credit demand exclusively from formal sector (7)

n¼ 129

n¼231

0.30 4.18

0.41 4.43

sources. Accordingly, households can be placed in one of the following categories (Table 1A), based on loans outstanding with the respective credit source(s): Working with a dataset of rural Indian households, Pal (2002) used a different approach and categorized surveyed households into four categories: a non-loanee, an exclusively informal sector borrower, an exclusively formal sector borrower, and a borrower from both formal and informal sectors. She employed a multinomial logit model to compare the probability of being in each borrower group, using a base outcome of the non-loanee group. Similar to Pal (2002), Doan et al. (2010) and Castellani (2014) also categorized the surveyed households from peri-urban areas in Vietnam and rural areas of Southern Ethiopia, respectively, in four broad categories: non-borrowing, borrowing from only formal lenders, borrowing from only informal lenders, and borrowing from both formal and informal credit sources. They too employed a multinomial logit model to compare the probability of being in each borrower group with a base outcome of the non-loanee group. Kumar et al. (2007) studied the sectoral choice of the credit market in India through a multinomial logit model by categorizing households among non-borrowers, borrowers of the formal financial sector, and borrowers of the informal sector. To capture the extent of participation in the rural credit market in Uganda, Mpuga (2010) expanded the response variable of the multinomial logit into seven unordered categories: individuals without loan facility, followed by individuals with loans from a bank, a cooperative/non-government organization, the government, traditional moneylenders, friends and relatives, and local community/group. When examining the empirical investigations that were undertaken to identify the distinguishing characteristics of rural households' sectoral choice of credit, we find the following limitations. First, the sample set of the earlier studies include only those borrowing from formal and informal sources. However, since the early nineties, developing countries – specifically, India, since 1992 – have experienced the evolution of a semi-formal sector, i.e., microfinance lending. This includes loans from both private microfinance institutions as well as the government-backed microfinance programs. Second, based on the assumption that credit sources are from unordered categories, multinomial logit models have been commonly used to assess credit access from various sources. However, the literature of the rural credit market postulates that access to formal credit sources is limited to the resource-rich households – who can offer marketable collateral – to overcome the market failure that arises out of the asymmetric information between the formal lender and the prospective borrower (Hoff and Stiglitz, 1990; Basu, 2006). Hence, the households with access to formal credit are expected to have a stronger signaling capacity, in terms of profit potential and asset holding, than the borrowers from the semi-formal and/or informal sectors. As the categories of the response variable can be ranked, these unordered limited dependent models become inappropriate, since they would not be able to account for the ordinal characteristics of the response variable (Liao, 1994). Furthermore, data for a majority of the studies conducted in an Indian context were collected nearly a decade ago. This article accounts for most of these issues. Unlike earlier studies, for which data was restricted to a single state, this study uses a dataset of 600 households covering six Indian states (see Section 3 for details). The sample is comprised of 75 non-borrowers and 525 borrowers falling under all seven possible categories that cover the formal, semi-formal, and informal sectors, as mentioned in Table 1A. Next, as those 525 households that have taken out loans may have different characteristics from the 75 households who did not have any loans, the resulting estimates may be biased if the analysis to assess the factors influencing access to credit is carried out only on borrower households. Hence, unlike the multinomial models employed in the earlier contributions, we use a sample selected ordered probit model (Greene and Hensher, 2010, p.307–308). This approach allows us to, first, correct the selection bias (following Heckman, 1979) and, second, to capture the ordered nature of the response variable. To the best of our knowledge, this methodology has not yet been employed to study the modeling of choice in the context of rural credit markets in developing countries. To ensure that the categories are finite, mutually exclusive, and exhaustive (Train, 1986) we rearrange them as either non-borrowers or borrowers within four broad categories (Table 1B). The justification for such an ordering of the response variable is that if a rational household opts for credit facility, it would prefer formal creditors since the nominal interest rate is lower than that offered by either the semi-formal or informal lenders. The nominal interest rate offered by Indian banks is reported to be around 11 percent per annum – less than half of the interest rate charged by the semi-formal lenders, which ranges between 24 and 36 percent (Basu, 2006). The

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

7

Table 7 Sample selected ordered probit results for access to credit. Parameter Explanatory variables

Parameter estimates Estimates Std. Error Selection equation Dependent variable: LOAND

α1

Intercept

0.326

0.375

α2 α3 α4 α5

Household attributes Household with minimum agricultural land holding of one hectare (Yes ¼1, No¼0) (LANDD1) Household whose agricultural land is covered under assured irrigation (Yes ¼1, No¼0) (IRRIDUM) Per capita large ruminants (PCLRUMI) Household mortgaged their asset during last two years (Yes ¼1, No¼ 0) (MTDUM)

0.301n 0.391nn 0.715nnn 0.519nnn

0.173 0.166 0.274 0.169

α6 α7

Village attributes Average distance (in km.) from agricultural infrastructures (AGRI) Average distance (in km.) from formal financial institutions (FININS)

 0.126nnn 0.220nnn

0.034 0.052

α8 α9 α10 α11 α12

State dummies Chhattisgarh (CGARH) Maharashtra (MAHA) West Bengal (WB) Andhra Pradesh (AP) Tamil Nadu (TN)

-0.606n 1.247nnn 1.150nnn  0.240  0.584

0.337 0.484 0.443 0.452 0.523

Ordered Probit Outcome Dependent variable: LOANCAT

β1

Intercept

12.214nnn

0.315

β2

Household attributes Household with minimum agricultural land holding of two hectares (Yes ¼1, No¼0) (LANDD2)

0.403nn

0.177

β3

Household mortgaged their asset during last two years (Yes ¼1, No¼ 0) (MTDUM)

0.292nn

0.120

β4

Number of key social leaders whom the household is familiar with (FAMTOT)

0.086nn

0.043

β5

Any member of the household migrated during 2009-10 to find a job outside (Yes¼ 1, No¼0) (MIGRATEDUM)

 0.369nn

0.154

β6

Household which has availed a loan tied with at least one credit complementary service (Yes ¼1, No¼0) (LINK) 1.869nnn

0.169

β7

Village attributes Average distance (in km.) from agricultural infrastructures (AGRI)

 0.132nnn

0.033

β8

Average distance (in km.) from formal financial institutions (FININS)

0.157nnn

0.053

β9

Average distance (in km.) from educational institutions (EDU)

 0.084n

0.050

β10

Percentage of upper caste households in the village (PUCASTE)

0.012nnn

0.003

β11

State dummies Chhattisgarh (CGARH)

 12.725nnn

0.123

β12

Maharashtra (MAHA)

 13.296nnn

0.228

β13

West Bengal (WB)

 13.082nnn

0.166

β14

Andhra Pradesh (AP)

 14.040nnn

0.190

β15

Tamil Nadu (TN)

 12.680nnn

0.311

μ1

Cut-point 1

0.902nnn

0.084

μ2

Cut-point 2

1.936nnn

0.108

ρ

Rho

 0.042

0.256

Log likelihood  676.701 Note:

nnn nn

,

and

n

indicate statistical significance at 1, 5 and 10 percent respectively. Gujarat state is considered as base.

interest rate charged by the informal lenders is even higher, sometimes as high as 90 percent (Ghate, 2007). This high interest rate is an observed default premium (Bottomley, 1975), which acts as a cushion to cover any possible default. Nevertheless, formal lenders operating under government control are not allowed to increase the interest rate beyond a certain point. Hence, formal lenders become more specific in choosing clients who can either offer marketable collateral or arrange a guarantee from any financially sound third party to ensure timely repayment. A semi-formal lender enjoys greater flexibility in fixing interest rates, even within a cap; thus, the semi-formal lender charges an interest rate that is higher than that offered by the formal lender. Hence, borrowers from the poor economic strata with limited production assets and without any marketable collateral cannot access formal lenders and depend solely on informal credit sources. In the next category, borrowers with comparatively higher signaling capacity – in terms of profit potential and asset holding – may gain access to the semi-formal sector, though they still remain outside the purview of the formal credit channels. Borrowers with

8

Independent variables

Marginal effect (solely dependent on informal sector)

Marginal effect (dependent on semiformal and/or informal sector)

Marginal effect (credit demand partly met from formal sector)

Marginal effect (credit demand solely met from formal sector)

Household with minimum agricultural land holding of two hectares (Yes¼ 1, No ¼ 0) (LANDD2) Household mortgaged their asset during last two years (Yes¼ 1, No ¼ 0) (MTDUM) Number of key social leaders whom the household is familiar with (FAMTOT) Any member of the household migrated during 2009-10 to find a job outside (Yes¼1, No ¼0) (MIGRATEDUM) Household which has availed a loan tied with at least one credit complementary service (Yes ¼1, No ¼ 0) (LINK) Average distance (in km.) from agricultural infrastructures (AGRI) Average distance (in km.) from formal financial institutions (FININS) Average distance (in km.) from educational institutions (EDU) Percentage of upper caste households in the village (PUCASTE)

 0.0664

 0.0263

 0.0017

0.0944

 0.0487

 0.0193

 0.0012

0.0692

 0.0143

 0.0056

 0.0004

0.0203

0.0608

0.0240

0.0015

 0.0863

 0.3062

 0.1210

 0.0078

0.4350

0.0218

0.0086

0.0006

 0.0309

 0.0260

 0.0103

 0.0007

0.0370

0.0135

0.0053

0.0003

 0.0191

 0.0020

 0.0008

 0.0001

0.0029

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

Table 8 Marginal effect of the ordered probit estimate of loanee households. Dependent variable: LOANCAT.

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

9

access to formal financial sources but still partially dependent on semi-formal and/or informal sources are endowed with higher asset holdings and improved production potential. At the top of the hierarchy are the resource-rich households who can manage their entire demand of credit from formal creditors. Finally, the empirical results presented here are relevant to policymakers, as the study uses recent data covering the credit experience of 600 rural households between April 2009 and March 2010.

3. Survey and data description The data used for our study is a survey dataset from the research project “Assessing Policy Interventions in Agri-Business and Allied Sector Credit versus Credit Plus Approach for Livelihood Promotion,” conducted by the Centre for Management in Agriculture at the Indian Institute of Management Ahmedabad and sponsored by the Government of India's Ministry of Agriculture. It was conducted between May and December 2010 across six Indian states – specifically, Chhattisgarh, Maharashtra, West Bengal, Andhra Pradesh, Tamil Nadu, and Gujarat. Two state-level indicator criteria were used in the selection of the states: the average population per bank branch at the end of June 2009 and the level of indebtedness of rural households (Table 2). Since, 1977 banks in India were directed to open four branches in unbanked area to obtain license of opening one branch in banked area. The assumption was less population per bank branch would improve access to the formal financial system. Second indicator, incidence of indebtedness, captures the extent to which a rural household could meet its credit demand from the institutional sources. Both the indicators were considered during state selection to ensure variation in access to as well as dependence on the formal financial sector. Against the national average of 6554 people served by each bank branch, states were chosen with an even spread on this criterion. In Chhattisgarh, the average population per bank branch is 10,410; the lowest state average is 2688, in Maharashtra. West Bengal is the closest to the national average. Hence, a variation on the penetration of the bank branches is well maintained in our sample. Similarly, sample states are evenly distributed at the average, over the average as well as below the average on the criterion of indebtedness of rural households. This ensures variation among the sample states. Tables 3 and 4 provide a brief description of the sample states to highlight the sample states' representativeness of the Indian agrarian economy. Table 3 highlights the 2009–2010 shares of the agriculture and allied sectors in the Gross State Domestic Product (GSDP) at current prices (2004–2005 as base year). Among the sample states, the highest percentage share of agriculture and allied activity in GSDP is in Andhra Pradesh, which is around 23.54 percent. The share ranges between 10.19 percent in Maharashtra and 23.54 percent in Andhra Pradesh. Table 4 covers the area, yield, and production of food grains, as well as the extent of irrigation coverage in the sample states. Among the sample states, West Bengal performs highest in both production of food grains and yield, followed by Andhra Pradesh. The Andhra Pradesh state, with 63 percent of the cultivable area under irrigation, tops the list for irrigation coverage, followed by Tamil Nadu. Hence, the sample states are evenly distributed at the average, over the average as well as below the average on both the criteria, as seen against the All India average food grain yield of 1798 kg per hectare and All India average irrigation level of 48.3 percent. After the selection of the states, state-wise within any agro-climatic region, a village or cluster of villages was selected where all credit sources – formal, semi-formal, and informal – were available (Appendix 1). A complete enumeration was undertaken of all households6 in a village to form a sampling frame, from which a representative sample of village households was drawn. The survey was conducted in five villages in West Bengal, three villages each in Maharashtra and Chhattisgarh, and one village each in Andhra Pradesh, Gujarat, and Tamil Nadu. In West Bengal, 30 households were randomly chosen from each village; in the other states, 50 households from each village were chosen through random sampling. Thus, the study covers 600 sample households. A structured questionnaire was administered to each sample household, and information was collected regarding the household's demographics, social characteristics, primary and secondary occupations, resource endowments, and differential access to credit sources between April 2009 and March 2010, along with the terms and conditions of those credit facilities. The survey finds households borrowing exclusively from formal, semi-formal, and informal sources, along with households obtaining credit facility from multiple sources. Out of the 600 sample households, 525 households are borrowers and 75 households are without any credit facility. Among the 525 loanee households, 231 households have met their credit demand entirely from formal financial sources. 129 households have met part of their credit demand from formal financial sources, hence meeting the rest of the credit demand from semi-formal and/or informal sources. 91 respondents are excluded from formal credit sources but have borrowed from semi-formal and/or informal sources. Finally, 71 households are found to borrow exclusively from informal sources (Table 1B). For a better understanding of the characteristics of the study's survey villages and households, we report summary statistics in Tables 5A and 5B. Table 5A shows that, on an average, 53 percent of the households belong to the Hindu upper 6 If a village is too large in terms of the number of households, some representative hamlets of that village (not exceeding 300 households) are used for complete enumeration. In that case, only those households covered under complete enumeration constitute the population from which the sample is drawn.

10

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

caste. The average distance from agricultural infrastructures, educational institutions, and financial institutions is found to be 5.15 km, 2.23 km, and 3.76 km, respectively. On average, a four-member household owns one large ruminant like cattle. Finally, on average, a household is familiar with at least four key social leaders. Table 5B shows that, among the 600 households, 199 respondents have at least one hectare of land, while 69 households possess at least two hectares. 31.33 percent enjoy assured irrigation, and about 43 percent of households have mortgaged land during the last two years. 22 percent of households have at least one person who had migrated during 2009–2010. In addition, nearly 85 percent of households have obtained at least one credit complementary service, such as insurance, input selling, output marketing, and extension service with a loan facility. Next, we show the variation in village and household attributes, first between non-borrowers and borrowers (Table 6A) and then among the four borrower categories (Tables 6B and 6C). Referring to Table 6A, we first compare the village attributes between non-borrowers and borrowers (columns 2 and 3). Borrower households are found to be in villages with higher representations of the upper caste Hindu community. Furthermore, among the borrowers, households exclusively borrowing from the formal sector belong to villages where nearly two-thirds of the households are from the upper caste Hindu community. On average, borrower households are closer to agricultural infrastructure, as well as educational institutions, compared to non-borrower households. Moreover, borrower households who can meet their entire credit demand from formal sources exhibit this tendency as well, as compared to borrowers dependent on semi-formal and informal credit sources. Surprisingly, we find that borrowers are located farther from the financial institutions, as compared to nonborrowers. Referring to Table 6B, we find that, among the 75 non-borrower households, only 15 households hold at least one hectare of land and only two households possess an operational landholding of two hectares. Among the 525 borrower households, 184 households hold at least one hectare of land, of which 91 households can meet their demand exclusively from the formal sector. In our sample, only 67 borrower households have an operational land holding of at least two hectares, and among them 55 households have access to formal financial sources. Among households who had obtained any type of credit facility, 59 percent were found to be bestowed with irrigation facility. In our sample, 72 percent of the households exclusively dependent on the formal sector and 64 percent of the households partly dependent on formal lenders have irrigation facility in their farmland. Close to half of the borrower households have mortgaged their assets to obtain credit facility. 54 percent of households with exclusive access to the formal sector have offered marketable collateral to avail a loan, against 34 percent of households solely dependent on informal credit sources. The data shows that as high as 85 percent of the borrower households are engaged in interlinked transactions, under which they have availed various credit complementary services along with loans. Referring to Table 6C, we observe that, among borrower households with exclusive access to the formal credit source, on average, a five-member household owns two large ruminants. Among the households without access to formal lenders, on average, a seven-member household owns only one large ruminant. Similarly, exclusively formal sector borrowers are found to be familiar with more key leaders compared to informal and/or semi-formal borrowers as well as non-borrowers.

4. Econometric model The sample selected ordered probit model (Greene and Hensher, 2010, p.307–308) adopted in this study uses two equations – a selection equation for credit choice and an ordered probit model – to examine the determinants behind whether a household can meet its credit demand entirely from the formal sector, partly from the formal sector, only from the semi-formal and informal sector, or entirely from the informal sector.7 The model system is as follows:

z ⁎ = t′α + μ , z = 1 if z ⁎ > 0 and z = 0 if z ⁎ ≤ 0 y⁎ = x′β + ε,

y = 1 if y⁎ ≤ 0, y = 2 if 0 < y⁎ ≤ τ1, y = 3 if τ1 < y⁎ ≤ τ2, y = 4 if y⁎ > τ2,

7 One of the referees raised the issue of possible endogeniety in the model application; for instance, it could be that if a household want to access some formal channel of credit, they decide previously to buy one more ruminant introducing endogeniety. However, we did not see any instance of such happening during the conduct of the survey and in other informal meetings with the rural households. Hence, we have ruled out of this kind of endogeniety from the model.

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

11

where z is an observed binary variable indicating whether or not a rural household has chosen to borrow (z ¼1 if the household borrows, 0 otherwise), and z ⁎is an underlying continuous variable related to z , ranging from  ∞ to ∞. y is an observed ordinal variable representing the categories of households according to their credit sources, and y⁎ is a latent continuous variable, ranging from  ∞to ∞; the value taken by y depends on y⁎ as described in the system above. t and x are vectors of explanatory variables, α and β are vectors of parameters to be estimated, and μ and ε are random errors distributed as bivariate normal with zero means, unit variances, and correlation coefficient ρ. This distribution is denoted by Φ2 (0, 0; ρ). Finally, τ1 and τ2 are unknown parameters. As we have four ordered categories with one category having a threshold 0, the model has only two unknown threshold parameters. The probabilities of y taking the values 1, 2, 3 or 4 are given by the following,

Prob(y = 1x) = Φ (−x′β)

Prob(y = 2x) = Φ (τ1 − x′β) − Φ (−x′β), Prob(y = 3x) = Φ (τ2 − x′β) − Φ (τ1 − x′β),

Prob(y = 4x) = 1 − Φ (τ2 − x′β). where Φ is the cumulative distribution function of the standard normal distribution. Note that for all probabilities to be positive, we must have 0 < τ1 < τ2. The marginal effect of the changes in the independent variables can be calculated by taking a partial derivative of the above stated expressions,

∂Prob(y = 1x) = − ϕ (x′β) β , ∂x

∂Prob(y = 2x) = [ϕ (−x′β) − ϕ (τ1 − x′β)] β , ∂x ∂Prob(y = 3x) = [ϕ (τ1 − x′β) − ϕ (τ2 − x′β)] β , ∂x ∂Prob(y = 4x) = ϕ (τ2 − x′β) β . ∂x where ϕ is the probability density function of the standard normal distribution. We estimate the parameters using the maximum likelihood method, where the log-likelihood function for a random sample (z1, y1), … , (zn, yn )is written as n

L (θ ) =

⎧ ⎪

4

∑ ⎨ (1 − zi ) ln (π0i (θ)) + ∑ zi I (yi ⎪

i=1

⎩

h=1

⎫ ⎪ = h) ln (π hi (θ)) ⎬ ⎪ ⎭

and where θ = (α, β, τ , ρ), π0 (θ) = 1 − Φ (t ′α), and

π h (θ) = (t′α, τh − 1 − x′β ; − ρ) − (t′α, τh − 2 − x′β ; − ρ) We set τ−1 = − ∞ , τ0 = 0, and τ3 = + ∞ (De Luca and Perotti, 2011). While assessing formal credit market access, the underlying latent variable is the household's capability to signal its creditworthiness. The cut points in the ordered probit model isolate the varying categories of borrowers with respect to their capacity to demonstrate their creditworthiness. In the lowest two categories, borrower households do not have access to formal credit, as they are not able to prove their creditworthiness. Agrarian households with maximum capacity to signal their creditworthiness can meet their entire credit demand from formal lenders. The selection equation identifies factors explaining whether a rural household would avail credit or not. The dependent variable LOAND in the selection equation takes the value one if the household has taken advantage of a loan during 2009– 2010; otherwise, it is zero. As an exchange mechanism, a credit transaction is guided by both the demand for credit and the lender's willingness to supply credit. Further, credit is an inter-temporal transaction, with repayment of the loan following its disbursement after a lag. This exposes the lender to post-contractual default risk, and she will be careful about the timely recovery of her dues. Since credit is extended to the agricultural households, who borrow based on their land holding, the land acts as collateral from the lender's perspective. The lender feels assured that a household with a large track of cultivable land would have a higher production potential and be able to repay the loan obligations without difficulty. Better irrigation facility is also expected to positively boost the demand for credit, as the household may desire to plant multiple crops and is likely to choose credit facility to meet the increased production expenses – including the purchase of inputs, payment of irrigation facility charges, and wages for hired laborers. From the lender's perspective, loan applications from households with better irrigation facilities may also be considered positively, in anticipation of a higher yield as well as the higher profit

12

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

associated with improved irrigation facility. Households having large ruminants – e.g., cattle – would demand a loan to meet the rearing expenses. At the same time, a lender may prefer these households, as the alternative earning source is associated with, in the case of crop failure, a lower probability of default. Lower value of per capita large ruminant would indicate that the household would have lower disposable income to repay loans from any earnings from cattle rearing during the crop failure. Similarly, the holding of assets available for mortgage would act as a strong signal of creditworthiness. Thus, the coefficients associated with land holding, irrigation facilities, per capita large ruminants, and offering of mortgage-able assets are expected to have a positive sign in the selection equation, as they indicate profit potential for any household who may choose to borrow for expanding the respective income-generating activity. The lender may favorably consider loan applications from such households over those of resource-poor applicants. Since households residing in the villages closer to agricultural infrastructure are expected to have higher business potential, they are also more likely to apply for credit to expand their agricultural or other businesses. As formal lenders favor production and enterprise loans over consumption loans, those households would carry a higher probability of receiving a loan. Finally, proximity to formal financial institutions is expected to lead to wider outreach of financial services because of reduced transaction costs to both the borrower and the lender. Thus, the coefficients associated with average distance from agricultural infrastructure and formal financial institutions are expected to carry a negative sign in the selection equation. The ordered probit outcome equation illustrates the factors that influence the likelihood of receiving a loan from the formal sector. The dependent variable is the source of the loan, which is modeled as an ordered variable having the four categories explained in Table 1B. While households with a higher acreage of cultivable land are expected to demand loans to meet their production expenses, their credit proposals are also likely to be approved by the formal lenders due to their profit potential. An ability to offer suitable marketable collateral may also stand as a strong signal of a household's creditworthiness, as the lender can liquidate the collateral in the event of any possible default. Familiarity with key social leaders is expected to improve the credential of the loan applicant as well. Out-migration is expected to dampen a household's probability of getting a loan from the formal sector, due to difficulty in monitoring the loan contract. The presence of interlinked credit is assumed to positively affect the likelihood of access to formal credit sources, as it would increase the lender's involvement in the non-credit market (i.e., insurance, input, and output) and reduce the production and market risks faced by a loanee. Interlinked credit is expected to relax the demand for marketable collateral, as insurance covers the unforeseen production shocks, provision of input would reduce adverse use of loan funds, and linking marketing of outputs would allow the inclusion of future receivables as collateral. Households from villages with better accessibility to agricultural infrastructure are expected to carry higher business potential. The formal lenders would therefore favorably consider the demand for credit from these households. Closeness to the formal financial institution, on one hand, ensures easier accessibility by the borrower; and, on the other, it minimizes the ex-post monitoring cost for the lender. Furthermore, proximity to educational institutions would result in improved educational levels among the villagers, who in turn would be better informed about the process of obtaining credit from the formal lenders. On the lending side, formal lenders may prefer to extend loans to these villages that are in proximity to educational institutions, as those villages are expected to have a higher number of educated households who are expected to make judicious use of loan funds. Moreover, these households would have better employment opportunities in the nonfarm sector and hence would have adequate income to settle the dues. Incidences of formal sector loans may be higher in villages with a higher percentage of the upper caste Hindu population – not due to caste biasness of the lending institution, but rather due to the resources held by these households. Thus, the coefficients of AGRI, FININS, and EDU, are expected to carry negative signs in the ordered probit outcome equation, while PUCASTE would yield a positive sign in the same. Following Greene and Hensher (2010, p.307–308), specification of the model is as follows: Selection equation:

z ⁎ = α1 + α2 LANDD1 + α3 IRRIDUM + α4 PCLRUMI + α5 MTDUM + α6 AGRI + α7 FININS + α8 CGARH + α9 MAHA + α10 WB + α11AP + α12 TN + u,

LOAND = 1[z ⁎ > 0] Ordered probit outcome

y⁎ = β1 + β2 LANDD2 + β3 MTDUM + β4 FAMTOT + β5 MIGRATEDUM + β6 LINK + β7 AGRI + β8 FININS + β9 EDU + β10 PUCASTE + β11CGARH + β12 MAHA + β13 WB + β14 AP + β15 TN + ε,

LOANCAT = 1 if y⁎ ≤ 0 = 2 if 0 < y⁎ ≤ μ1 = 3 if μ1 < y⁎ ≤ μ2 = 4 if y⁎ > μ2 Here, μ1 and μ2 are threshold parameters or cut-points. Observation mechanism: y, x observed when LOAND = 1

(ϵ, u) ~N [(0, 0), (1, ρ , 1)]

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

13

5. Results and discussion The model parameters are estimated using Full Information Maximum Likelihood (FIML) and is implemented in the paper using QLIM procedure of the SAS software. The estimation of the model is presented in Table 7. In the estimated selection equation, all variables except for FININS are observed to have the expected signs on their coefficients; these are significant at 5 percent level. The variable FININS also exhibits the opposite sign than expected in the estimated order probit model. The positive sign of FININS's coefficient is contrary to the common belief that households closer to financial institutions would exhibit greater financial inclusion (see Burgess and Pande, 2005). This finding may imply that even a household physically closer to a financial institution may remain excluded from financial services, unless it has better production resources. While a minimum holding of one hectare of land has come out as a significant determinant for entering the credit market, households with at least two hectares of land are the favorable choice of the formal creditors. As expected, households with better irrigation facility have chosen to obtain credit, as they may require a higher amount of money to meet their production expenses in multiple-crop seasons. Possession of large ruminants is found to have a positive and significant effect on credit choice. This fact, on one hand, boosts the household's demand for credit in order to meet the rearing expenses; on the other, it is likely to strengthen the applicant's creditworthiness to the lender. Unsurprisingly, the ability to offer marketable collateral is observed to positively influence the credit choice of households, as lenders would have chosen those households who are able to mortgage a marketable asset as collateral. Similarly, closeness to agricultural infrastructures, such as markets, has turned out to be a significant variable. This finding seems plausible, as not only have households belonging to villages closer to markets and other agricultural infrastructure chosen to borrow money to expand their agricultural activities and other business avenues, but also their higher profit potential has helped them to gain access to loans from the formal lenders. Familiarity with key social leaders – namely, head of village administration, government extension officers, headmaster of local school, and officials of financial institutions – appear to be crucial for deciding whether a borrower household can obtain a loan from the formal sector. In the rural credit market, this familiarity with key social leaders helps borrowers overcome the information asymmetry and improve their credential as a good borrower in the eyes of the lender. An instance of migration outside the village dampens the household's possibility to negotiate a loan from formal lenders, as expected. Households in which some members have migrated outside the village are found less likely to be serviced by the formal lenders – possibly in anticipation by the lender of potential difficulty in monitoring those households for repayment. The positive sign on the coefficient of LINK shows that a loan linked with at least one credit complementary service would boost the household's possibility of obtaining a loan from the formal sector. This finding seems plausible, as the linking of credit with another complementary service would relax the demand for marketable collateral and hence positively influence the formal sector's approval of a household. Physical proximity to educational institutions, as expected, has emerged as a significant variable in influencing the household's choice of credit from the formal sector. This finding seems plausible, as villages closer to educational institutions are expected to have a higher number of educated households, who are in turn likely to be more aware of credit programs, as well as of the formalities required to obtain a loan from formal sources. Similarly, lenders would also place more confidence in educated households, as they are likely to enjoy better employment opportunities in the non-farm sector and thus have income from additional sources to repay the loan. As expected, in villages with a higher percentage of the upper caste Hindu community, which are traditionally resource-rich, loan applications carry a higher probability of receiving favorable consideration from the formal sector lenders. Table 8 highlights the marginal effects of the independent variables of the ordered probit outcome equation. The marginal effects show, by how many percentage points, the probability of being in one category differs with changes in the value of the response variable (Greene, 2003, p.737–739). The values, in Table 8, were calculated on sample enumeration. Possession of a minimum of two hectares of operational land and property available for mortgage are strong signals of an endowment and demonstrate creditworthiness. Borrower households with at least two hectares of land have a lesser probability by 2.63 percentage points of remaining excluded from formal sector credit sources, as compared to landless borrowers or borrowers with less than two hectares of land. Additionally, borrower households of this category have a higher probability by 9.44 percentage points of meeting their credit demand exclusively from formal sector credit sources, as compared to those with less than two hectares of land. Borrower households with interlinked credit have a lesser probability by 12.10 percentage points of remaining excluded from formal sector creditor sources, as compared to borrowers with non-linked credit. Additionally, interlinked borrowers are found to have a 43.50 percentage higher probability of meeting their credit demand exclusively from formal credit channels, as compared to non-linked borrowers. As noted earlier, the marginal effects corresponding to distance from formal financial institutions have yielded the opposite signs than expected.

6. Conclusion This article draws its motivation from the growing debate on the large-scale exclusion of rural households from the domain of the formal credit delivery channels, even after government interventions have attempted for over a century to

14

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

make credit inclusive. The objective of this paper is to identify the factors affecting rural Indian households' choice of credit from formal sector financial sources over other credit institutions, namely semi-formal and informal lenders. The main contributions of this paper are four-fold. First, compared to previous studies, it uses a more recent and broader dataset covering multiple states and representing the Indian agrarian economy, thus allowing for the results to be more useful for policy formulation. Second, it differs from earlier studies by its use of the sample selected ordered probit model, and we believe this application is the first of its kind in analyzing differential access to credit institutions in an agrarian economy context. Third, the results are contrary to the established wisdom that opening more branches in an unbanked area – i.e., taking the delivery window closer to the target group – would ensure access to formal financial institutions. Finally, the study finds corroborative evidence that offering credit along with other complementary services – namely insurance coverage, provision of inputs, procurement of output, and access to extension services – would help in expanding the domain of formal creditors by relaxing the demand for marketable collateral. The study finds that the formal credit market in rural areas is sensitive to an applicant's household and production characteristics, along with village characteristics. Households that have an operational land holding of at least two hectares, are capable of offering marketable collateral, are familiar with key social leaders, exhibit non-migratory nature, and are engaged in interlinked transactions are more likely to secure credit from formal sources. Households residing in villages that are far away from agricultural infrastructure – namely markets and input stores – and educational institutions are less likely to be covered by formal sector lenders. The present study did not find closeness to the formal financial institutions, which to date is assumed one of the crucial factors for expanding the coverage of formal lenders, to be a factor influencing the credit choice from the formal sector. Instead, we find that resource-rich households (e.g., capable of offering marketable collateral) from distant locations, however, are enjoying better access to formal credit sources over resource-poor households that are closer to formal sector lenders. This finding seems plausible, as formal sector lenders operating under an interest rate regime controlled by the government do not have the flexibility to add any default premium over that of the nominal interest rate to cover any possible default (Beck et al., 2008). Furthermore, unlike informal sector lenders who take advantage of the closely-knit village economy and localized operations that allow them to keep more personal information about their clients, bank officials are mostly outsiders who hold limited information about the borrowing group. Though among the formal financial institutions, RRBs and PACS are supposed to be run by local employees, those institutions are frequently criticized for being controlled by political groups that have vested interests (Vaidyanathan, 2013). Furthermore, a majority of the formal financial institutions in rural areas are single-person branches where an officer or secretary operates either alone or maybe with help from a single support staff. Hence, officials of formal financial institutions in rural areas often blame a heavy workload as a reason for the limited time available to them to procure new business or monitor the existing loan accounts (RBI, 2011). Given this backdrop, to safeguard their interest, formal sector lenders quite often ask for marketable collateral to ensure that the borrower repays the loan on time without much follow-up work on the part of the lender. In extreme cases, the lender may also liquidate the collateral to recover the dues in case the borrower fails to meet her obligations. In a nutshell, even if the government pushes for financial inclusion through opening new branches in unbanked areas and offering credit at a relatively cheaper interest rate, access to formal credit is determined by profitability of the farmer's individual/household projects, as well as the resource endowments that can be offered as suitable collateral to the formal sector lender. Interestingly, offering credit as a package with complementary services, such as insurance and input and output marketing services, improves resource-poor borrowers' chances of being serviced by formal sector lenders. This result may be relevant to policymakers in the present context, when the central bank of India is pushing the bankers to come out from traditional mortgage-based lending and focus on financial inclusion. While banks are presently resorting to credit-alone approaches, they may be suitably advised by policymakers to focus more on a ‘credit plus’ approach, where they would offer credit along with insurance, input selling, output purchasing, and extension services. This approach would boost borrowers' credit demand by reducing their exposure to production and market risk, and it would also relax the demand for marketable collateral. Insurance would help mitigate the production risk, provision of input would control adverse use of loans, and linking output procurement with loans would cover receivable as collateral. While multi-purpose cooperative societies may offer part of these services through their own banks, RRBs, and MFIs may offer them through third parties or subsidiaries. Since the present study deals with the determinants of rural households’ choice of an institution for securing credit, as future research we want to study the effect of access to formal sector credit on agricultural production and other allied activities. Furthermore, while we base our analysis on the traditional utilitarian framework, investigating sectoral credit choice by the rural households from the perspective of the behavioral economics may also be explored in future.

Acknowledgments Authors acknowledge comments from the editor. We would like to thank three anonymous referees for their valuable comments, which have led to substantial improvements to this paper. Authors express their indebtedness to Samar K. Datta, Vasant P. Gandhi, and William H. Greene for their valuable suggestions leading to enrichment of this paper. We have also benefited from the comments and suggestions of the participants of the 88th Annual Conference of the Western Economic Association held at Seattle.We acknowledge doctoral fellowship and dissertation support grant extended by Indian Institute

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

15

Table A1 Distribution of lenders across sample villages in India. State

Village

Type of formal and semi-formal creditora

Chattisgarh

Kendri, Tarpongi Bhatagaon Bahirewadi Amboda Chohottobazar Dhuchnikhali Chorabidya Khulna Metiyakhali Madhusudankati Kesharajpally Eenathi Kotha

RRB, PACS, SHG RRB, MPACS, MFI SCB, MFI, PACS MPACS, SCB PACS, MFI SCB, RRB RRB, PACS, SHG SCB, MFI RRB, PACS, SHG MPACS, SCB, MFI SCB, MPACS SCB, MFI, SHG SCB, SHG SCB, MPACS

Maharashtra

West Bengal

Andhra Pradesh Tamil Nadu Gujarat a

Substantial presence of informal credit sources are found in all villages.

of Management, Ahmedabad to the first author. We are grateful to Samar K. Datta for making available the dataset used in this study. This dataset is collected as part of the Centre for Management in Agriculture, Indian Institute of Management Ahmedabad study titled “Assessing Policy Interventions in Agri-Business and Allied Sector Credit versus Credit Plus Approach for Livelihood Promotion” sponsored by Ministry of Agriculture, Government of India

Appendix A See Table A1.

References Barslund, M., Tarp, A., 2008. Formal and informal rural credit in four provinces of Vietnam. J. Dev. Stud. 44 (4), 485–503. Basu, P., 2006. Improving Access to Finance for India’s Rural Poor. The World Bank, Washington, D.C. Beck, T., Demirguc–Kunt, A., Peria, M.S.M., 2007. Reaching out: access to and use of banking services across countries. J. Financ. Econ. 85 (2), 234–266. Beck, T., Demirguc-Kunt, A., Peria, M.S.M., 2008. Banking services for everyone? Barriers to bank access and use around the world. World Bank Econ. Rev. 22 (3), 397–430. Besley, T., Coate, S., 1995. Group lending, repayment incentives and social collateral. J. Dev. Econ. 46, 1–18. Bottomley, A., 1975. Interest rate determination in underdeveloped rural areas. Am. J. Agric. Econ. 57 (2), 279–291. Boucher, S., Guirkinger, C., 2007. Risk, wealth and sectoral choice in rural credit markets. Am. J. Agric. Econ. 89 (4), 991–1004. Burgess, R., Pande, R., 2005. Do rural banks matter? Evidence from the Indian social banking experiment. Am. Econ. Rev. 95 (3), 780–795. Castellani, D., 2014. Shocks and credit choice in Southern Ethiopia. Agric. Financ. Rev. 74 (1), 87–114. Doan, T., Gibson, J., Holmes, M., 2010. What Determines Credit Participation and Credit Constraints of the Poor in Peri-urban Areas, Vietnam? Working paper, Economics Department, The University of Waikato. De Luca, G., Perotti, V., 2011. Estimation of ordered response variable with sample selection. Stata J. 11 (2), 213–239. Fisher, T., Sriram, M.S., 2002. Beyond Micro-credit: Putting Development Back into Micro-Finance. Sage Publications, New Delhi (Vistaar Imprint). Ghate, P., 2007. Consumer protection in Indian microfinance: lessons from Andhra Pradesh and the microfinance bill. Econ. Polit. Week. 42 (13), 1176–1184. Ghosh, P., Mookherjee, D., Ray, D., 2001. Credit rationing in underdeveloped countries: an overview of the theory. In: Mookherjee, D., Ray, D. (Eds.), A Reader in Development Economics, Blackwell., New Delhi. Government of India, 2011. Agricultural Statistics at a Glance 2011. Ministry of Agriculture, New Delhi. Guirkinger, C., 2008. Understanding the coexistence of formal and informal credit markets in Piura, Peru. World Dev. 36 (8), 1436–1452. Guirkinger, C., Boucher, S.R., 2008. Credit constraints and productivity in Peruvian agriculture. Agric. Econ. 39 (3), 295–308. Greene, W.H., Hensher, D.A., 2010. Modeling Ordered Choices: A Primer, 1st ed Cambridge University Press, Cambridge. Greene, W.H., 2003. Econometric Analysis, 5th ed Pearson Education, New Delhi. Heckman, J., 1979. Sample selection bias as a specification error. Econometrica 47 (1), 153–161. Hoff, K., Stiglitz, J.E., 1990. Imperfect information, and rural credit markets – puzzles and policy perspectives. World Bank Econ. Rev. 4 (3), 235–250. Hassan, M.K., 2008. The microfinance revolution and the grameen bank experience in Bangladesh, financial markets. Inst. Instrum. 11 (3), 205–265. Johnston, J.D., Morduch, J., 2008. The unbanked: evidence from Indonesia. World Bank Econ. Rev. 22 (3), 517–537. Kumar, A., Singh, D.K., Kumar, P., 2007. Performance of rural credit and factors affecting the choice of credit sources. Indian. Indian J. Agric. Econ. 62 (3), 297–313. Liao, T.F., 1994. Interpreting Probability Models: Logit, Probit, and Other Generalized Linear Models. Sage University Press, California. Mahajan, V., Ramola, B.G., 1996. Financial services for the rural poor and women in India: access and sustainability. J. Int. Dev. 8 (2), 211–224. Mpuga, P., 2010. Constraints in access to and demand for rural credit: evidence from Uganda. Afr. Dev. Rev. 22 (1), 115–148. Mohieldin, M.S., Wright, P.W., 2000. Formal and informal credit markets in Egypt. Econ. Dev. Cult. Change 48 (3), 657–670. National Sample Survey Organisation, 2005. Household Indebtedness in India as on 30-06-2002, All-India Debt and Investment Survey, NSS 59th Round (January December 2003), Report No.501 (59/18.2/2). Ministry of Statistics andProgramme Implementation, Government of India.

16

D. Pal, A.K. Laha / Journal of Choice Modelling 14 (2015) 1–16

Pal, S., 2002. Household sectoral choice and effective demand for rural credit in India. Appl. Econ. 34 (14), 1743–1755. Pal, D., 2012. Managing Rural Institutional credit: Lessons from Interlinked Transactions (Doctoral dissertation). Indian Institute of Management Ahmedabad. Reserve Bank of India, 2011. RBI Central Board meets Today at Agartala; Governor meets State Government and Bankers. Press Release May 19. Sarap, K., 1990. Interest rates in backward agriculture: the role of economic and extra economic control. Camb. J. Econ. 14 (1), 93–108. Stiglitz, J., Weiss, A., 1981. Credit rationing in markets with imperfect information. Am. Econ. Rev. 71 (3), 393–410. Stiglitz, J., 1990. Peer monitoring and credit markets. World Bank Econ. Rev. 4 (3), 351–366. Train, K., 1986. Qualitative Choice Analysis: Theory, Econometrics, and an Application to Automobile Demand. MIT Press, Cambridge. Vaidyanathan, A., 2013. Future of Cooperatives in India. Econ. Polit. Week. 48 (18), 30–34. Yadav, S., Otsuka, K., David, C.C., 1992. Segmentation in rural credit financial markets: the case of Nepal. World Dev. 20 (3), 423–436.