An approach using primary evidence from low-income ... · Vaishnavi Prathap and Rachit Khaitan...

Toward Developing Unsuitability Guidelines for Credit Provision

An approach using primary evidence from low-income households in India

Vaishnavi Prathap and Rachit Khaitan

WORKING PAPERAPRIL 2016

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

TOWARD DEVELOPING UNSUITABILITY GUIDELINES FOR CREDIT PROVISION

Table of ContentsList of FiguresList of Tables1. Overview2. Policy Motivation 2.1. Consumer lending in India 2.2. Existing customer protection norms for credit provision in India 2.2.1. Regulatory guidelines 2.2.2. Credit bureau reporting norms 2.2.3. Self-regulatory norms2.3. Inadequacies of the current customer protection regime3. Academic Motivation 3.1. Heterogeneous impacts of microcredit 3.2.Debt-inducedfinancialdistress 3.3. Debt management and negative coping mechanisms4. Research Design5. Economic Characteristics 5.1. Livelihoods and Income 5.2. Consumption expenditure 5.3. Income Surplus 5.3.1. The scope to expand surplus 5.4. Incidence of shocks and adverse events6. Financial Sector Characteristics 6.1. Access to credit in Krishnagiri 6.2. Credit access and use among sample households7. Evidence of Debt-Related Financial Distress 7.1. Debt affordability measures: Debt-to-income and debt-to-surplus ratios 7.2. Delinquency on outstanding loans 7.3. Liquidating assets 7.4. New borrowing 7.5. Noninstitutional borrowing and social networks8. Debt Affordability Assessment at Point of Sale 8.1. Implementing debt affordability assessments and its challenges9. Implications for Policy Design and Financial Services Provision 9.1. Toward a suitability-based customer protection regime 9.2. Toward better delinquency management and recourse 9.3. Implementing debt affordability assessments and its challenges10. Directions for Future Research References Annexure 1. Data Collection Annexure 2. Additional Sample Description Annexure 3. Socioeconomic Indicators of Study Location

........................................................................................................................................................ iv......................................................................................................................................................... v

.......................................................................................................................................................... 1........................................................................................................................................... 2

............................................................................................................. 2................................................ 3

..................................................................................................... 3...................................................................................... 4

...................................................................................................... 4................................................................... 5

................................................................................................................................... 6....................................................................................... 6

.................................................................................................... 7........................................................... 9

............................................................................................................................................ 9............................................................................................................................. 10

................................................................................................................. 10............................................................................................................. 12

................................................................................................................................ 13......................................................................................... 14

.................................................................................... 14.................................................................................................................. 17

........................................................................................................ 17................................................................... 20

............................................................................................. 23................................ 23

............................................................................................... 27............................................................................................................................ 28

............................................................................................................................... 30......................................................................... 32

........................................................................................... 29............................................ 33

............................................................... 34.......................................................... 34......................................................... 35

............................................. 36.................................................................................................................... 36

..................................................................................................................................................... 37...................................................................................................................... 38

............................................................................................. 40...................................................................... 42

iii

WORKING PAPER

List of Figures

iv

Figure 1. MFI lending in India

Figure 2. Occupation profile of sample households

Figure 3. Household income (6 month total)

Figure 4. Monthly income of household with median volatility

Figure 5. Consumption expenditure, including on social events (6-month totals)

Figure 6. Household surplus (6-month total)

Figure 7. Comparison of current income surplus vs. potential surplus

Figure 8. Financial costs of major and/or unplanned events

Figure 9. Opportunity costs of major and/or unplanned events

Figure 10. Incidence of indebtedness

Figure 11. Average outstanding debt per indebted household

Figure 12. Access points in and around Krishnagiri district (Map showing 4 South-Indian states)

Figure 13. Gross Loan Portfolio of Krishnagiri and surrounding districts (Map showing 3 South-Indian states)

Figure 14. Density distribution of DIR, by type of repayment structure

Figure 15. Composition of DIRs by income quintile

Figure 16. Scatterplot of monthly repayment obligations with available household surplus

Figure 17. Some indicators of debt-related financial distress

Figure 18. Savings withdrawals as a coping mechanism, by DSR group

Figure 19. Sale of livestock as a coping mechanism, by DSR group

Figure 20. Sale of land as a coping mechanism, by DSR group

Figure 21. New borrowing as a coping mechanism, by DSR group (Secured borrowing)

Figure 22. New borrowing as a coping mechanism, by DSR group (Unsecured borrowing)

Figure 23. New borrowing as a coping mechanism, by DSR group (Noninstitutional borrowing)

Figure 24. Social network transfers as a coping mechanism, by DSR group

Figure 25. Distribution of household size

Figure 26. Map of Krishnagiri district from Census 2011

................................................................................................................... 2

.............................................................................. 10

.......................................................................................... 11

............................................................. 12

............................. 13

............................................................................................ 13

................................................ 14

................................................................ 15

........................................................... 16

....................................................................................................... 17

............................................................. 17

................................................................................................................... 19

....................................................................................... 19

.................................................. 24

................................................................................. 24

............. 25

................................................................. 27

.............................................. 29

..................................................... 29

............................................................ 29

................... 30

.............. 30

..................................................................................................... 32

........................................ 33

.................................................................................................... 40

........................................................................... 42


List of Tables

v

Table 1. Credit profile of Tamil Nadu

Table 2. Occupation profile of Subiksha’s household

Table 3. Relative composition of consumption expenditure for median household

Table 4. Types of adverse events recorded

Table 5. CRISIL Inclusix scores

Table 6 Sources of credit for JLG and Non-JLG households, by lender

Table 7. Loan amount and cost of borrowing

Table 8. Collateral requirements

Table 9. Terms of repayment

Table 10. Distribution of sample households by overall DSR

Table 11. Incidence of delinquency on outstanding loans, by DSR

Table 12. Purpose of new borrowing, by DSR group

Table 13. Retrospective debt affordability assessments for loans disbursed, July-September 2015

Table 14. Characteristics of the head of household

Table 15. Access to basic facilities

Table 16. Access to Insurance

Table 17. Use of Savings

Table 18. Income profile

Table 19. Occupation profile

Table 20. Vehicle ownership profile

Table 21. Consumer durable ownership profile

Table 22. Property ownership

Table 23. Dwelling characteristics

Table 24. Access to facilities

...................................................................................................... 3

.......................................................................... 11

........................ 12

.......................................................................................... 15

................................................................................................................ 18

............................................ 20

........................................................................................ 21

............................................................................................................ 21

................................................................................................................... 22

.................................................................. 26

..................................................... 28

............................................................................... 31

................................................................................................................. 33

............................................................................... 40

........................................................................................................... 41

.................................................................................................................. 41

............................................................................................................................. 41

............................................................................................................................. 42

.................................................................................................................... 43

......................................................................................................... 43

....................................................................................... 43

.................................................................................................................... 44

........................................................................................................... 44

.................................................................................................................... 44

WORKING PAPER


Consumer lending in India has rapidly expanded in the past decade and has been motivated by a strong focus on financial inclusion, services a large number rural, low-income clients with little former formalfinancialsectorexperience.Thereisastrongimperativeforcustomerprotectionnormstoaccompanythisrapidgrowth,tosustainthebenefitsoffinancialinclusionthataccruetocustomers,communities,andthelarger economy. Through this research we seek to inform and facilitate the development of guidelines for suitable credit delivery to low-income households—guidelines that will help to prevent adverse outcomes to borrowers that might arise from an unsuitable sale.

We collect rich primary data through repeated structured interviews with 411 rural households in Tamil Nadu. The analysis of primary data and additional analysis of secondary data reveal that these households areactive users of both formal and informal financial services,more so than theaverage. Further,weoversample microcredit borrowers to ensure adequate representation of various products and practices thatdominatemicrolending in India today. Specificallyonmicrocredit, sample respondents reportedavarietyofcontractfeatures—includinglongertenures,bothweeklyandmonthlyrepaymentschedulesandloans with variable equated monthly installments (EMIs) that allow customer to front-load prepayments. Respondentsalsoreportedborrowingfrommultiplesourcesofinstitutionalandinformalcredit,oftenatthesame time.

Wefindthat13percentofsamplehouseholds(bothJLGborrowersandnonusers)werecarryingunaffordablelevelsofdebt,wheremonthlyrepaymentobligationswereinexcessoftheirincomeandtomeetalltheserepaymentsinfullandon-time,borrowerswouldneedtomakecostlyadjustmentsincludingreductionsinconsumption expenditure beyond a reasonable minimum living standard. Simulating a retrospective debt affordabilityassessmenttoasubsetofloansdisbursedinthelatermonthsofthestudy’sobservation,wefindthatnearly61percentofthoseloans,thoughtheywerereceivedatatimeofneed(andsomeproportionoftheseloanswouldbeusedtorepayotherloans),lefthouseholdsworseoff,strictlyinacashflowsense—asperestimatesatthepointofsale,borrowersdidnothavethemeanstoservicetherepaymentsontheseloans through their naturalcash flows.Wealsopresentevidenceofover-indebtedhouseholdsdrawingdowntheirsavings,sellingassets(livestock,land)andtakingonmuchmoredebttoeasecashflowsandfinancerepayments.

Further,webelievethatbyconstruction,thismeasurewouldunderestimateover-indebtedness.Someformalloans(suchasthepopularjewelloan)andmanyinformalloanshaveflexiblerepaymentschedules—loantenuresorrepaymentsfrequenciesarenotspecified,sometimesboth.IfourcalculationsweretoassumethathouseholdswouldservicealloutstandingloansasifthroughEMI-stylemonthlyrepayments,casesofdebt unaffordability would rise. Our current analysis also does not account for the seasonality of income flows over the year nor thedifferences in income frequencywithin amonth. Both of theseare criticalelementsofproduct suitabilityand theaffordabilityof repayments,althougheffective savingsproductscould smoothen away some of the distress.

Theresearchpresentedinthisreportraisesseveralquestionsforfurtherinvestigation,mostsalientlyontheimpact of access to microcredit on borrowers’ resilience vis-à-vis the vulnerability of debt entrapment. Early insightssuggestthatboththeseimpactswouldbeheterogeneousandamongmanyotherfactors,localfinancialsectordevelopment(ofcredit,savings,andinsurance)andpracticeswillsignificantly influenceborrowers’ vulnerability to negative outcomes.

Thoughthisresearchwillcontinuetobeupdatedoverseveralmonths,thereareseveralearlytakeawaysforpractitionersandpolicymakers.First,wepresentuniqueevidenceonthelevelandqualitiesofdebt-inducedfinancialdistressamongmicrocreditborrowers,whilealsorepresentingthelocalfinancialsectorcontextinwhichsimilarlevelsmightbeexpected.Second,teamsmonitoringandregulatingover-lending,wefindthatpubliclyavailabledata,whenwelltriangulated,mightbeahelpfultooltofocusfurthermonitoringefforts.

1. Overview

1

The motivation for this work stems from the current state of consumer lending in India which is rapidly expanding and increasingly providing access to credit to rural households in India. Where there is an imperative for strongcustomerprotectionnorms toaccompany such stronggrowth, it is unclearwhether theexistingregimeisadequateinprotectinghouseholds,particularlyinruralIndia,fromadversefinancialoutcomes.The following subsections elaborate on the inadequacies and present the rationale for developing suitability guidelines for more responsible credit provision.

2.1. Consumer lending in IndiaThepenetrationofformalcreditpenetrationinIndiaremainslowanddominatedbythejoint-liabilitygroup(JLG)loanproduct.Thisismostlysoldthroughtheregulatedchannelofmicrofinanceinstitutions(MFIs),butalsoothernonbankfinancialcompanies(NBFCs).

With a gross loan portfolio of Rs. 57 billion, Tamil Nadu (TN) endedFY 2015 as the second largestmicrocreditmarketinIndia,secondonlytoWestBengalatRs.60billion.Sector reports analysing the growth pattern in top states have drawn attention to the deepening of client relationships as the key driver ofgrowthinseveraltopstates,1 over branch expansion or new client acquisitions. Consider for example that the loan portfolio grew by 51 percent in TN and 67 percent in Karnataka(twostatesadjacenttoourstudylocation),primarilydrivenby equivalent or higher increases in loandisbursement,followedbytheaddition of new clients in existing service areas.

2. Policy Motivation

Third,weexploretheconstructionofadebtaffordabilityassessmentforlow-incomeborrowersandinfutureoutputsfromthisresearch,willseektodemonstratethefeasibilityofimplementingsuchanapproach.Wecall for researchersandpractitioners in the larger financial inclusioncommunity toconsider replicationssimilar to debt affordability measurement in their own contexts—many of the fundamental implementation challengessuchastheneedforhigh-qualitycreditinformation,theneedtostandardizewell-developedtools and adequately train loan officers, and the informality and volatility of incomes for low-incomehouseholdsaresimilaracrossseveralcountries,andthereisatremendousopportunityforinsightfulresearchtoinforminternationalfinancialinclusionpracticeandcustomerprotectionpolicy.

2

1. Sriram,MS,InclusiveFinanceIndiaReport(2015).

Figure 1. MFI Lending in India

10,000 20,000 30,000 40,000 50,000 60,000

Average outstanding loan per client

Clients pre branch

Loan disbursed per branch (in Rs. thousands)

Number of clients per employee

Loan disbursed per employee (in Rs.

thousands)

WORKING PAPER

3

2. Nondistinct3. RBIMasterCircularonNBFC-MFIs:https://www.rbi.org.in/scripts/BS_ViewMasCirculardetails.aspx?id=9012#f14

Table 1. Credit profile of Tamil Nadu

2013–14 2014–15 Growth (%)

Tamil NaduGrossloanportfolio(inRs.billion) 37.86 57 51

Number of clients2 3,825,377 4,868,056 27Number of branches 1,250 1,281 2

Loans disbursed (in Rs. billion) 45.96 69.41 51Number of loans disbursed 3,350,517 4,604,561 37

KarnatakaGrossloanportfolio(inRs.billion) 26.21 43.7 67

Number of clients 2,492,444 2,942,949 18Number of branches 881 1,040 18


All IndiaGrossloanportfolio(inRs.billion) 241.3 390.79 62

Number of clients 24,356,209 29,885,196 23Number of branches 65,278 79,396 22


2.2. Existing customer protection norms for credit provision in IndiaCurrentnormstoprotectcustomersofmicrofinanceproductsthroughtheNBFC-MFIchannelaregovernedprimarilybytheregulatorReserveBankofIndia(RBI).

2.2.1. Regulatory guidelines

ThelatestRBIMasterCircularonNBFC-MFIs3definesthefollowingguidelinesoncustomerprotection:

• Caps on loan price▫ Loanpricemarginnot toexceed10percent for largeMFIs (loanportfoliosexceedingRs. 100

crore) and 12 percent for others.▫ Interest ratetobe lowerof(a)thecostof fundsplusmargin(b)averagebaserateofthefive

largestcommercialbanks(advisedbyRBIeveryquarter)byassetsmultipliedby2.75.


WORKING PAPER

4

4. RBI directive on Membership of Credit Information Companies (January 2015): https://www.rbi.org.in/Scripts/NotificationUser.aspx?Id=10227&Mode=0

5. InclusiveFinanceIndiaReport(2015)6. RBIcircularonSROsforMFIs:https://www.rbi.org.in/scripts/BS_PressReleaseDisplay.aspx?prid=300527. MFCodeofConduct:http://mfinindia.org/wp-content/uploads/2015/12/2nd-Edition-of-the-MFIN-CoC.pdf

• Caps on loan size and tenure ▫ LoanstobedisbursedtoruralhouseholdswithincomenotexceedingRs.1lakh,andurbanand

semi-urbanhouseholdswithincomenotexceedingRs.60,000.▫ LoanamountnottoexceedRs.60,000infirstcycleandRs.1lakhinsubsequentcycles.▫ Total indebtedness not to exceed Rs. 1 lakh (loans toward meeting education and medical

expenses shall be excluded in indebtedness calculation).▫ Tenure not to be less than 24 months for loan amounts in excess of Rs. 15 thousand with prepayment

penalty.• Caps on loans per borrower

▫ BorrowercannotbeamemberofmorethanoneSHG/JLG.▫ NotmorethantwoNBFC-MFIsshouldlendtothesameborrower.

• Repayment▫ Minimum period of moratorium between disbursement and due date for repayment of first

installment should be less than then frequency of repayment.▫ Loanisrepayableonweekly,fortnightlyormonthlyinstallmentsatthechoiceofborrower.

• Loan use▫ Aggregateamountofloansgivenforincomegenerationshouldnotbelessthan50percentof

disbursed.

2.2.2. Credit bureau reporting norms

Incompliance,MFIsaremandatedtoreportallnewloanapplicantstoacreditbureauandconfirmtheirloan eligibility before approving or disbursing a loan.4Whilespecificweaknessesintheauthenticationandupdationofcreditinformationcontinuetopersist,almostallMFIshaveimplementedthesenorms.

Creditbureausalsotrack,aggregateandfrequentlyreporttheincidenceofnon-compliantinquires,i.e.,potentialloanapplicationsthatwererejectedonthebasisthattheapplicantalreadyheldactiveloansinexcess of the lending limit or was in default of one or more loans.

Asperdatareportedin2015,5 Tamil Nadu ranked second in the incidence of noncompliant loan applications when8.57percentof the inquiries submittedwere rejected,alongsideMaharashtra (8.97percent)andKarnataka(8.55percent).Comparablemarketsegmentsthathadsimilarstatisticsweredistrictswith26–30MFIs(8.84percent),thetop50districtsasrankedbythenumberofborrowers(9.28percent)andthenext50districts(8.39percent).

2.2.3. Self-regulatory norms

Two self-regulatory organisations, M-FIN and Sa-Dhan, as RBI-approved industry associations6 are also responsible for ensuring customer protection measures through adherence to a Code of Conduct. The latestCodeofConduct for theMicrofinance Industry7 includes norms on “avoiding over-indebtedness” includingself-regulatingcommitmentsasMFIsto:


5

1. Conduct proper due diligence to assess the need and repayment capacity of client and make loans commensurate with the client’s ability to repay

2. Not be the third lender to a client3. NotbreachthedebtlimitprescribedbyRBI(theregulator)4. RestricttheloanamountdisbursedtoRs.60,000perborrowerforloansonthebasisofjoint-liability,with

the possibility of exceeding the limit as long as long is given out on an individual basis5. Carryouttestchecksofefficacyofprocessesrelatingtoavoidanceofover-indebtedness6. ReduceerrorsinidentificationthroughAadhaar-basedKYC

2.3. Inadequacies of the current customer protection regimeIn thecurrent regime, thereare severalaspects thatmaynotadequately safeguardcustomersof JLGloansfromadversefinancialoutcomes,whicharehighlightedasfollows.Thereisincreasingevidencefromtheindustryabouttheseinadequaciesmanifestinginadverseclientoutcomes,sometimestoanextremedegree.

Debt affordability

• Assessment based on credit bureau checks: The assessment of debt affordability when carried out through querying the credit bureau profiles of customers provides an incomplete at best pictureof household indebtedness.While there has beena recent significant regulatory development tomandatereportingoftheSHGchannel,bureauinformationremainsfarfromacompletepicturewheninformal credit constitutes a large part of household indebtedness in rural India.

• Restrictions based on caps and limits: The regulatory and self-regulatory norms specify thresholds which seem fairly arbitrary and may not be applicable to the situations of different borrowers with a variety of incomeprofilesandsituatedindifferentregional locations.Thehardcapsonloanamountsandnumber of active loans could exclude households with otherwise genuine needs and ability to repay. Conversely,suchstarkrestrictionsthiscouldalsoleadprovidersto“lendtothelimit”withoutassessingdebtaffordabilityinmoremeaningfulways.Ultimately,suchlendingrestrictionsbasedoncapsandlimitscouldcreate incentives forprovidermisreportingandborrowerKYCmanipulation, resulting inover-borrowing and issues like “ghost-borrowing” or “pipelining.”

Specifications of loan obligations

ObligationsofJLGloanproductswhichwheninteractedwithhouseholdportfolioscanresultinadversefinancialoutcomes.

• Criteriaforloanutilization:Theregulatorymandateforatleast50percentofaprovider’sloanportfoliobeing toward income-generation isdifficult tomonitorafter thepointof sale. Thismandatebeingrelaxed over the years is in itself might be testament to loan use for consumption owing to the fungibility of loan amounts.

• Repayment schedules: While structured repayment schedules could encourage customers to make disciplinedrepayments,loanuseforconsumptionsmoothingoftentimessupersedesthepointofgoodmoneymanagementintoacuteliquiditymismatchesresultinginfinancialdistress.

WORKING PAPER

6

Thisresearchismotivatedbythreebroadthemeswellrepresentedintheacademicliteratureonfinancialservicesforlow-incomehouseholds.Thefirstthemedrawsoninsightsfromrigorousevaluationsofmicrocreditaccessanduse.Moststudieshavereportedmodestpositiveyet,nottransformativeeffectsonfinancialwell-being,whilesomehavefurtherexploredpopulationsubgroupsforwhommicrocreditwasmoresuccessfulthan on average. The second theme references a large and multidisciplinary body of work that evidences negativeoutcomesofmicrocreditforcertainpopulationsubgroups,alsomakenoteofsystematicweaknesson both the demand and supply side that leave low-income borrowers vulnerable to a range of negative outcomes.Followingfromthis,thethirdthemeexploresinmoredetailvariousmanifestationsofdebt-inducedstress,andthestrategiesthatlow-incomehouseholdsadopttomanage,ifnotmitigate,financialdistress.

3.1. Heterogeneous impacts of microcreditRecent impact evaluations have questioned and successfully countered the claim that microcredit interventions, through unleashing latent entrepreneurship, could eradicate want in low-incomecommunities. A common premise of the studies reviewed here is that latent entrepreneurial talent is both unobservableandheterogeneousatthetimeofgrantingnewcreditaccess.Whengivenaccesstocredit,microentrepreneursmayinvestmoreintheirbusinesses,orexperimentwithnewtechniquesandovertheshort-term, both actions generate valuable insights on entrepreneurial capability: both internally withinthefirmandtoalimitedextent,externallytolenders.However,aprimaryoutcomeofincrementalcredit

3. Academic Motivation

Transparency and client education

Whileregulatorymandatesareinplacetoincreaseproducttransparencyandenhanceclienteducation,relying on such approaches still places too much responsibility on customers to decide their level of indebtedness and ultimately prevent bad outcomes.

Delinquency as a signal of financial distress

The industry has relied on repayment frequency (or lack of delinquency) as a measure of overall success. However,delinquencymightbe,atbest,abelatedmeasureofcustomerfinancialdistress.

“We see a lot of heterogeneity. We see some people seem to be doing really well when they get access to microcredit, [say] 5% of the population, maybe. And for the rest it seems to have very little impact. So how do we target that 5%? This is a set of issues that seems to be about product design. What is it that we want microcredit to be?”

—Prof. Abhijeet Banerjee, February 20158

8. At‘FinancialServicesforthePoor:LessonsandImplicationsoftheLatestResearchonCredit’,adisseminationeventhostedbytheWorldBank.Onlinerecordingavailableathttp://live.worldbank.org/financial-services-for-the-poor


7

access,subsequentinvestmentandexperimentationisthelikelihoodoffailureandinfact,asKarlan,Knight,andUdry (2012) show, theaveragemicroentrepreneur’sattempts togrowaremore likely to fail that tosucceed. Through their own research and a review of three similar studies they report that only a small set ofmicrobusinessesreceivingarandomizedcapitalinfusionoutperformedthecounterfactualgroup,whilemany in the treatment group experienced negative returns.

Towardabetterpredictionofthepositivereturns(whichcouldtherebyallowmoretargetedmicrolending),Banerjeeetal. (2014)marka fundamentaldistinctionbetween two typesofmicroentrepreneurson thebasis of their entrepreneurial intent.9Throughthelensofthisdistinction,theauthorsdemonstratethatalmostall the positive impacts of microcredit access (increased business profits, household consumption anddurableassetownership)accruedonlytothosewhowerenaturalentrepreneursand/orwerelikelytohavebusinessesevenbeforegainingnewaccesstocredit.Together,theseresultssuggestthatentrepreneurialabilityandefficiencymightbeakeypreconditionforcreditaccessinitselftotranslateintowell-being.

Other factorsbesidesentrepreneurshiphavealsobeenshowntosystematically influencethetypesandmagnitudeofbenefitsthataccruetomicrocreditusers,andthesebecomeespeciallysalientinthecontextofborrowerswhouse their loans to smoothconsumptionormakenonbusiness investments, including inhumancapital.Angeluccietal.(2013)documentbothindividualcharacteristics(suchashavingpriorcreditexperiencewith the formal sector,borrowers’ levelof risk toleranceandpatience)andaswell, overallcommunity characteristics that mediate the impact of microcredit on borrowers.

In this research featured in this report we acknowledge heterogeneity in both (a) how low-income borrowers use their loans and (b) how these loans might interact with borrower characteristics to mediate subsequent outcomes. An in-depth understanding of both,weargue, can informmarket standards for responsiblecustomer acquisition and underwriting.

3.2. Debt-induced financial distressWhilethereisnowstrongevidenceofheterogeneouspositiveimpacts,thereisalsoevidenceofharmfulornegativeoutcomes,althoughthisevidenceismixedwhencomparingevaluationsconductedindissimilarsettings.Forexample,respondentsintreatmentclustersinHyderabad,India,reportedonanaveragethattheyweremoreworriedandlesshappy(bothmeasuredthroughconstructedindicesinBanerjee[2014]),although it is unclear whether higher levels of debt are a possible cause for the reported distress.

It is importantat this stagetonotethedistinctiondrawnbetweenfinancialandpsychologicaldistress—creditwouldlikelybelinkedtobothandindeed,bothfindtheirplaceinevaluationsofmicrocredit.Theoriginal data collection and analysis presented in later sections of this report are limited to the antecedents andconsequencesoffinancialstressonly,fortworeasons.First,studiesontheincidenceofmentalhealthproblemsrevealedthatfinancialstressisnotonlyaleadingcauseforconcernworldwide,butparticularlysoindevelopingcountries(Fieldetal.2012).Second,mostcomparableinternationalpolicyandregulationtoprotectcustomersarticulatetheneedforfinancialservicesproviderstoassessandmitigatefinancialdistressfor their customers.

Scholarsassertthattounderstandtheroleofmicrocreditindebt-inducedfinancialdistress,researchfocusmustshiftfromimpactscreatedaftertheloan,totherepaymentsmadeduringtheloan(Dattasarmaetal.2015;Fieldetal.2012;Venkataetal.2010).Venkataetal.(2010)documentthatmicrocreditborrowers

9. Thefirsttype is themuch-celebratedsegmentofpoormicrofinanceclientsasnaturalentrepreneurswhocouldrapidlydeveloplocal,profitablemicrobusinesses—theauthorscallthem“gung-ho”entrepreneurs.Incontrastarereluctantentrepreneurswhomightactuallyprefertoworkelsewhere(inregularwageemployment,forexample)but are instead called on to start businesses or other pursuits to repay emergency or consumption loans.

WORKING PAPER

8

attributedfinancialstresstheyexperiencedduringtheloantenuretoseasonalityrelatedincomevolatility,life-cycle events and related expense volatility and the additional burden of repayment in the event of delinquencybyothersintheirjoint-liabilitygroups.

On loan repayments andcredit contracts, twoaspects findprominence in the literature. First,whetherthe loan’sfeaturesareappropriatetoborrowers’circumstances inamannerthatminimizesharm, ifnotencouragesuccess.Fieldetal.(2012)findthatclientswithamonthlyrepaymentschedulescored45percentloweronaFinancialStressIndexthanclientswithrepaymentsdueeveryweek.Further,theyobservedthatthis differential was particularly large toward the end of the loan (correlated with a large differential in theaverageincomesearnedbythetwosubgroups),demonstratinghowmicrocredit’sfeaturesholdthepotentialtobothinduceandmitigatefinancialstress.

The second feature of credit services that are salient in the discussion of repayment burden is the question ofwhetheranadditionalloan,orloanaccess,posesariskofover-indebtedness.Theconsequencesofover-indebtednessforborrowerscanbesevere,butwilllikelydifferdependingon,amongotherfactors,theextentof over-indebtedness and the strategies they adopt to manage this debt. Transitional over-indebtedness maycharacterizedbylevelsofdebtthatarehighrelativetoincome,butnotoverwhelminglyso;borrowersmightbeabletousevariousstrategiestostabilizeandreducetotaldebtoverthemediumterm(Guerinetal.2011),withoutsignificantlyerodingtheirassetholdings.However,borrowerscarryingpersistentlyhigherlevelsofdebt(suchaswhentherepaymentsdueareequaltoincome,orhigher)wereobservedjugglingvarioussourcesof funds tomaintain theircreditworthiness,andwerealso likely toexperiencesignificantasseterosion(Grammling2009;Guerinetal.2011),pauperization,orextremedependenceonsupportfromsocial networks.

SchicksandRosenberg(2011)reviewedsixfieldstudiesofmicrofinanceover-indebtednessandreportedthatanywherebetween12percentand85percentofthestudypopulationswereclassifiedasover-indebted.10 Thedefinitionsofover-indebtednessineachstudy,however,werequitevaried:theycouldbemeasuredbytheoutcomesofdebt(delinquencyordefault,althoughthesearetypicallylaggingindicators),byborrowers’financialcharacteristics(numberofloans,debtserviceratio,anddebt-assetleverageratio,tonameafew)orbythenatureofsacrificesmadebyborrowerstomeetrepayments.Schicks(2013)furtherexploresthemeritsofsubjectivemeasuresofover-indebtednesssuchaswhenborrower’sself-identifycostlyadjustments,undulyhighsacrificesthatwerenotanticipatedatthetimeoftakingontheloanand“regrettable”debt.

However,theuseofsubjectivedefinitionstopreventover-indebtedness(say,atthetimeofloanappraisal)hasstronglimitations,andweobservethat inapplication,regulatorshavetendedtopreferquantitativemeasures.Forexample,regulationsinMalaysiaandNicaraguaprescribeaceilingontheratioofdebt-to-income,particularlyforconsumer lending(CGAP2009). InSouthAfrica,theNationalCreditActrequiresthat all providers conduct a thorough affordability assessment of prospective borrowers through a check on both their incomes and outstanding debts: a borrower would be considered over-indebted through the new loan if the total of monthly debt payments was in excess of income net of minimum living expenses (WorldBank2012).

The research featured in this report will consider both loan features and indebtedness levels in relation to financialdistress.Wealsoexploresomeofthechallengesinimplementingdebtaffordabilityassessmentsinlow-incomesettings,andsimilarregulationsaimedatmitigatingdebtstress.

10. Theauthorsnotethattheseestimatesofover-indebtednessarenottobeconsideredrepresentativeofmarketseverywhere. Several of the studies reviewed were conducted in areas known to have over-lending or lending crises.


9

3.3. Debt management and negative coping mechanismsLow-incomehouseholdsareknowntofacehighincomevolatility,extremevulnerabilitytoillnessandinjuryandunpredictablelivelihoodopportunities—andtocope,useavarietyoffinancialstrategiestomanagetheircashflows.Copingstrategiesmight includeonsmoothening incomethrough incomediversificationorbalancingfarmandnon-farmemployment.Alternatively(orsometimes,additionally)copingstrategiesmightinsteadfocusonsmootheningexpenditurebysupplementingincomegapsthroughsavings,insuranceand borrowing arrangements (Morduch 1995). These mechanisms help protect essential consumption not only from incomevolatility,butalso from shocks to livelihoodsor tomanage largeexpenses. There isagrowingbodyofevidenceonthevibrantfinanciallivesoflow-incomehouseholdsandthevarietyofwaysinwhichtheyuseformalandinformalfinancialservicestomanagevolatilityandbuild-inadditionalflexibilityintimesofneed(Collinsetal2009;Zollmanetal.2015).

Specificallyinthecontextofmanagingthestressofdebtrepayments,afewrecentstudieshaveexpressedconcern over certain coping mechanisms that could potentially prove harmful if persistent over the mediumorlong-term.Exploitinganunexpectedtemporaryshut-downofmicrofinanceinKarnataka,India,Dattasarmaetal.(2015)documentdramaticchangestothehouseholdconsumptionexpenditurepatternswhen repayment obligations were relaxed. The relaxation of repayment obligations during a temporary banwasobservedtosignificantlybridgeexpensedifferentialsacrosshouseholdonessentialconsumption—notably,grainsandvegetableswhilethesignificantgapsthatremainedwereondiscretionaryspendingundertheheadssnacks,“giventohusband,”fuel,meatandsweets.Thesefindingssupportselectborrowertestimonies (MIX 2015) where about 5 percent respondents self-reported that they struggled to makerepaymentson-timeandinordertodoso,theywouldsometimesreduceorskipmeals.Respondentsinthissurveyalsoreportedthatintimesofstress,repaymentswouldbemadethroughborrowingfromanothersource,however,itisunclearunderwhatcircumstanceseachofthese(oranyothercopingmechanisms)were available to stressed borrowers and how they were chosen.

Inthisresearch,oneofourhypothesesisthatlocalsocioeconomicandculturalcharacteristics,thehistoryoflocalfinancialdevelopmentandthestrengthofcommunitysupportnetworkswillsignificantlymediatehowborrowerscopewithfinancialdistress.

Thedesignofthisstudyispremisedonthefoundationthatfinancialproductsaredeeplyintegratedwithlivesof theirusers. There isamarked role for FSPs tobemore responsible for theproductsand servicesthey provide to vulnerable customers and recent regulatory changes have required lenders to proactively prevent adverse outcomes to the borrower that might arise from an unsuitable sale.11 The outputs of this research are intended to inform and facilitate the development of suitability or unsuitability guidelines for lending to low-income households.

Inlightoftheimperativetomeaningfullyimplementsuitabilityguidelinesforcreditprovision,thisresearchseeks to address the following questions:

1. Howdoesacreditproductinteractwiththefinancialportfoliosoflow-incomehouseholds?2. Isthereevidenceofadverseoutcomescorrelatedwith,orasaresultofcertaintypesofcredit?3. Canthisevidenceinformresponsiblecreditdelivery?

4. Research Design

11. A more detailed discussion of a suitability-based customer protection regime is available in the Sahasranamam et al.(2014)positionpaperoncustomerprotectionandthechapteronCustomerProtectionintheRBICommitteeofComprehensiveFinancialServicesreport.

WORKING PAPER

10

The primary data collection for this study was located in Krishnagiri district in Tamil Nadu, South India.Householdswererecruitedintothesamplefollowingastratifiedrandomsamplingprotocol,thoughover-sampling JLGborrowersandalso takingcare to includecompletegroups, i.e., allmemberhouseholdsof a borrowing group were included in the sample. Through detailed monthly interviews with 411 panel households, we construct a longitudinal dataset presenting a comprehensive picture of economiccharacteristics,financialaccess,financialbehaviorandcashflows.Adescriptionofthesamplingprotocolis presented as an appendix and summary descriptive characteristics pertinent to sample households are discussed in the next chapter.

Inanalysisweseek to test the followinghypotheses.First,weseek todevelopmeasuresofdebt-relatedfinancialdistressexperiencedbyborrowers.Wetestwhethertraditionalmeasuresofdefaultordelinquencyfairlyestimatelevelsofunderlyingfinancialdistress.Second,weseektounderstandtheroleofunexpectedeventsthatmayoccurduringthe loantenuretoexacerbatefinancialdistress.Third,weseektopresentevidenceonthevarietyofcopingmechanismsthatborrowersadopttomitigatestress,includingonesthatcould prove harmful over the medium term or longer term.

5.1. Livelihoods and IncomeThe largestportionofsamplehouseholds reportedworking incasualwage labor(23percent), followedbyhouseholdenterprises(18percent),andregularsalariedprofessions(16percent).Thesefivecategoriesalone(seeFigure2)accountedfortheoccupationsof76percentofindividualsintheoverallsample.

5. Economic Characteristics

Figure 2. Occupation profile of sample households


11

ConsiderSubiksha,whosehouseholdis at the median of the income distribution. She reported that about 71 percent of her family’s income from casual labor and about 30percent of their income through self-employment.

Basedonmonthlyincomereports,wealso estimate high month-on-month income volatility, similar to the richevidence from financial diaries of

Table 2. Occupation profile of Subiksha’s household

Occupation Income (%)

Casual labor (nonpublic works) 67

Household enterprise (as own account worker) 22

Household enterprise (as employer) 8

Casuallabour(publicworks/NREGA) 4

Miscellaneous 0

Total (Rs.) 100

Figure 3. Household income (6-month total)

In addition to employment income, households also receive income through rent on landor property,remittancesfromhouseholdmembers,pension,andotherinstitutionalorgovernmentwelfaresupport.ThemedianmonthlyhouseholdincomethusreportedforallsamplehouseholdsrangedbetweenRs.3,893atthelowerendtoRs.24,209atthehigherend,whilethemedianwasRs.9,614.

WORKING PAPER

12

low-incomehouseholds.Narayan’shousehold,atthemedianofthevolatilitydistribution,mightrepresentthe average experience of low-income households. Even within a short observation period (approximately sevenmonths),hisfamily’smonthlyincomefell28percentbelowtheiraverageinonemonth,whilerose66percenthigherjusttwomonthslater.

5.2. Consumption expenditureIneachsurveyvisit,householdsreportedtheexpensesthey incurred under various heads, including adetailed schedule of food and nonfood essential items, tuition and related education expenses,and medical expenses. Large purchases such as durable assets, personal transport vehicles, andexpenses incurred on ceremonies and celebrations arerecordedseparately.Forthemedianhouseholdamajorityofconsumptionexpenditurewasonfooditems (47 percent), followed by celebrations andceremonies (33 percent), and nonfood essentials(13 percent).

Relative composition of consumption Table 3. expenditure for median household

Category % total expenditure

Food 47

Medical 6

Regular (nonfood) 13

Events 33

Total (Rs.) 100

Figure 4. Monthly income of household with median volatility


13

5.3. Income SurplusThe median monthly household surplus (calculated as total household income net of total household consumption)reportedacrossallsamplehouseholds inthestudy’sobservationwasRs.3,936.Thesurplusrangedfrom(negative)Rs.1,488toRs.14,982.Thisincludesincomegeneratedfromprimaryandsecondaryoccupations,miscellaneous sources, andconsumptionof food,medicineand treatment, other regularnonfooditems,andeventssuchasweddings,festivals,andfunerals.

Figure 5. Consumption expenditure, including on social events (6-month totals)

Figure 6. Household surplus (6-month total)

317

266

WORKING PAPER

14

5.3.1. The scope to expand surplus

Manystudieshavedocumentedavarietyofwaysinwhichlow-incomehouseholdsstretchtheirincomes,building inmoreelasticity in timesofneed (seeZollmanetal.2015 foradetaileddiscussiononbudgetelasticity).Mostofthesestrategiesfallintoafewbroadcategories—increaseincomes,reduceexpenses,receiveothernonincomeresourcesthroughsocialnetworks,orusecredit.

Specificallyon reducingexpenses,weexplorehowmanyhouseholds in the samplehave theability toreduceexpensesfurther,estimatingtheminimumlivingstandardasthe10thpercentileofthemonthlypercapitaexpenditurereportedforTamilnadu(NSSO2011).Eighteenpercentofhouseholds’currentreportedconsumption in thesurvey isbelowthisminimumthreshold;a further reductionofexpensesasacopingmechanismcouldhaveseriousconsequencesoverthelongerterm.Oftheremaininghouseholds,manyarejustabovethethresholdandalsohavelittleroomtoreduceconsumption,thoughiftheywerealltodroptheirconsumptiontotheminimumlevel,onaverage,itwouldfreeupanadditional25percentoftheirincomes for other payments.

Figure 7. Comparison of current income surplus vs. potential surplus

5.4. Incidence of shocks and adverse eventsThehousehold surveys record the incidenceofanycelebrations, unplannedevents, oradverse shocksoccuringinthehouseholdinthepastsevenmonths.Theeventsrecordedfall intofivebroadcategories,listed in Table 4.


15

Table 4. Types of adverse events recorded

Event category ExamplesFamily or social

events Birthdaysandweddings,religiousceremonies,localsocialcelebrations,andallnotablefestivals.

Health shocks Illnessorinjuryofahouseholdmember,ordeathofahouseholdmember

Shocks to assets or property

Repairsofmaintenancetohouseordurableassets,incidenceoflivestockillness,andthelossofassets (physical and livestock) to crime

Shocks to livelihoods

In the case of agriculture or self-employment:Delaysinprocurementofinputsorinpaymentsreceivable,unexpectedpricechanges,lossordamage suffered to crops

In the case of laborers or service providers:Impossibletofindwork

Climatic changes Extremeheatorcold,delayed/excessiverain,floods,andothernaturalcalamities

Foreachevent,thesurveysalsorecordedthetotalexpensesincurredandnumberofworkingdayslostduringtheevent(foranestimateoffinancialandopportunitycosts,respectively).Whereverincomewasforegoneoradditionalexpenseswereincurred,investigatorsprobedtounderstandhowhouseholdsmanagedtheirbudgets during that period.

Coping strategies were marked in a precoded list and used a reference for the administration of the rest ofsurvey,suchastheincome,savings,andborrowingmodules.Sincecopingstrategieswereself-reported,theywillalsoneedtobetriangulatedwithdetaileddataonobservedincomes,expenses,andborrowing,and are not reported here.

Figure 8. Financial costs of major and/or unplanned events

WORKING PAPER

BetweenMarchandNovember2015, themost frequentadverseevent that respondents reportedwasillnesstohouseholdmemberswhilealsoreportingthattheyfrequentlyhostedguests,celebratedfestivalsandceremonies,andundertooksometravel.

Foreachoftheseevents,weestimatedboththemedianandupperboundoftotalexpensesincurredineachsurveyperiodthattheseeventswerereported.Forexample,wefindthatwhilerepairstopropertyanddeathofahouseholdmemberarerelativelyinfrequentevents,themedianexpensesincurredinbothcasesare much larger than most other events and these event-related expenses have the potential to escalate quiterapidly(ashighasRs.80,000inthecaseofonepanelhousehold).

Ofthemorefrequentlyreportedevents,spendingonfestivalsandceremoniestendedtobehigherthantherest,wherethemedianexpensesincurredduringanillnessorwhilehostingguestswasRs.500,onfestivalsRs.1,000andonceremoniesRs.700.Asbefore,weobservedthatexpensesonfestivalsandceremonieshadthe potential to escalate more rapidly and for more households than in the case of health-related events.

Illnessandinjurieswouldknockhouseholdmembersoutoftheworkforceforanywherebetweenonedayandfivedays(atthemedian)tothreeweeks(atthe90thpercentileinthecaseofinjuries).Travel,eithertovisitfriendsandrelatives,insearchofworktendedtotakeanywherebetweentwodaysatthemediantoeight days at the upper end.

Inthecaseofhouseholdsdependentonincomefromdailywagelaborordailyincomeself-employment,longertripsmightmeanareductioninincomeduringthisperiod.However,thebiggestshockintermsofopportunitycostsistheinabilitytofindincome-generatingemploymentitself—thoughthiswasreportedbyonlytwotothreesamplehouseholdsineachround,theopportunitycostsofthisshockwerehigherthanmostothers,evenatthemedian.

16

Figure 9. Opportunity costs of major and/or unplanned events


17

6.1. Access to credit in KrishnagiriKrishnagiridistrict,lyingatthejunctionofTamilNadu,Karnataka,andAndhraPradeshandbenefitingfromthestrongfinancialdevelopmentofallthreestates,hasavibrantfinanciallandscape.

6. Financial Sector Characteristics

Figure 10. Incidence of indebtedness

Figure 11. Average outstanding debt per indebted household

WORKING PAPER

18

The All India Debt and Investment survey is a decadal survey that results in rich and detailed insights on financialbehaviorofIndianhouseholdsatdistrictandsubdistrictlevels.Fromthedatareleasedinthelatestround(2013),wefindevidencethatTamilNaduoutperformsall-Indiaaveragesonbothmeasuresofaccesstoformalcredit.WithinTamilNadu,themedianamountofoutstandinginstitutionaldebtwas71percenthigher inKrishnagirithanthestateaverage,eventhoughtheincidenceof indebtednesswasalmostthesame,indicatingthatruralhouseholdssampledfromKrishnagiriaresignificantlymorelikelytohaveaccessto,andactivelyuse,institutionalcreditservices.

There is mixed evidence of the effects of formal sector financial development on informal borrowing.IncidenceofborrowingfrommoneylendersismuchlowerinKrishnagirithanthestateaverage,butforthosehouseholdsthatreportedoutstandingloans,theloanswere28percentlarger.Similarlyforloansoutstandingtofriendsandrelativeswithinsocialnetworks,theincidenceofindebtednessisonlymarginallylower,butoutstanding amounts are higher. Overall, indebted households in this population carried much higheroutstanding debt than comparable state or national averages.

TheCRISILInclusixreportconsolidatesmultiplemeasuresofformalfinancialdevelopmentincludingbranchpresence,useofformalsavingsaccounts,andcreditpenetrationintoacompositeindexatbothstateandsubstate/districtlevel.In2013,Krishnagiribrokeintothetop100districtsinIndia(manyofthesedistrictsincludeurban economic centers) while Tamil Nadu itself consistently ranked among the very best performing states. In2013thesecalculationsbegantoincludecreditdisbursedthroughMFIsbutevenwithMFIsexcluded,aconsistentupwardtrendisobservedinbank-ledfinancialaccessinKrishnagiri.

Specificallyonmicrofinance,veryfewsourcesreleaseaccessandusestatisticsforsubstateunitsandmostthatdo(suchasMIXMarket’sFinclusionLabinitiative)relyonself-reporteddatafromasubsetofproviders.Evenfurther,thegovernment’sownSelf-HelpGroup(SHG)bank-linkedmicrocreditprogrammaybeatleasttwiceaslargeascommercialmicrofinancebutremainsunreportedincreditbureausandsectorreports.Measurementchallengesnotwithstanding,wereviewpublicdataavailableonKrishnagiriandfindthatthedistrict is fairly well served through local institutions present in the district. Anecdotally and through primary interviews,weunderstoodthatresidentsarealsoservedbymobileagentsfrominstitutionsinneighboringdistricts—NorthKrishnagiriislessthananhour’sdrivefromBangalore,whileintheSouthseveralvillagesshareborderswithVellore.AsFigure12andFigure13show,bothneighborshave largefinancialmarkets thatwouldhavenaturallyextendedtoKrishnagiriovertime,easilyserviceablewithout investing inadditionallocalphysicalinfrastructure.Forthisreason,statisticsaggregatedonthebasisofbranches’physicallocationand not on the basis of the location of customers may underestimate the true penetration of microcredit in districts such as Krishnagiri.

Table 5. CRISIL Inclusix scoresa

Krishnagiri All Tamil Nadu Karnataka Andhra Pradesh

2013(includingMFIs) 66.7 79.2 74.4 69.2

2013(onlybank-led) 61.9 - - -

2012 57.2 64.8 61.4 64.8

2011 52.4 60.5 57.7 61.3

2010 50.3 57.6 54.7 57.6

a. SeeCRISILInclusiveathttps://www.crisil.com/about-crisil/crisil-inclusix.html.


19

Figure 12. Access points in and around Krishnagiri district (map showing four South-Indian states)

Figure 13. Gross loan portfolio of Krishnagiri and surrounding districts (map showing three South-Indian states)

Source:FinclusionLab

Source:MIXFinclusionLabreports,June2015

WORKING PAPER

20

6.2. Credit access and use among sample householdsTheloanportfoliosofsamplehouseholdsarereportedforeachsubsample—JLGborrowersandnon-JLGhouseholds.TheJLGgroupcomprisesthosehouseholdsthatself-reportedtobeactiveusersofanewJLGloan(definedbyhavingusedanewJLGloanduringthemonthofenlistment).Thenon-JLGgroupcomprisesthosehouseholdsthatself-reportedtobenonusersofaJLGloanduringthemonthofenlistment.Throughthestudy’sobservation,only6percenthouseholdsinthenon-JLGsamplewentontoreportreceivinganewloanfromanMFI.

In concurrence with overall district-level access to credit statistics cited earlier in this chapter, samplehouseholds, too, reportedhigh incidenceof institutionaldebt, inboth samples. Loans fromalldomesticbankswere frequently reported in both subsamples, aswere loans received through SHGs. In terms ofinformalborrowing,loansfromkinwererelativelylowerinincidencethanloansfromfriendsorothernonkinsocial connections.

Table 6. Sources of credit for JLG and Non-JLG households, by lender

JLG sample(% households)

Non-JLG sample(% households)

Banks

Public sector banks 67 65

Regional rural banks 16 10

Co-operative banks 23 33

Private banks 11 15

Foreignbanks 4 5

OtherFormalInstitutions

MFIs 95 6

SHGs 37 25

Nonbankingfinancialinstitutions 6 2

Chit funds 1 1

Other formal 5 4

Relatives*

Elder sister 7 3

Elder brother 4 3

Wife’s elder brother 1 7

Other kin relationship 7 10

Friends&Nonkin*

Friend 32 46

Employer 4 2

Neighbor 4 9

Other nonkin relationship 46 37

* only most prominent relationships listed


21

JLGand non-JLG subsamples seemevenlymatched in terms of overall debt, though stark differencesemerged around certain types of unique institutions. We are yet unable to explain the reasons for some of thesedifferences;especiallyinthecaseofcertaintypesofbanksorotherfinancialinstitutions,incidenceofindebtednessmightbecorrelatedwithphysicalaccessoraccessthroughanMFIorasocialconnection.However,thisisahypothesisweareunabletofurtherexplore.

Acrossthesample,householdsreportedhavingtakenthelargestloansfrombanks(Rs.50,000),followedbySHGs(Rs.30,000)andco-operatives(Rs.28,000).

Thesurveyalsoaskedaboutanyprocessingortransactioncoststhathouseholdsincurredtoreceivetheloan,typically includingmembership fees, transportationcosts,costofprocuringdocuments,processingfees,etc.ForlargeportionsofloansfrombanksandMFIs,41percentand55percent,respectively,respondenthouseholdsreportedhavingincurredacostofmorethanRs.50,andaveragecostforthesesourcestendsto be quite high.

Formostloans,notablyincludingthosefromkinandnonkin,householdsreportedthatinterestwaschargedandamajority of borrowerswerealsoable to recall andclearly specify the typeand level of interestcharged.

Table 7. Loan amount and cost of borrowing

Kin Nonkin Bank Co-op MFI SHG Other

Loan principal Median (in Rs.) 20,000 10,000 50,000 28,000 20,000 30,000 10,000

Transaction costs

Incurred costs >Rs.50 % Loans 6 5 41 24 55 18 20

Average costs Median (in Rs.) 175 200 100 100 106 50 75

InterestInterest is charged % Loans 58 59 100 99 99 100 88

Borrowerabletorecall interest rate

% Loans with interest 89 96 92 93 85 94 89

Table 8. Collateral requirements


Collateral required % Loans 2 5 53 68 1 0 54

Type of collateral given

Immovable property % Collateralized loans 60 29 7 7 - - 1

Movable property % Collateralized loans - 4 3 11 - - 13

Gold,andjewelry % Collateralized loans - 21 57 52 33 - 39

Shares % Collateralized loans - 21 25 22 67 - 33

Others,incl.documents % Collateralized loans 53 25 6 9 - - 12

WORKING PAPER

22

Morethanhalfofloanstakenfromformalinstitutionsrequiredsomeformofcollateral,withtheexceptionofMFIsandSHGs.Ofthesecollateralizedloans,mostloansweretakenagainstgold,ornaments,andjewelryor movable property (including livestock). The high prevalence of gold as collateral required might be a tellingindicatorforthepopularityofgoldloanproducts,rendereddesirablebytheirquickprocessingtimeand low interest rates.

Across thesample,most formal loansare reportedtospecifyfixed repaymentperiods in theircontractsandwherethispercentageislowerforsomeloansfrombanksandco-operatives,thisisperhapssomesignoftherespondents’lowawarenessofloanterms.Ontheotherhand,lessthanhalfofloansfrominformalsourcesarereportedtospecifyfixedrepaymentperiods.

Intermsofloantenure,informalloansarereportedtohavemanymoreshorttenureloans(1–6months)thanformal loans.

Mostloansfrombanks(63percent),co-operatives(83percent),andMFIs(75%)haveloantenuresupto12months.It isnoteworthythatasignificantproportionofloansfromMFIs(22percent)werereportedtohavetenuresbetween13and24months,asoneindicatorofinnovationfromtheclassicjoint-liabilityloanproduct offered for only 52 weeks.

Table 9. Terms of repayment


Fixedrepaymentperiod % Loans 32 36 86 78 100 95 65

Loan tenure

1–6 months % Fixed-tenure loans 44 59 4 2 4 4 36



More than 24 months

% Fixed-tenure loans 0 7 5 6 2 5 2

Repayment frequencyAt regular intervals:

Weekly % Fixed-tenure loans 1 3 0 1 6 1 7

Monthly % Fixed-tenure loans 7 8 31 17 44 48 19

Others % Fixed-tenure loans 8 11 31 18 49 49 26

Irregularly % Fixed-tenure loans 74 65 16 23 1 1 32

Bulletrepayment % Fixed-tenure loans 10 13 22 41 0 0 16


23

Study respondents reported loanswithavarietyofdifferent repayment terms.Overall, loans taken fromfriends in timesofneed tendedtobeflexibleeither in loan tenureor the frequencyofpayments tobemade—orusuallyboth.Institutionalloansoftenrequiredfixedweeklyormonthlyrepaymentsalthoughwelearned of contract innovations that deviated from the typical EMI-type structure.

7.1. Debt affordability measures: Debt-to-income and debt-to-surplus ratiosFollowingfromtheliterature,wesoughttoestimatehouseholddebtburdenastherepaymentobligationforalloutstandingloansrelativetothehousehold’sincomesandexpenses.Forconvenience,weaggregatedrepaymentobligationsandcashflowsbymonth,evenwhererepaymentsaredueorincomesareearnedatahigherfrequency.Broadly,ourreviewofloanfeaturesindicatedthreetypesofrepaymentstructures:(i)fixedtenureandregularrepaymentschedules,(ii)fixedtenureand“bullet”repaymentschedules(whereboth interest and principal are paid once at the end of the contract or a moderate rate of interest alone is servicedthroughfrequentpaymentsbutprincipalispaiddownattheendofthecontract),and(iii)flexiblerepaymentscheduleswhereeithertenureorpaymentfrequencyorbothareleftunspecified.Purelyflexiblepayment schedules are largely limited to informal lending whereas bullet repayments are increasingly becomingmorepopularinformalbanking,throughthewidespreadsuccessofthejewelloanproduct.12

Foreachof the three typesof loans,weestimateand report simpledebt-to-income ratios (DIR)as thesum total of all estimated debt obligations for outstanding loans divided by average monthly household income. In doing so we rely on two accounting assumptions:

1. Wherethe loantenure isfixedbutrepaymentamountsorfrequenciesarenotspecified,suchas inbulletrepaymentloans,weassumedthattheburdenofrepaymenttobeevenlydistributedoverthetenure.

2. Whereloantenureisnotspecified,suchasinsometypesofinformalloans,weassignedareasonabletenure (based on loan amount)13 and thereby assumed that the burden or repayment to be evenly distributed over the tenure.

Bothassumptions,while limiting,areanecessaryprerequisite tostandardizeandcompare relativedebtburdens. Figure 14 presents comparable distributions of debt-to-income ratios that first exclude flexibleloans,thenconsiderthemseparatelyand,inthelastcolumn,aggregateflexiblewithfixedEMI-typeloans.Comparing the distributions, it is evident that considering repayments due toward fixed-schedule (andmostly institutional) debts greatly underestimates the overall household debt burden for the whole sample. Further,aspresented inFigure15,pooresthouseholdscarry thehighestestimateddebtburdenandyetagain,anyanalysisthatexcludesflexibleloanswouldsignificantlyunderestimatedebtburdenand,thereby,debt-related stress for the poorest and most vulnerable households.

7. Evidence of Debt-Related Financial Distress

12. Ajewelloanisaloanagainstcollateralandistypicallymarketedasprovidingshort-termliquidityreliefandwidelyoffered at both rural and urban bank branches. Several features explain the popularity of the product—many low-incomehouseholdsowngoldjewelery,interestratesaremuchlowerotherinstitutionalloans,andtheloanoftenrequiresonlyonelump-sumpaymentwithinayeartoredeemthejeweleryor,alternatively,asettlementoftheinterestto-datewhereintheloanis“rolledover,”orsimplyrenewed.

13. ForloansuptoRs.15,000weassumedatenureof12monthsandforlargerloans,24months.ThisisconsistentwiththepracticesofMFIs,asdirectedbyregulation.

WORKING PAPER

24

Figure 14. Density distribution of DIR, by type of repayment structure

Figure 15. Composition of DIRs by income quintile


25

However,itisherethatourassumptionsaroundhowflexibleloansmightberepaidbecomesalient—indeed,they may neither be repaid in the tenure assumed nor as frequently as assumed. A common feature of certaintypesofflexible,informalloansisalsothattheyareoftenrenegotiatedwiththelender.

Debt-to-surplus ratio

While DIRs are popular largely as personal finance rules of thumb, it is unclear whether they can beappropriately applied to underwrite loans for low-income households. Poor people in developing countriesfacesignificantincomevolatilityandlackaccesstoreliableformalsafetynets.Debtaffordabilityassessmentsmight therefore need to include a buffer to coverminimum living expenses, and preventborrowerimpoverishmentthroughover-indebtedness,andthiscouldtaketheformofadebt-to-surplusratio(DSR)whereahealthylevelofdebtischaracterizedbyaratiolessthan1.

Asbefore, thenumeratorofanoveralldebt-to-surplus ratio remains the sum totalofpaymentsdueonoutstandingloansineachmonth.Inthedenominator,weestimateaveragemonthlyincomesnetofminimumliving expenses for all permanent household members.14Figure16isascatterplotofdebtobligationsonthey-axisandsurplusonthex-axis,wherethequalityofdebt-to-surplusisplottedasthereferenceline.Householdsinthelowertriangleofthescatterplotare“intheclear”:theirincomesare,onaverage,adequatetomeetbothminimumhouseholdexpensesandloanrepaymentsdue.Asbefore,limitingtheanalysistoonlythepayments dueon fixed EMI-type loans could significantly underestimateoverall indebtedness. Even so,consideringonlyfixed,regularpaymentobligationsservestoidentifyhouseholdsintheuppertriangleoftheplot—somethatwereonthemarginwithonlyasmallincomedeficit,whileothersreportedamuchwidergap between disposable income available and estimated payments due.

14. Permanentmembersare thosewhohavecontinued to residewith thehousehold foramajorityof the study’sobservation and contributed to/shared in household resources. Minimum living expenses for a household iscomputedbasisthe10thpercentileofmonthlyper-capitaexpenditure(approximatelyRs.800perperson)forruralTamilNaduasreportedintheNSSOSurveyofHouseholdConsumerExpenditure,2012.

Figure 16.Scatterplot of monthly repayment obligations

with available household surplus

WORKING PAPER

26

Overall,however,forabout78percentofsamplehouseholds,overallmonthlyrepaymentobligationsonfixed-tenureloanswithregularrepaymentobligationswaslessthan80percentoftheirincomesurplus(i.e.,incomenetofminimumlivingexpenses).Thesehouseholds, it seemsex-ante,shouldhavebeenabletocomfortably manage to make their payments on these loans on time without facing extreme duress and over and above these payments might have also been able to spend above the minimum standard and perhaps,savetowardpayingdownsomeoftheotherloansoranyothergoal.AsnotedinTable10,evenhouseholdswithcomfortableDSRstatuscarrymoderatetohighlevelsofflexibledebt—topaydownthoseloans,borrowerswillneedtosavelumpsumsormobilizefundsfromothersourceswhenthosepaymentsbecome due.

Foryetanother7percentofhouseholds,DSRissomewherearound1,thatis,theirdebtserviceobligationsonfixedandregularloanswillmoreorlessdepletetheircompletehouseholdsurplus,leavinglittleexcess(orasmalldeficit)tomanageanyotherexpensesthatmayariseduringthecourseoftheseloans.Shouldthesehouseholdfaceahealthshockoranadverseevent,orshouldanylargepaymentsfalldueaboveandbeyondtheverybareminimumlivingexpenses(suchasfortravel,celebrations,orceremonies)orevenifoneoftheircreditorswithintheirsocialnetworksasktoberepaid,thesehouseholdsmightstruggletofindthe resources to make those payments.

Foralloftheotherhouseholds(15percentofthesample),DSRseemstoindicatethattheymightexperiencesignificantfinancialdistressinthecourseoftheloantenure,foraslongascurrentloansremainoutstandingand income sources remainunchanged.All of thesehouseholdsalsocarry high levels of flexibledebt,likely incurred to supplement inadequate incomes in times of stress. Notice that this group has the highest incidence15ofbulletrepaymentloans(themostcommonisthejewelloan)andanevenhigherincidenceofflexibleborrowing,typicallyfromfriendsandrelatives.

15. Incidence isdefinedas“presenceoruseat leastonceduringtheobservationperiod”(roughlysevenmonths).Amounts,however,arereportedasmonthlymedianstobecomparablewithmonthlyincomesandsurplus.

Table 10. Distribution of sample households by overall DSR

Number of households %

Debt-to-Income Ratio

Fixed tenure & regular payment Bullet repayment Flexible loans

% HH Median %HH Median % HH MedianDSR0–0.4 216 59 65 0.1 23 0.2 63 0.3

DSR0.4–0.8 69 19 100 0.4 30 0.2 72 0.2

DSR0.8–1.2 26 7 100 0.6 31 0.5 62 0.4

DSR 1.2–1.6 15 4 100 0.7 27 0.7 80 0.2

DSR 1.6–2 8 2 100 0.7 50 0.2 88 0.2

DSR >2 24 7 100 1.2 38 0.6 79 0.7

Incomedeficit 7 2 100 2.8 43 0.5 71 1.5

Total 365 100


27

Fora smallnumberofhouseholds,wewereunable tocomputedebt-to-surplus ratios; itappeared thatthesehouseholdswere inanoverall incomedeficitposition.Theydidhave incomesourcesbutnoneofthese was regular and this created an extremely volatile surplus that was overall negative. Earlier we had discussedincomeandsurplusvolatilityasamonthlyphenomenon.Here,weobservethatfor2percentofindebtedhouseholdsinthesample,muchofthisvolatilitywaspersistentforalargepartoftheyear,ifnotlonger.Thesehouseholdsalsoreportedfrequentlywithdrawingfromsavingsaccounts,reportedthehighestincidenceofnewborrowingduringthestudy’sobservation,andoneofthemevenreportedsellingtheirland (a follow-up qualitative interview will seek to understand if this was a distress sale).

7.2. Delinquency on outstanding loansDelinquencyratesinmicrofinanceareseldomreportedordiscussedinpublic,whereaslowdefaultrateshave been widely reported and celebrated within the sector. We therefore lack the means to benchmark or compare the rates of the self-reporteddelinquencyobserved in our sample. Further,we report self-reported delinquency that was captured during interviews at frequent intervals—it is possible that very short-termdelinquency(delayedpaymentsbutwithinsevendays,forexample)wherenotsystematicallyself-reported as delinquency in our survey.

Duringtheobservationwindow(roughlysevenmonths),7percentofhouseholdsreportedat leastoncethattheyweredelinquentonapaymenttowardafixed-tenureregular-repaymentloan,i.e.,thetypeofloanonwhichtheytypicallymademonthlypayments.Incontrast,otherindicatorsofrepaymentdifficultiesreturned much higher estimates.

Ourhypothesisisthatasdebtlevelsincreaserelativetohouseholdsurplus,theincidenceofdelinquencywould also increase, but even so, this would remain an underestimate of underlying financial distress.Evidence reviewedearlierchapter in this report suggests thatmanyborrowers,whenunder stress, raiseresources through other means to make their payments on time.

Figure 17. Some indicators of debt-related financial distress

WORKING PAPER

28

InTable11showsresultsrelevanttothefirstpartofthishypothesis.Theincidenceofdelinquency(atleastonce during the study’s observation) is generally higher for households that have debt obligations in excess ofabasicsurplus.DelinquencywasfairlyhighevenforhouseholdsintheDSR0.8–1.2group,butthisisnotunexpected. Surpluses were computed as net of income and a bare minimum living expense standard—several sample households were (fortunately) at a slightly higher level even at the baseline and may not be abletomakefurtherreductionstotheirconsumptionexpenditurewithoutsignificantlyalsoreducingfoodconsumptionandinvestmentsinhealthandhumancapital.Further,householdscomputationallylocatedat the margin were vulnerable to a wide variety of shocks during the study’s observation—any or all of whichcouldhavetemporarily reducedlaborsupplyor increasedexpenses,causingadditionalfinancialdistress through a temporary shrinking of resources.

Inaddressingthesecondpartofthehypothesis,weinvestigatethreecopingmechanisms:(i) liquidatingassetsforcash,(ii)borrowingtosmoothencashflows,and(iii)transactingwithpersonswithintheborrowers’social or community networks.

Table 11. Incidence of delinquency on outstanding loans, by DSR

On Secured borrowingOn Unsecured institutional

borrowingOn Noninstitutional

borrowingDSR0–0.4 5% 3% 18%

DSR0.4–0.8 1% 3% 22%

DSR0.8–1.2 15% 8% 19%

DSR 1.2–1.6 - 13% 33%

DSR 1.6–2 25% 13% -

DSR >2 21% 13% 25%

Incomedeficit 14% 14% 43%

7.3. Liquidating assetsThreetypesofassetreductionswererevealedinthedataanalysisand,whilemanyotherswereinvestigated(suchasthesaleofdurableassetsandfarmmachinery),perhapsthelackoffunctioningsecondarymarketsfor certain assets implied that even households that were moderately asset-rich were unable to leverage thisposition to smoothcash flows. Themost liquidassetswere savingsaccounts, livestock,and inwhatweexpectarerarecases,thesaleof land.Householdsusedsavingsaccountsnotonlytomakesavingsdepositsbutalsotoreceivewagesandwelfarebenefitsfromthegovernment.ManyhouseholdsalsousedtheirformalsavingsaccountsascollateralundertheSelf-HelpGroupBank-Linkagescheme,whereholdinga lump sum in the account acted as security against a group loan.

Figure18presents the resultson theuseof savings, Figure19on the saleof livestock,andFigure20ontheoccasionallandsale.Whiletheincidenceofsavingswithdrawalsandlivestocksalearehigh,bothofthese may in some part be related to livelihood activity and in entirety cannot be interpreted as coping mechanisms. For example, where weekly wages are received under the National Rural EmploymentGuaranteeSchemeoranyother subsidiesorbenefitsare received through theaccount,withdrawaloftheseamountsmustbeinterpretedwithcaution.Similarly,householdsengagingincertainkindsoflivestockfarmingwillbemorelikelythanotherstoselltheir livestockfortrade,andagain,thismustbeinterpretedwith caution. We propose to account for some of these livelihood factors in future iterations of this analysis.


29

Figure 18.Savings withdrawals as a coping

mechanism, by DSR group

Figure 19. Sale of livestock as a coping


Figure 20. Sale of land as a coping


WORKING PAPER

30

7.4. New borrowingIncidence of new borrowing through the study observation period is correlated with DSR groups: groups with higher DSR are also more likely to borrow and to borrow larger amounts. Where collateral is not a constraint,thistrendisborneoutclearly(seeFigure22).

Figure 21. New borrowing as a coping mechanism, by DSR group (Secured borrowing)

Figure 22. New borrowing as a coping mechanism, by DSR group (Unsecured borrowing)


31

However,sincecomputedDSRisanaverageofdebtobligationsduringthesametime-periodforwhichwe report incidence, thiscorrelationmaynotbe interpretedasacausaleffectof repayment stressonborrowing.Rather,wecautiously infer thathighmonthly repaymentpressurecoupledwithacorrelatedrepaymentpressureonflexibleloansandasystematicvulnerabilitytovolatilecashflowsandadverseshocksleaves many households seeking to borrow additionally as a means of enhancing income and as a coping mechanism.ThisissupportedbytheevidenceinTable12,whereamajorityofrespondenthouseholdswhoborrowed new loans from institutions during the study’s observation period reported that they were either borrowing to make repayments or to smoothe consumption.

Table 12. Purpose of new borrowing, by DSR group

DSR 0- 0.4

DSR 0.4- 0.8

DSR 0.8- 1.2

DSR 1.2- 1.6

DSR 1.6- 2

DSR >2

Income deficit

Other

Reported as % households in each group

Secured borrowing

To repay other debts 5 13 12 7 38 21 29 36

Forconsumption 6 7 15 13 25 21 53

To meet health expenses 4 3 8 14

To meet event expenses 2 4

Foragriculture 1 4 4

Forabusiness 7

Investmentsinlandorjewelry 2 1 7 8 9

Home improvement 4 2

Unsecured institutional borrowingTo repay other debts 6 33 23 33 38 54 43 7

Forconsumption 17 35 23 27 25 50 29 19

To meet health expenses 6 6 8 13 14 26

To meet event expenses 2 1 4 32

Foragriculture 3 3 4 7 4 16

Forabusiness 1 9 4 7 17

Investmentsinlandorjewelry 2 1 13 13 14

Home improvement 4 6 4 13

Non-institutional borrowingTo repay other debts 10 23 23 20 75 50 43 7

Forconsumption 22 26 42 33 63 58 71 19

To meet health expenses 14 14 31 20 25 33 29 26

To meet event expenses 3 3 4 4 32

Foragriculture 2 6 4 16

Forabusiness 2 1 12 4

Investmentsinlandorjewelry 1 7 8 14

Home improvement 13 13 15 7 13 43

WORKING PAPER

7.5. Noninstitutional borrowing and social networksAsinthecaseofinstitutionalborrowing,asimilartrendisobservedinnoninstitutionalborrowing,alargepartof which is borrowing from friends and relatives. The incidence of this type of borrowing is much higher than either types of institutional borrowing but the amounts tend to be much lower. While resources available throughsocialnetworksare limitedandevenwhentheyare interest-free,theyareattachedwithsocialcosts.However,theymaybesignificantlyeasierandquickertoaccessthaninstitutionalloans(earlierwereported transactions costs for access).

Interestingly,householdswithhigherDSRwerealsomorelikelytorequestandreceivenonloantransfersfromtheirsocialnetworks,andbothincidenceandaveragemonthlyresourcesavailablethroughthischannelrank-order with the magnitude of debt-surplus ratio.

Figure 23. New borrowing as a coping mechanism, by DSR group (Noninstitutional borrowing)

32


Figure 24. Social network transfers as a coping mechanism,

by DSR group

With a view to understanding the role of debt affordability assessments at the point of sale (POS),weretrospectively evaluated a subset of new loan disbursements reported—how many of these loans might havefailedaformaldebtaffordabilityassessment?Atotalof170newloanswasreportedinthefifthandsixthsurveyrounds,ofwhich60werefrominstitutiontypesgovernedbycustomerprotectionregulationsofsomeform.Foreachoftheborrowerswhoreportedtheseloans,wecomputedDSRatPOSastheratioscomputedbasedondataobserveduptothelatestsurveyroundatthattime;resultsarereportedinTable13.LessthanhalfoftheloansdisbursedclearedaretrospectiveDSRassessmentasspecifiedinthefollowing:

(Borrowers’ existing fixed and regular debt obligations + Additional debt obligations from the new loan) < (Average monthly income – Minimum living expenses for family)

Wefindthispatternconsistentinboththesurveyroundsthatsupportedretrospectivechecks.

8. Debt Affordability Assessment at Point of Sale

Table 13. Retrospective debt affordability assessments for loans disbursed, July-September 2015

Number %

New institutional loans disbursed 60

Loans cleared a relaxed DSR check

DSR<0.4 17 28

DSR0.4–0.8 7 12

DSR0.8–1.2 3 5

Loans did not clear DSR check

DSR 1.2–1.6 4 7

DSR 1.6–2 1 2

DSR > 2 6 10

IncomedeficitatPOS 22 37

33

WORKING PAPER

8.1. Implementing debt affordability assessments and its challengesToenablebetterdebtaffordabilityassessmentsatPOS, thequalityandaccuracyofcredit informationavailabletoprovidersneedtobesignificantlyenhanced.WhileeffortsareunderwaytointegrateformalbankingcreditbureauswiththedatafromnonbankingfinancialinstitutionsandaparallelefforttodigitizeandintegratedataontheperformanceofSHGloans,loansdisbursedthroughco-operativebanksremainsoutsidethepurviewofanycentralizedrepository.Thereisaneedtorevisitandstandardizecreditreportingrequirementsbyfunction,ratherthanbyinstitutionaltypes.

However,relyingonlyoninstitutionaldataonhouseholddebtcouldbeinadequate.Asourresultsshow,low-income households carry high amounts of debt outstanding to friends, relatives, employers, andmoneylendersandyetdonothaveanyinsightsonwhethertheseagreementsaretrulybinding,whethertheycanberenegotiated,andatwhatcosts.Theseinsightswouldalsocriticallyinformthedevelopmentoffair debt affordability assessments.

OnFSPs’readinesstoimplementdebtaffordabilityassessments,itseemsthatthereisahugedisparityinhowprovidersconductduediligence,whatdatatheycollectoncustomers,andhowmuchofthisinformationfeedsintomeaningfulloanunderwritingcriteria.Often,incrementsinloansizeareautomatedbyrepaymenthistory (intended as a dynamic incentive) without much regard for borrowers’ own repayment capacity. We believethatthereisscopetoharvestbestpracticesfromwiththesector,bothinIndiaandinternationally,and to learn from previous experience how these assessments should be best designed and communicated toloanofficers.

9.1. Toward a suitability-based customer protection regimeDevelopingguidelinesforunsuitabilityisinthespiritoftheHippocraticOath,toensurethatfinancialservices“first,donoharm.”UpholdingthisprincipleisparticularlysalientinIndianlow-accessenvironmentswherehouseholdsarevulnerabletoplacingthemselvesinfinancialdistressowingtovastasymmetriesofinformationandpowerwithfinancialservicesproviders.16

9. Implications for Policy Design and Financial Services Provision

34

For the loans thatdidnotclear theDSR threshold, two scenariosemerged. First,where the servicingofalready-existingfixedandregular repaymentsalonewasconsumingmorethantheborrower’saverageincome surplus (suggesting that the borrower would either be delinquent on loan payments or meet expensesthroughnonoccupationalsourcesoffunds,orboth).Second,whereservicingexistingfixedandregularpaymentloansdidleavesomeroomforadditionaldebt,butperhapsnotasmuchaswasreceivedresultingintheadditionofthenewloanpredictinganincomedeficitinthenearfuture.

Severalconcernsarisehere.First,thisassessmentcountsonlyloansthatspecifiedfixedtenureandrequiredregular,frequentpayments—mostoftheseareinstitutionalloanswhereasamajorityofloansfromfriendsandrelatives,orevenjewelloanswouldbeexcludedfromsuchadebtaffordabilityassessment.Ifwewereto add all outstanding loans to this assessment and apply the same assumptions as before (regarding how flexible loansmightbepaiddown),even fewerborrowerswouldhavebeendeemedfit to receive theloans that theydid.Second,manyborrowers seemedto reportan incomedeficitorwerepredictedtoexperienceadeficitasaresultofnewdebtobligationsintheobservationwindow.Specificchallengesinmeasuring and implementing debt affordability assessments in India are also discussed as follows.

16. Sahasranamanetal.(2014)providesamuchmoredetaileddiscussion.


Ratherthanplacingtheresponsibilityoncustomerstopreventadverseoutcomesoffinancialservices,thereisamarkedroleforprovidersoffinancialservicestobemuchmoreresponsiblefortheproductsandservicesthat they recommendbasedona systematicunderstandingof thecustomer’sfinancial situation.Whileproviderscannotberesponsibleforultimatecustomeroutcomes,adheringtopreventiveprocesses,suchas implementingabetterdebtaffordabilityassessmentatPOS,willmarkasignificantstep inenhancingsuch provider responsibility.

TheIndiannationalimperativetorapidlyexpandthereachoffinancialservicestounderservedpopulationshas recently been accompanied by a remarkable evolution of its customer protection regime from one thatwascharacterizedby increasingproductdisclosureandclienteducation (generally,according tothe principle of caveat emptor or “buyer beware”) to one that is increasingly placing legal and regulatory responsibilityonproviders (generally,caveat venditor or “seller beware”) to steer customers away from harmful financial outcomes.17 This shift is consistent with new regulatory approaches in several other jurisdictions,mostnotablyAustralia,SouthAfrica,theUnitedStates,andtheUnitedKingdom,toensurethatfinancialservicesprovidersareactinginthebestinterestoftheircustomers.

Thedraft IndianFinancialCode2015,which iscurrentlyawaitingParliamentaryapproval,establishedalegallybindingrightforretailfinancialconsumerstoreceivesuitableadviceinrelationtothepurchaseofafinancialproductorservice,andthattheprovidermustcollectallrelevantinformationontheneedsandsituation of the consumer in making its recommendation.18TheReserveBankofIndiainitslatestCharterofCustomer Rights also includes a Right to Suitability such that the products offered should be appropriate totheneedsofthecustomerandbasedonanassessmentofthecustomer’sfinancialcircumstancesandunderstanding.19

Both these norms, legal and regulatory, rest on amandate for every retail financial services provider,regardlessofitstypeorcustomerbase,toinstitutionalizeaBoard-approvedsuitabilitypolicythatoutlinesprocesses itcanbeheldaccountableto, ratherthanactualoutcomesforcustomers.Suchapolicywillideallyincludeclearguidelinesonallend-to-endaspectsofafinancialservicesprovider’sinteractionwithacustomer,includingcollectionofinformation,productrecommendationandsale,customerunderstandingandinformedconsent,productpost-saleservicing,incentivesforcustomer-facingstaff,grievanceredressal,andaudit,monitoring,andcompliance.20

Our model of debt affordability assessment to prevent unsuitable credit provision will seek to inform normativepracticesoffinancialservicesprovidersonaspectsofproductrecommendationandsaleaswellascollectionofinformation.Ultimately,ourmodelaswellasourevidenceonborrowerindebtednesswillseek to serve as a building block for developing regulatory guidelines on suitable credit provision as well as developingsimilarguidelinesforsuitableprovisionofotherproductclassessuchasinsurance,investments,and pension.

9.2. Toward better delinquency management and recourseOur evidence shows that low-income households are vulnerable to several shocks on an ongoing basis and are subject to significant income volatility, which could greatly affect their ability tomake timelyrepayments. Even beyond reasonable debt affordability assessments at POS, timely monitoring and

17. The Report of the RBI Committee on Comprehensive Financial Services for Small Businesses and Low IncomeHouseholds2014includesadetaileddiscussionofthesetwoapproachesinitsChapteronCustomerProtection.

18. SeeDraftIndianFinancialCodeathttp://finmin.nic.in/fslrc/fslrc_index.asp.19. SeeRBICharterofCustomerRightsathttps://www.rbi.org.in/scripts/BS_PressReleaseDisplay.aspx?prid=32667.20. The Report of the RBI Committee on Comprehensive Financial Services for Small Businesses and Low Income

Households2014includesamodelSuitabilitypolicyonpage182.

35

WORKING PAPER

appropriate handling of incidents of delinquency will always remain a critical component of responsible lending,andonethatcansignificantlymitigateloan-relateddistressforborrowers.Asystematicassessmentof delinquency management practices in microlending could be insightful in this regard.

The advent of a new bankruptcy framework in India (pending Parliamentary approval) will allow eligible borrowerstofileforpersonalbankruptcyandavaila“freshstart”thatwouldentaildischargingborrowersfromqualifyingdebtsandallowforanopportunitytostartfinanciallyafresh.21 This is an important development inthelightofourfindingsthatnewborrowingisago-tocopingmechanisminenvironmentswherethereisreadyaccesstocredit,andwhilethisaccessisanimportanttoolforconsumptionsmoothing,thereisariskthat this could lead to entrapment in debt over the medium term.

Inaddition,financialservicesprovidersneedtohaveinplacebetterprovisionsformanagingunforeseenborrower delinquency. These could include practices such as providing top-up loans and contract renegotiationtoeasetemporaryfinancialdistress.Timelydebtcounsellingandfrequentmonitoringcouldalso help borrowers manage their debts over the loan tenure and may prevent further escalating distress.

9.3. Implementing debt affordability assessments and its challengesThere has been an unequal trend in the expansion of financial services to low-income households inIndia,withaccesstocredithavingrisenrapidly,whileaccesstoreliablesavings,investment,pension,andadequate insurance instruments remain wanting. Our results suggest that low-income borrowers may be substituting the rolesofother financial instruments inmitigating riskandproviding liquiditywithcredit,afinancialstrategythatisoftentimesexpensiveandsuboptimal.Commensurateandadequateaccesstoamorecomprehensivesuiteoffinancialserviceswillallowlow-incomehouseholdswithlimitedresourcestobettermeettheirfinancialgoalsofbuildinglumpsumsandcopingwithadverseeventsbyusingmoreoptimalfinancialstrategies.

36

DatacollectionwillcontinueintoMay2016,extendingthislongitudinaldatasetbyfivemonthsandallowingthe observation of panel households for more than a year. Data updates for all the completed analysis willservenotonlytovalidateinferences,butalsotoremoveanysystematicseasonaleffectsonthedata.Upcominganalysiswillalsoexploreinmoredetailtheroleofborrowingasacopingmechanismforvariousadverse events that low-income households face.

This study generates both data and research questions for further exploration. This study makes a unique contributionbypresentingearlyinsightsonthevarioustrajectoriesthatindebtedhouseholdsmighttake,astheyactivelyusenewfinancialservices.Throughthelensofdebtstress,westudythecopingmechanismsthatlow-incomehouseholdsadoptandlearnthatreadyaccesstofinancehasthepowertobothmitigateandexacerbatefinancialdistressoverthemediumtermforborrowers.Similartotheevidenceonmicrocredit’simpactsonfinancialwell-being,wealsoseeearlytrendsofheterogeneouseffectsoffinanceonborrowers’resilience.Furtherrigorousandinsightful researchontheseeffectswouldbebothtimelyandrelevanttoinformmicrofinancepracticeandcustomerprotectionregulation.

10. Directions for Future Research

21. SeeDraftInsolvencyandBankruptcyBill2015athttp://finmin.nic.in/reports/DraftInsolvencyBankruptcyBil2015.pdf.


37

Angelucci, M., D. Karlan, and J. Zinman. 2013. Win Some Lose Some? Evidence from a Randomized Microcredit Program Placement Experiment by Compartamos Banco. Tech. rep., National Bureau ofEconomic Research.

Banerjee, A., E. Breza, E. Duflo, and C. Kinnan. 2014. Does Microfinance Foster Business Growth? TheImportance of Entrepreneurial Heterogeneity.

CGAP.2009.Measuring Access to Financial Services around the World.CGAP.

Collins,D.,J.Morduch,S.Rutherford,andO.Ruthven.2009.Portfolios of the Poor: How the World’s Poor Live on $2 a Day.PrincetonUniversityPress.

Dattasharma,A.,R.Kamath,andS.Ramanathan.2016.“TheBurdenofMicrofinanceDebt:LessonsfromtheRamanagaramFinancialDiaries.”Development and Change,47(1),130–56.

Field, E., R. Pande, J. Papp, and Y. J. Park. 2012. Repayment FlexibilityCan Reduce Financial Stress: ARandomizedControlTrialwithMicrofinanceClientsinIndia.PloS one,7(9),e45679.

Grammling, M. 2009. Cross-Borrowing and Over-Indebtedness in Ghana: Empirical Evidence fromMicrofinanceClienteleandSmallEnterprises.Technical draft for discussion.ProCreditHoldingFrankfurtamMain.

Guerin, I., and M. Roesch. 2011. In Venkatasubramanian and S. Kumar, The Social Meaning of Overindebtedness and Creditworthiness in the Context of Poor Rural South India Households (Tamil Nadu). Tech.rep.,RUMEWorkingPaper.

Karlan,D.,R.Knight,andC.Udry.2012.Hoping to Win, Expected to Lose: Theory and Lessons on Micro Enterprise Development. Tech.rep.,NationalBureauofEconomicResearch.

MIX. 2015.Voice of the Client: An Analysis of Client Satisfaction and Consumer Protection across Four Microfinance Institutions in India.Tech.rep.,MIX.

Morduch,J.1995.“Incomesmoothingandconsumptionsmoothing.”The Journal of Economic Perspectives, 9(3),103–14.

Reserve Bank of India. 2014. Report of the Committee on Comprehensive Financial Services for Smalll Businesses and Low Income Households.

Sahasranaman,A.,D.George,D.Rajendran,andV.Prasad.2014.A New Framework for Financial Customer Protection in India.

Schicks, J. 2013. “The Sacrifices of Micro-Borrowers in Ghana—A Customer-Protection Perspective onMeasuring Over-Indebtedness.” The Journal of Development Studies, 49(9),1238–55.

Schicks, J.,andR.Rosenberg. 2011. “TooMuchMicrocredit?A Surveyof IssuesandEvidenceonOver-IndebtednessamongMicro-Borrowers.”CGAPOccasionalPaper.

Sriram,M.S.2015. Inclusive Finance India Report. ACCESS.

Venkata,N.,V.Yamini,andS.K.Krishna.2010. Diagnosing Financial Stress in Group Methodology. Tech. rep.

WorldBank.2012.Good Practices for Financial Consumer Protection.

Zollmann,J.2015.Kenya financial diaries. Tech. rep.

References

WORKING PAPER

38

Annexure 1. Data CollectionPrimary data were collected through in-person interviews with 411 households (in the initial sample). Householdswere typical revisitedwithin 30–50daysand follow-up surveysonchangesoccurring in thehouseholds since the last visit were administered.

Sampling

ThisresearchstudyislocatedinKrishnagiridistrictinTamilNadu,SouthIndia.ThedistrictofKrishnagiriwasformedasrecentlyas2004(whenitwascarvedoutfromthelargerDharmapuriadministrativedivision)andhas the unique distinction of sharing borders with two other Indian states—Andhra Pradesh and Karnataka. TogetherwithTamilNadu,thesestatesareknowntohaveavibrantfinanciallandscapeandstrongpresenceoffinancialinstitutions,includingmicrofinance.

Toselectthehouseholdsinthesample,wefollowedamultistagerandomsamplingprotocol:

1. First,weconductedlistingsurveysin42villagesinKrishnagiridistrict,selectedinarandomorderfromtheCensus2011villagelist.ThelistingsurveywasadministeredtokeyinformantsintheentirevillageandgatheredinformationonallcurrentlyactiveJLGborrowinggroups.Ofthese,28villageshadatleastoneJLG,andin18villageswefoundatleastoneJLGthatreceivedtheirdisbursementwithinthelastmonth.SincewewerekeentostudyJLGborrowersthroughthetenureoftheirloans,weselectedvillages and groups such that borrowers would be at the beginning of a loan cycle.

2. Ineachof the18villages,wesublistedonly theJLGsthat receiveda loandisbursement in the lastmonthandrandomlyselectedbetweenoneandfivegroups22 per village. All member households of each selected group were surveyed.

3. Ineachvillage,anequivalentnumberofJLGnonparticipanthouseholds23 were also selected using a systematicrandomsamplingtechnique,appliedtotheentirepopulationofthevillage.Forexample,inavillagewherewesampledtwoJLGsoffivememberseach,10JLGnonparticipanthouseholdswere also sampled.

4. Inall,wesampledandsuccessfullycompletedbaselinehouseholdsurveys for200households thatweremembersofJLGgroupsinanewloancycleand211JLGnonparticipanthouseholds.

Survey design

Intotal,foursurveyinstrumentsweredesignedandadministered.First,aquicklistingsurveywasadministeredbetweenJanuaryandFebruary2015throughwhichweunderstoodthegeographyofeachvillage,anychangesinpopulationsincethe2011census,andthepresenceoffinancialinstitutionsincludingmicrofinanceprovidersandjoint-liabilitygroups.

Followingthesamplingprotocols,thefirstinterviewwithsamplehouseholdsfollowedtheBaselineHouseholdQuestionnaire.Subsequenttothebaselineround,itwasplannedthatsamplehouseholdswouldberevisited

22. ThenumberofgroupsselectedintothesampleineachvillagewasnormalizedbybothvillagepopulationandJLGloanaccess,allowingustoover-samplefromlargervillagesandfromvillageswheremultipleJLGproviderswerepresent.

23. AJLGnonparticipanthouseholdwasidentifiedasonethathadeitherneverborrowedthroughaJLGbeforeand/oronethatdidnothaveaJLGloan(orapplicationforaloan)outstandingatthetimeoflisting.NonparticipanthouseholdsthatsubsequentlyborrowedthroughJLGsduringthestudyperiodwereretainedinthesamplewithnoconsequentchangestothesurveyorobservation.Uptothefirstsevenmonthlysurveyrounds,only6oercebtofinthenon-JLGsamplereceivedanewJLGloan.


39

every month24andamorestreamlinedFollow-upQuestionnairewouldbeadministered.Thestreamlinedquestionnairewasedited to remove redundancies in thedatacollectedonhouseholdcharacteristics,whilefocusinginsteadonrecentevents,dynamiccashflows(incomesandexpenses),andanychangesinhouseholds’ socioeconomic situation since the previous visit.

Finally,arichcommunity inventoryofbasicfacilities inthevillage, livelihoodopportunities,andactivitiesofformalandinformalfinancierswillbeadministeredtokeyinformants(e.g.,Panchayatleaders,VillageAdministrativeOfficebearers,schoolheadmasters,SHGleaders,villageelders,etc.).Thevillagesurveywillalsocollectrichdataonvillage-levelcharacteristics,villageassociations,socialgroups,andrecentadverseevents.Thefirstpartofthissurveywasadministeredalongwiththelistingsurveyin2015,whiletheremainingmoduleswillbeadministeredinMarch2016.

24. Inpractice, revisit interviewsweretypicallyconductedwithin30–50daysof thepreviousvisit.Field investigatorsregularlyfollowedafixedrouteandvillageorderforsurveyadministrationbutoccasionalchangesweremadeto accommodate respondents’ availability or unavoidable delays between survey rounds. If a respondent was unavailable for the interview (oftendue to temporarymigration forwork, social events outside their villageortheincidenceofadverseeventsintheirfamily),theinvestigatorscoordinatedwithrespondentstocompletetheinterview later at their convenience.

WORKING PAPER

40

Annexure 2. Additional Sample Description

Figure 25. Distribution of household size

Table 14. Characteristics of the head of household*

Characteristics Study sampleGender

Male 83

Female 16

Age20–39years 33

40–60years 51

60–90years 16

EducationNo schooling 32

Uptoprimaryschool 11

Primary completed 34

Higher secondary completed 23

Characteristics Study sampleReligion

Hindu 89

Muslim 8

Christian 3

Others 0

Social categoryScheduled caste 27

Scheduled tribe 2

Backwardclass 39

Most backward class 29

Others 3

*Thesurveydidnotformallydefinetheroleofheadofhousehold,butinvestigatorsweretrainedtoprobeforthepersoninchargeofhouseholdmanagementanddecision-making.Asaresult,wealsoobservedwomenself-identifyingasthekeydecision-makerofthehouseholdinsteadoftheirhusbands,whichweperhapsmaynothaveobservedhadweusedotherstandarddefinitions.


41

Table 16. Access to Insurance

% households

Enrolled in government health insurance scheme 69

Purchased any other insurance policy 22

By type of insurance cover

Life insurance 22

Health insurance 90

Personal accident insurance 5

Asset insurance 17

Table 15. Access to basic facilities

Taluk 1 Taluk 2

Number of rooms Median 3 2

Area of dwelling (in sq. ft.) Mean 861 872

Rights over current residence

Owner with title %HH 60 66

Owner without title %HH 12 20

Owner with house or site received from government %HH 3 5

Tenant %HH 23 5

Other %HH 2 3

Toilet facilityFlushtoilet—Sharedwithotherfamilies %HH 8 3

Flushtoilet—Exclusiveaccess %HH 52 22

No toilet %HH 38 74

Drinking water sourcePiped to dwelling %HH 26 6

Piped to yard or public pipe %HH 57 81

Lighting sourcePrimary lighting fuel—Electricity %HH 99 97

Primary cooking fuelLPG %HH 70 47

Wood %HH 24 49

* Including scheduled commercial banks (nationalised andprivate), co-operative banks, and regional ruralbanks.

Table 17. Use of Savings

Product usage

Current savings balance

% HHs with at least 1 account

Median (Rs.)

Self-help groups 46 10,000

Banks* 33 1,500

Postofficedepositaccounts 13 1,000

NBFCsincludingchitfunds 10 20,000

Cash-in-hand 29 1,000

WORKING PAPER

42

Annexure 3. Socioeconomic Indicators of Study LocationThenationalCensus2011providesahostof indicators that revealuseful socioeconomiccharacteristics.Undereachofthecharacteristics,KrishnagiridistrictiscomparedtoruralTamilNadustateandruralIndia.

Figure 26. Map of Krishnagiri district from Census 2011

Table 18. Income profile

Monthly income bracket of highest earning member (Rs.) India Tamil Nadu state Krishnagiri district

<5000 75% 78% 77%

>=5000and<=10000 17% 16% 17%

>10000 8% 6% 6%

>5000withgovernmentjob 5% 4% 5%


Table 19. Occupation profile

Income source India Tamil Nadu state Krishnagiri districtCultivation 30% 18% 27%

Manual casual labor 51% 66% 56%

Part-time or full-time domestic service 3% 2% 4%

Nonagricultural own account enterprise 2% 4% 4%

Others 14% 11% 10%

Payingincometax/professionaltax 5% 4% 4%

Landlesshouseholdsderivingmajorpartoftheirincome from manual casual labor 38% 56% 40%

Table 20. Vehicle ownership profile

Vehicle type India Tamil Nadu state Krishnagiri districtBicycle 46% 46% 35%

Scooter/motorcycle/moped 14% 26% 29%

3-wheeler 1% 0% 0%

4-wheeler 2% 2% 1%

Table 21. Consumer durable ownership profile

Consumer durable type India Tamil Nadu state Krishnagiri districtRadio/Transistor 17% 19% 9%

Television 33% 85% 84%

Computer/laptopwithinternet 1% 1% 1%

Computer/laptopwithoutinternet 4% 4% 4%

Mobile only 68% 78% 83%

Without any phone 28% 18% 15%

Refrigerator 11% 12% 10%

TV,computer/laptop,telephone/mobilephoneandscooter/car 1% 2% 2%

43

WORKING PAPER

Table 22. Property ownership

Ownership type India Tamil Nadu state Krishnagiri districtOwn land 67% 42% 73%

Own cultivable land (irrigated and unirrigated) 55% 31% 56%

Own house 95% 91% 92%

Rented house 3% 8% 7%

44

Table 23. Dwelling characteristics

Dwelling characteristic India Tamil Nadu state Krishnagiri districtPermanent structure 51% 60% 57%

Semi-permanent structure 31% 19% 30%

Goodcondition 44% 63% 64%

Livable condition 46% 32% 32%

Dilapidated condition 6% 2% 2%

One room 39% 42% 41%

Two rooms 32% 29% 32%

Three rooms 13% 11% 12%

Table 24. Access to facilities

Facility type India Tamil Nadu state Krishnagiri districtTap water from treated source 18% 46% 42%

Tap water from untreated source 13% 33% 33%

Lighting from electricity 55% 91% 89%

Lighting from kerosene 43% 8% 10%

Cookingusingfirewood 63% 67% 74%

CookingusingLPG/PNG 11% 29% 21%

Latrine facility within the premises 31% 23% 19%

Public latrine 2% 4% 2%

Open 67% 73% 79%

Advancing financial inclusion to improve the lives of the poor

1825IStreetNW,7thFloorWashington,DC20006

www.cgap.org

Correspondence: Lisa Stahl ( [email protected] )CoverPhoto:SudiptoRana,2014CGAPPhotoContest

Date post:	16-Oct-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

An approach using primary evidence from low-income ... · Vaishnavi Prathap and Rachit Khaitan...

Documents