Experiments in Islamic microfinance
Post on 14-Dec-2016
Journal of Economic Behavior & Organization 95 (2013) 252 269
Contents lists available at ScienceDirect
Journal of Economic Behavior & Organization
j ourna l ho me pag e: www.elsev ier .com/ locate / jebo
Experiments in Islamic micronance
Mohamed El-Komi , Rachel CrosonUniversity of Texas at Dallas, 800 W. Campbell Road, Richardson, TX 75080-3021, United States
a r t i c
Article history:Received 12 AReceived in reAccepted 14 AAvailable onlin
Keywords:Islamic nancExperimental MicronanceInformation asInterestProt sharingJoint venture
This papIslamic (maendeavor isU.S. and abr
MicronBarua, 2006the reach othat over 1
We thankthe conceptuaAdam Pierce fthe participanEconomic Assoand Finance in
CorresponTel.: +44 0191
0167-2681/$ http://dx.doi.ol e i n f o
ugust 2010vised form 13 June 2012ugust 2012e 7 September 2012
a b s t r a c t
Micronance has been identied as an important tool in increasing the productivity ofthe poor and in aiding economic development. However, a large proportion of the poorare practicing Muslims, and are thus unable to take advantage of traditional micronancecontracts which involve the payment of interest. This paper describes and experimentallytests Islamic-compliant micronance products in the context of information asymmetryand costly state verication. We nd signicantly higher compliance rates for the Islamic-compliant contracts (prot-sharing and joint venture) than for the traditional contract(interest-based). We believe that there is great promise for these types of loans in themicronance context, for both Muslims and non-Muslims.
2012 Published by Elsevier B.V.
er describes and experimentally tests Islamic-compliant micronance products. While both micronance andcro)nance are well-developed elds of study, very little previous research has examined their intersection. This
important for understanding the nancial attitudes and behaviors of low-income consumers, both within theoad.ance has been identied by many as a useful tool in aiding economic development (Comim, 2007; Dowla and; Wright, 2000; Islam, 2007), even though the actual impact on poverty is still a matter of debate. However,f micronance (and especially microcredit) may be more limited than was previously realized. It is estimated/3 of the worlds poor are Muslims (CIA World Factbook, 2010 and Economist, 2008), and thus are unable to
Ali Abdul-Rahman, Ahmed Al-Zahrani, Daniel Arce, Catherine Eckel, Mahmoud El-Gamal, Uri Gneezy, and Dean Karlan for their help inlization and design of this paper. Special thanks to the CBEES lab staff; Sheheryar Banuri, Tara Larson, Wendy Lee, Elizabeth Pickett andor outstanding support. We particularly thank three anonymous reviewers for their invaluable comments and suggestions. We also thankts at the Southern Economic Association Conference in San Antonio, the Economic Science Association Conference in Chicago, the Europeanciation Conference in Glasgow, the African Econometrics Association Conference in Cairo, the International Conference on Islamic Economics
Qatar and the Symposium on Analytical and Empirical Research in Islamic Finance at Aston Business School for helpful comments.
ding author. Present address: Department of Economics, the American University in Cairo, AUC Avenue,P.O. Box 74, New Cairo 11835, Egypt. 334 5433; fax: +44 0191 334 5201.resses: email@example.com (M. El-Komi), firstname.lastname@example.org (R. Croson).
see front matter 2012 Published by Elsevier B.V.rg/10.1016/j.jebo.2012.08.009
M. El-Komi, R. Croson / Journal of Economic Behavior & Organization 95 (2013) 252 269 253
take advantage of the existing, interest based, microcredit products (CGAP, 2008). The Consultative Group to Assist thePoor (CGAP) conducted a global survey in 19 Muslim countries in 2007, in which 2040 percent of the respondents citedreligious reasons for not using conventional microloans. Constructing and making available Islamic-compliant micronanceproducts would extend the reach of micronance and aid in the economic progress of the Muslim poor in all countries, andthe economic development of nations with large Muslim populations.
Karlan and Morduch (2009) stress the importance of the way micronance choices are offered, and argue that they maymatter more than the pure economics of the choices with regards to take-up rates. The products tested in this paper rely onprot and loss sharing (PLS) contracts, and show how these options can contribute to the variety of micronance choices,which should result in higher take-up rates of micronance products. Our results indicate that these products can be moreefcient (annon-Muslim
The useasymmetrythe problemand costly sthese PLS an
PLS cont(such as: leaand Liu (20most of thebehind the PLS is inher
That saidern Nigeria of random p
Our goalthat PLS cofranchise agthe intereststronger ob
In this pexperimentbehavioral the eld. Th
We devethem with not to the lthree contrthe borrowprojects ouborrower if
Our resuall three concontracts din the micro
The papeSection 3 deand Section
Islamic industry. Its
1 This is by nchoice of proje
2 We thank d more protable for lenders) than traditional interest based contracts which are primarily offered, even among populations.
of PLS contracts has traditionally been minimized in Islamic banking due to risks resulting from information (such as enterprise failure, adverse selection, moral hazard and costly state verication) (El-Gamal, 1997). Whiles of enterprise failure and adverse selection are common for both PLS and interest based nance, moral hazard
tate verication are more acute risks of PLS nance (e.g. Gale and Hellwig, 1985). Our paper focuses on comparingd interest based contracts when the lender needs to verify the projects state when the loan has not been repaid.1
racts are surprisingly sparse in the eld. El-Gamal (1997) notes that Islamic banks over-rely on non-PLS contractssing, cost-plus and deferred payment sales), while using an accounting interest rate for their bookkeeping. Chong09) show that less than 1 percent of Islamic banking operations in Malaysia use PLS-based products and thus,
operations are interest based or -pegged. Dar and Presley (2000) cite agency problem as one of the main reasonslack of PLS in Islamic banking. However, they conrm that there is no theoretical reason to make us believe thatently less efcient than interest based contracts., sometimes interest contracts incorporate PLS features. For example, Udry (1990) cites an example from North-
(where Muslims are predominant) in which risk is shared by varying the interest rates that based on the realizationroduction shocks.
in this paper is to investigate the feasibility of PLS contracts in the presence of information asymmetry. We arguentracts pool risk between the lender and the borrower, much as a sharecropper agreement or a prot sharingreement. This risk pooling generates in a different power structure between the borrower and the lender than
contract. As a result, PLS contracts are seen as more equitable than interest contracts, and borrowers will feel aligation to comply with their terms and, in particular, to repay the loan when their project succeeds.2
aper, we develop and test Islamic-compliant micronance contracts that are based on PLS. We use laboratorys as a testbed or wind-tunnel for these contracts (Plott, 1987). The laboratory can test, at very low cost, theresponse to different types of contracts. These contracts can then be adjusted or amended before being used ine laboratory thus serves as an intermediate step between theoretical development and eld-testing.lop two micronance contracts which are Islamic-compliant (prot sharing and joint venture) and comparean interest based loan. Borrowers invest these loans in risky projects, whose outcome is known to them butender. We examine compliance rates, the rate at which individuals comply with the terms of the loan, in theacts under three conditions. A no-enforcement condition involves a simple repayment or reporting decision byer. An enforcement condition allows the lender to follow up on the decision, and to force collection or audit thetcome if appropriate. An enforcement with penalty condition also allows the lender to charge a penalty to the
he has defaulted on the loan or misreported the success of his project.lts indicate that the Islamic-compliant loans induce at least as much compliance as the interest based loans inditions (no-enforcement, enforcement and enforcement with penalty), although the differences between the three
ecrease as the possibility of penalty is introduced. We believe that there is great promise for these types of loansnance context, and discuss strategies for eld implementation in the conclusion of the paper.r is organized as follows. Section 2 discusses the practice of Islamic nance, micronance, and their combination.scribes our theoretical model and Section 4 our experimental design. Section 5 presents our experimental results
nance, micronance, and their combination
nance is in a phase of growth and change. On one hand, Islamic nance is being celebrated as a promising nancial assets exceed $1.3 billion (compared to $160 billion in 1997) and the market is estimated to be growing at 1015
o means the only interesting or important problem in micronance or microcredit. Adverse selection of borrowers, and moral hazard of thect funded with the borrowed money are also relevant issues, but are beyond the scope of this single paper.an anonymous reviewer for pointing out asymmetric power as a causal mechanism for the increased sense of obligation felt in PLS contracts.
254 M. El-Komi, R. Croson / Journal of Economic Behavior & Organization 95 (2013) 252 269
percent per year (The Economist Online April 4, 2012). On the other hand, despite its boom, the Islamic banking industrysuffers from two crucial shortcomings.
The rst is its failure to meet the developmental needs of the Muslim poor. Islamic nance is not designed for micronance,but focuses on large dollar investments and loans. The second shortcoming is the reliance of the Islamic banking industryon mark-upthe principreliance has
Micronthis paper wto the very to contribucosts, leadinbetter funct
Howeveprograms. Spercent ann10 percent these challeBanerjee et
A few prthe takeup They foundincome and
Abbink eOn the one the groups(2009) ndrates in thesensitive to
AnotherMorduch, 2nd no affeexperiment
In contror PLS baseFurthermorexhibit non
2.3. The com
The aimpoor that arpoor. Survereligious re
We belieoften exclu20 percent the variety (2009) abov
The maiularly PLS cprot sharican observe
To compsharing as s (murabaha) products that resemble interest based loans in structure (Appendix A provides a description ofles of Islamic nance and different types of products that have been developed, for the interested reader). This
reinforced the inability of Islamic nance to reach individuals who are seeking contracts without interest.
ance has attracted signicant attention during the last two decades from academics and policy makers alike. Ine will be focusing on a particular type of micronance; microcredit, in which small loans ($100 or so) are made
poor. Microcredit has the potential to include the poor in the formal nancial system, which is widely believedte to economic development and growth, because formal nancial sectors reduce transaction and informationg to a more efcient economy. Levine (2005) reviews both theoretical and empirical studies demonstrating thationing nancial sectors allow countries to grow faster.r, microcredit has its challenges. First, it suffers low take-up rates; relatively few people participate in microcreditecond, in apparent contrast with its mission, microcredit is not cheap. Interest rates between 20 percent and 50ually are quite common in the micronance industry, even in places where annual ination rates do not exceed(Dehejia et al., 2009); these high interest rates reect the additional risk of this type of borrower. Overall, givennges, the impact of microcredit on poverty alleviation has not yet been conclusively demonstrated (see, e.g.
al., 2009).evious studies have examined one or another of these issues. For example, Karlan and Zinman (2009) examinedproblem, and estimated the impact of enhancing access to consumer loans in a eld experiment in South Africa.
signicant positive effects of expanded access to credit on the overall borrower outcome, in terms of employment, food consumption. The marginal loans given through the credit expansion were also protable for the lender.t al. (2006) used laboratory experiments to investigate the impact of interest rates on default in group settings.hand, they nd that higher interest rates increase default rates. On the other hand, higher interest rates increase
punishment of defaulters. Thus, higher interest rate had a mixed effect on default rates. Similarly, Dehejia et al. that different categories of the poor respond differently to increases in interest rates. They increase interest
slums of Dhaka, Bangladesh from 2 percent to 3 percent per month. They nd that poorer borrowers were more the increase in interest rate (with elasticity of .86) than relatively wealthier borrowers (with elasticity of .26).
important choice in microcredit is between individual and group-lending arrangements (Armendariz and007). Group-lending arrangements are the most popular and well-developed. However, Gin and Karlan (2010)ct of group liability on repayment rates; individual loans are repaid at the same rate as group loans in their eld
in the Philippines.ast with this previous research, we will focus on the type of microcredit contract being used; interest basedd. We will assume that all individuals have access to these contracts, and will hold constant the cost of funds.e, we will focus on individual lending, as we believe that this is the setting where individuals are most likely tocompliance.
of this paper is to combine micronance and Islamic nance in order to develop nancial products for thee consistent with Islamic principles and involve PLS contracts. Almost half of the 1.6 billion World Muslims areys by CGAP (2008, 2009) indicated the 40 percent of the Muslim poor reject interest based microcredit loans forasons.ve that identifying effective PLS contracts will enhance access to nance amongst the Muslim poor, who are
ded from banking services. 80 percent of Muslims in India are excluded from banking services, as compared withof non-Muslims. Beyond the Muslim population, identifying potentially new microcredit products will add toof the available products, which could enhance take-up rates of microcredit, as stressed by Karlan and Morduche.
n question of this research is whether using microcredit products that are compatible with Islamic law (partic-ontracts) will increase the rate of default compared with interest based contracts. PLS contracts, which adoptng and joint venture mechanisms, are especially prone to the problem of costly state verication; the borrower
the projects return at no cost, but it is costly for the lender to observe (Gale and Hellwig, 1985).are these different contracts theoretically, we constructed game-theoretic models with differing degrees of risktipulated by different microcredit contracts. The games are one-shot, played between a lender and a borrower.
M. El-Komi, R. Croson / Journal of Economic Behavior & Organization 95 (2013) 252 269 255
The lendproject has > 1); and tdecides whthe lender c(equivalentby auditingwith m prodepends onon the probof the borroverify the r
Below, wgames for tnance (musome agreecalculationas will be sh
3 For simpliverication pr
4 As Cason enot monitorinFig. 1. Interest loan.
er has the opportunity to make a loan of size L to the borrower, who invests the loan in a productive project. The two states of nature: one is successful with probability p, in which it returns L (where the projects multiplierhe other one is failure, where it pays nothing back. The borrower observes the states of nature at no cost, then sheether to repay the loan according to its terms (comply) or not.3 In the last stage (in some treatment conditions),an follow up (Wydick, 2008). In the interest based treatment, he can do so by forcing collection, which is costly
to declaring bankruptcy in Gale and Hellwig, 1985). In the PLS treatments, the lender can verify the state of nature the projects returns ex post. Both acts of costly verication (forcing collection and auditing) are probabilisticbability of success and v cost (v = L, where 0 < < 1). It is worth noting that we assume that verication cost vly on the size of the loan L, whilst Cason et al. (2008), for instance, assume that the cost of monitoring dependsability of a successful monitoring.4 We also assume that the probability (1 m) includes bankruptcy (inabilitywer to repay) in addition to the lenders failure to force the collection of the loan (in case of interest loans) or
eal outcome of the project (in case of PLS).e describe the models for each of our three types of nance. The extensive form representation of the sequential
he interest loan model is shown in Fig. 1; the prot sharing (mudaraba) model is in Fig. 2; and the joint venturesharaka) model is in Fig. 3. The lender starts with L loan that he can choose to lend to the borrower in time t ford upon interest at rate i (in the interest loan model) or share s of the projects returns (in the PLS models). Thes of i and s depend on the lenders required rate of return R (his reservation price or opportunity cost of capital),own below.
city, we control for adverse selection (the borrower type) and moral hazard (the riskiness of the project chosen) to focus on the costlyoblem.t al. (2008) mentions, monitoring can work in different ways in practice. However, since we are concerned with forcing collection/auditing,g per se, we assume that the higher the stakes are, the higher the cost the bank is willing to pay in order to retrieve its loan.
256 M. El-Komi, R. Croson / Journal of Economic Behavior & Organization 95 (2013) 252 269
The borrthere are twthe loan plucalibrate allrepay their and (6), butsome variat
This debextensive fobranch show
(1 + i = 1
Fig. 1 shof nature), defaults onconditions,probability
5 The lenderFig. 2. Prot sharing (P-S).
ower can invest L loan in a risky project that yields L in time t + 1 with probability p or zero otherwise. Thus,o states of nature that can only be observed at no cost by the borrower. The borrower then chooses to pay backs interest i (in case of interest based) or share s (in case of PLS) at nodes (3) and (4) of the extensive form. We
of our models using a myopic lender (Stiglitz, 1990; Wydick, 2008) and assuming (naively) that borrowers willloans when their projects are a success, and not otherwise. The lender only knows if he is paid or not at nodes (5)
does not know the true state of nature. The game proceeds similarly through the following three models withions.
t contract requires that the borrower repays (1 + i)L regardless of the state of nature. The left branch of the gamesrm in Fig. 1 shows the good state of nature, when the project returns L with probability p, whilst the rights the bad state of nature when the project returns zero. The lender calculates his interest rate i as follows:
R)L = p(1 + i)L p + R
ows the extensive form of the complete game of a model with costly verication. On the left branch (good statethe borrower decides in node (3) at t + 1, after learning the real outcome of the project, whether she repays or
the loan. In the no-enforcement condition, the game ends here. In the enforcement or enforcement with penalty the lender can attempt to force collection if the loan is not repaid. The lender succeeds in forcing collection withm at cost v, but at no charge for the borrower.5 In this case, the borrowers payoff is ( 1 i)L if collection is
has no incentive to force collection if the loan is repaid, of course.
M. El-Komi, R. Croson / Journal of Economic Behavior & Organization 95 (2013) 252 269 257
successful awhile he ge
On the rno-enforcemv and probawas succesfor repaymno loan repin the enfor
The Bayeassumed. Kcollection a
EL(DefauThis is s
This is a at node (5) and force co
m Fig. 3. Joint venture (JV).
nd L if collection is unsuccessful. The lender gets (1 + i)L v for successful collections/repayment of the loan,ts (L v) for unsuccessful collection with no loan repayment.ight branch of the tree (bad state of nature), the borrower decides, at node (4), whether to pay or default. In theent condition, the game ends here. In the other conditions, the lender chooses to force collection or not (at costbility m for successful collection). The borrowers payoff is (1 + i)L, if she decides to pay back or if collection
sful when defaulting. She earns zero if defaulting when collection is not successful. The lender earns (1 + i)L vent and collection, or (1 + i)L for repayment with no collection. He earns (L v) for unsuccessful collection withayment. In the enforcement condition, the lender bears the cost of verication v, while it is born by the borrowercement with penalty condition.sian Nash equilibrium is for the borrower to default in both the good and bad states of nature whenever m < 1, asnowing this, the lender will force the collection of an unpaid loan whenever the expected returns when forcingre higher than those without forcing collection:lt Default, Collect) > EL(Default Default, Dont Collect),atised if:
1 + p + Rnecessary condition for forcing repayment, but not sufcient for the lender to lend. The lenders expected payoffsshould be greater than his expected payoffs when not lending (keeping his capital L). Therefore, he should lendllection whenever:
efault Default, Collect) > L
p(2 + )1 + p +
258 M. El-Komi, R. Croson / Journal of Economic Behavior & Organization 95 (2013) 252 269
Then, the equilibrium is: (Default, Default, Collect), if m (p(2 + ))/1 + p + R and m < 1 (when p < (1 + R)/1 + ).Otherwise, the equilibrium is: (Default, Default, Dont Lend), when: m < (p(2 + ))/1 + p + R. This means that the probability
of successful (costly) collection should be high enough for the lending to be feasible.We will choose parameters to induce an equilibrium of (Default Default, Collect) in this model, and parallel parameters
in the other
3.2. Prot s
Accordinto the interoutcome (th
If the booutcome ofaudit is sucwith a pena
(1 + R
Assuminre-write it a
f (L) = (1 s =
So, for aBoth theThe exte
contract. InsL and lea
In the bpay, whethborrower (iNash equilito default wif:
Thus theAs ment
3.3. Joint ve
This modlenders cap
(1 + s =
Then two models.
g to the prot sharing contract, the borrower must repay the loan only if her project succeeds. However, similarest based contract, only the borrower knows the outcome of her project. The borrower self-reports the projectsus determining her repayment requirements).
rrower reports that her project has failed and thus does not repay, the lender can choose to audit (or verify) the the project. The lender succeeds in its audit with probability (m) and pays auditing (verication) fee (v). If thecessful and the borrower has incorrectly reported the outcome of the project, she must repay the loan (possiblylty v in the enforcement with penalty condition).er calculates his share in the projects outcome (s) as follows:
)L = psf (L)
g that the investment project is divisible and the loans production function f(L) is linear (for simplicity), we cans follows:
+ R)L = psL1 + Rp
, s < 1
prot sharing loan to be feasible, the probability of the projects success should be large enough (p > (1 + R)/). lender and the borrower face the game in Fig. 2.nsive form of the game for the prot sharing contract has only a few differences from the one for the interest
the good state of nature and when the borrower repays, the lender receives his share in the projects outcomeves the rest (1 s)L to the borrower.ad state of nature, the situation is different, as the lender will always get (L v) if the borrower does noter auditing is successful or not, which is the distinguishing feature of the prot sharing contract. In this case, thef she chooses not to repay) will earn zero (as opposed to the loss of (1 + i)L in the interest contract). The Bayesianbrium for this game (here, assuming m < 1/p; since p < 1 by denition, this reduces to m < 1), is for the borrowerhether her project succeeds or fails. Knowing that the borrower will Default, the lender will Audit (at node 6)
lt Default, Audit) > EL(Default Default, Dont Lend),
2 + 1 + p + R
lender will audit as long as: m > (2 + )/1 + p + R and m 1.ioned earlier, we will choose parameters to induce this equilibrium of (Default Default, Audit) in this model.
el is the same as prot sharing, except that the borrower contributes capital (k) to the investment alongside theital (l). The lenders share in the projects outcome (s) is calculated as follows:
R)l = ps(l + k), where : l + k = Ll(1 + R)
< 1 :