competition between microfinance ngos: evidence from kiva

World Development Vol. 40, No. 3, pp. 643–655, 2012� 2011 Elsevier Ltd. All rights reserved

0305-750X/$ - see front matter

www.elsevier.com/locate/worlddevdoi:10.1016/j.worlddev.2011.09.009

Competition Between Microfinance NGOs: Evidence from Kiva

PIERRE LYUniversity of Puget Sound, Tacoma, WA, USA

and

GERI MASON *

Seattle Pacific University, WA, USA

Summary. — This paper investigates empirically the effect of competition between microfinance NGOs seeking subsidized capital fromindividual social investors. Although NGO behavior in competitive environments is often subject to controversy, there is little formalempirical evidence to illustrate theoretical findings and inform policy. Using data from Kiva, an online peer-to-peer (P2P) microfinanceplatform, we find that competition has a sizable negative impact on projects’ funding speed and that the effect is stronger between closesubstitutes. This is important because competition for subsidized capital implies organizational pressures similar to those faced whenNGOs compete for donations.� 2011 Elsevier Ltd. All rights reserved.

Key words — Nongovernment Organizations, microfinance, competition, charitable giving, developing countries

* We are grateful to the editor of World Development and five anonymous

referees for insightful comments and suggestions. We benefited from the

comments of Cynthia Howson, Sang-Hyop Lee, James Roumasset, and

seminar participants at the University of Puget Sound and the University

of Hawaii at Manoa. We thank Tipan Verella for accessing the API and

sharing the data. Final revision accepted: August 8, 2011.

1. INTRODUCTION

The promise of microfinance is to deliver poverty alleviationwhile allowing its providers to be financially sustainable,maybe even profitable. 1 However, this popular idea has beenquestioned. In India, microfinance organizations have been ac-cused of seeking profits at the expense of the poor, and evenblamed for the suicides of borrowers in distress (Polgreen &Bajaj, 2010). In fact, empirical evidence suggests that microfi-nance institutions face a trade-off between profitability andreaching the poorest customers (Cull, Demirguc-Kunt, &Morduch, 2007). Then, subsidies can play an important roleto help organizations that seek to serve the poorest (Cull,Demirguc-Kunt, & Morduch, 2009). Subsidies may be explicit,in the form of donor and government grants, or implicit, whenmicrofinance organizations have access to capital at below-market interest rates. In fact, Nongovernment Organizations(NGOs) compete for such implicit subsidies, much like theydo for donations. Yet, to our knowledge, the impact of com-petition on NGOs’ fundraising effectiveness has never beenstudied empirically. This paper examines the role of competi-tion in the case of a novel and growing source of such implicitsubsidies: the American NGO Kiva, an online peer-to-peer(P2P) microlending website. Kiva works with microfinanceNGOs around the world registered as its “field partners.”Field partners select entrepreneurs and post their projectsand profiles on Kiva. On Kiva, prospective individual lenderscan browse through entrepreneurs’ profiles and create a useraccount (formally becoming a “Kiva lender”) to lend and con-tribute to funding projects. Kiva acts as an intermediary be-tween individual Kiva lenders and the field partners whocollect funds on behalf of their entrepreneurs. Figure 1 sum-marizes the different layers involved in Kiva lending. 2 In theevent of repayment, individual Kiva lenders only recover theloan principal and forgo all financial returns. In other words,Kiva lenders are social investors, and Kiva provides its partnerorganizations with access to capital at zero interest, thanks toindividual lenders’ altruism.

The growth of P2P microlending (Bishop & Green, 2009)has created a new space for microfinance organizations to

643

compete for access to capital subsidized by the generosity ofindividuals. This has created a bridge between microfinancepractitioners and individuals interested in charitable giving.Given the importance of the debate regarding subsidies inmicrofinance, it is critical to understand how projects competeon these platforms. This is especially important because thecompetition between microfinance NGOs for subsidized capi-tal implies that they face pressures similar to those faced bynonprofit organizations that compete for donations. Often,NGOs’ behavior in competitive environments generates con-troversy. For example, in the wake of the 2010 earthquakein Haiti, The Lancet criticized NGOs for “jostling forposition” and failing to coordinate humanitarian efforts (TheLancet, 2010). Furthermore, using several case studies, Cooleyand Ron (2002) have found that NGOs’ fierce competition forfunding can distort their incentives in ways sometimesdetrimental to their primary missions. Meanwhile, economictheory has shown that intense competition between NGOsfor donor funds may lead to inefficiently high spending onfundraising and the diversion of funds away from develop-ment projects (Aldashev & Verdier, 2010; Rose-Ackerman,1982).

In this paper, we investigate the impact of competition be-tween projects posted on Kiva on their ability to raise funds,using a rich data set on 132,495 projects between April 2007and September 2009. We find that an increase in the numberof competing projects is associated with significantly slowerfunding times, and the effect is stronger between projects thatare closer substitutes. This finding has important implications.If fiercer competition for social investors makes subsidiesharder to get, the potential trade-off (discussed by Cull et al.(2007)) between profitability and serving the poorest implies

http://dx.doi.org/10.1016/j.worlddev.2011.09.009

Figure 1. The Kiva model.

644 WORLD DEVELOPMENT

that NGOs may have to focus on better off borrowers toremain financially viable. Alternatively, NGOs may increasetheir fundraising efforts to grab a share of the scarce subsidies.Theory suggests this may be associated with a shift of re-sources away from project implementation and a reductionin project impact (Aldashev & Verdier, 2010). In the case ofmicrofinance, it could lead to fewer resources being devotedto screening entrepreneurs or providing them with additionalservices such as business training. 3

Furthermore, we account for the role of the prospective len-der side of the market. First, consistent with economic intui-tion, we find that an increase in the number of registeredKiva lenders helps reduce funding time and relaxes the effectof competition. When competitive pressure is relaxed on themarket for subsidized capital, microfinance NGOs may beable to spend fewer resources on ensuring financial survival,and thus focus on their social impact and outreach to thepoor. However, conversely, this also implies that in timeswhen the pool of prospective social investors is reduced, com-petitive pressure increases, holding the number of competingprojects constant.

In addition, we examine the impact of two specific eventsthat induced large increases in traffic to the website. In eachcase, we focus on a subset of the data, starting one monthprior to the event and up to one month after the event. We dis-cuss how the difference in content between the two events mat-ters. First, on September 4, 2007, Kiva founders appeared onthe Oprah Winfrey Show. We find that the resulting increasedpublicity was associated with a decrease in funding time. Inaddition, it led to a reduction in the effect of competition onfunding time.

Second, on June 10, 2009, Kiva advertised its first microlo-ans to entrepreneurs based in the United States. The event,which specifically advertised a specific subset of projects, waswidely publicized in the media. We find that like Oprah, ithelped relax the marginal effect of competition. However, weshow that average funding time only improved for loans fromNorth America, while it worsened for those from the rest ofthe world. Since the latter represented the vast majority ofthe sample over that period, overall, the event led to an in-crease in funding time. Thus, these results suggest the effectof media exposure depends on its content.

Although the role of subsidies has been discussed in themicrofinance literature, the present paper is the first to exam-ine the impact of competition for subsidies between microfi-nance organizations. Moreover, despite the vast economicliterature on nonprofits, there is little empirical evidence onthe effect of competition to illustrate theoretical findings andinform policy. 4 To the best of our knowledge, this paper isthe first formal empirical investigation of the impact of compe-tition on NGOs’ fundraising success. By shedding light on the

effect of competition on Kiva, our paper contributes to under-standing the functioning of an increasingly popular source ofsubsidies to microfinance organizations.

Our results have clear policy implications. First, since P2Pdevelopment finance makes it easier for registered NGOs toincrease their fundraising effort to advertise projects, it maybe desirable for platform managers to regulate the entry ofnew projects.

Second, we provide evidence of the power of media expo-sure. More specifically, as the case of the US loans event illus-trates, advertising can be used to compensate for the initialdisadvantage of projects that are inherently harder to sell todonors. However, this also means that advertising a specifictype of project to improve its fate may come at the expenseof the fundraising effectiveness of other projects, at least tem-porarily.

The rest of the paper is organized as follows. Section 2briefly reviews related literature to clarify the analytical frame-work. Section 3 describes the data and explains our empiricalmethodology. Section 4 discusses the results and Section 5concludes.

2. COMPETITION BETWEEN NGOs

The issue of NGO entry on the market for donations hasbeen analyzed in theoretical papers. However, to the best ofour knowledge, this is the first econometric study on the effectof competition between charitable projects on their fundrais-ing success. Several papers provide related empirical insightson competition between charities. In the case of US medicalresearch charities, Feigenbaum (1987) finds that more compe-tition is associated with more spending on fundraising but lesson administration. Castaneda, Garen, and Thornton (2008)find similar results in a larger sample of nonprofits across sec-tors. Nunnenkamp and Ohler (2010) find similar results forUS NGOs, using an index of overlap between NGOs’ in for-eign countries. The case of Kiva allows us to use a more directmeasure of competition, by counting the number of entrants atdifferent points in time.

Our analysis contributes empirical evidence illustrating sev-eral important findings in the theoretical literature. Aldashevand Verdier (2010) show that entry of new NGOs on the mar-ket for donations decreases the marginal benefit of fundrais-ing. Here, we provide empirical evidence of this negativeexternality on the Kiva credit market, and its magnitude islarge. Aldashev and Verdier’s (2010) results suggest this hasimportant implications. In particular, when increased entrytriggers fiercer competition for funds, NGOs may shift theiruse of time more toward fundraising and less toward projectimplementation. In fact, posting profiles on Kiva can be time

COMPETITION BETWEEN MICROFINANCE NGOS: EVIDENCE FROM KIVA 645

consuming for field partners (Flannery, 2009), and fiercercompetition may lead them to spend more time and resourcespolishing profiles to appeal to Kiva lenders.

This issue has specific implications in microfinance. Whenaccess to Kiva’s subsidized capital becomes more competitive,partner NGOs may have to compensate by devoting more ef-fort on cost recovery by other means. First, they may have toseek other subsidies, such as grants or other subsidized capitalfrom donor agencies, which require costly applications andfinancial reports. Second, they may have to generate more oftheir own program revenue by raising interest rates and/or fo-cus on more profitable, better off borrowers. 5 As Morduch(2000) argues, implicit subsidies may be essential for somemicrofinance organizations to achieve social missions, insteadof just financial growth.

Moreover, our findings are related to normative theoreticalresults suggesting that there can be too many charitiesentering the market for donations, and that an oversupplyof projects becomes more likely when fundraising and entrycosts decrease (Aldashev & Verdier, 2010; Economides &Rose-Ackerman, 1993; Pestieau & Sato, 2006). 6 This isrelated to Rose-Ackerman’s (1982) argument that competi-tion for funds between charities leads to an excessive numberof brochures in the mail. These papers suggest that theremay be a rationale to regulate entry on charitable markets. 7

Kiva co-founder, Flannery (2009), argued that althoughposting profiles adds costs for field partners, this cost maybe minimal relative to the benefit. Thus, the potential foroversupply on P2P platforms may be high.

Our results also illustrate the importance of the interplaybetween the NGO and the donor side of the market, as themodel by Aldashev and Verdier (2010) makes clear. In theirmodel, an increase in the size of the pool of donors relaxesthe pressure of competition between NGOs, an effect thatour paper documents empirically. 8 Their model also accountsfor the fact that the fundraising efforts of NGOs with similarprojects compete fiercely with each other. This is related to ourresult that a project’s funding time is more highly impacted bycompetition from similar projects, which may be perceived ascloser substitutes by lenders. Van Diepen (2008) provides a re-lated empirical result. She finds that a charity’s own mailingsare short run substitutes. She also finds that competitive mail-ings are short run complements in the sense that they help in-crease the pie available for all charities. In our setting, thecompetition between projects is different because prospectiveindividual lenders have to come to the website first to browsethrough the listings, and new projects tend to reduce the visi-bility of existing ones.

The present paper contributes to an emerging literature ana-lyzing P2P development finance. Desai and Kharas (2009)compare the patterns of official aid with the selection of pro-jects by philanthropic citizens on Kiva and Global Giving.They show that while official aid is driven by country charac-teristics, individual donors on P2P websites respond primarilyto project characteristics. 9 Our paper provides new insights onP2P development finance by focusing on the impact of compe-tition.

More broadly, our paper is related to the literature onthe welfare consequences of competition between microfinanceproviders. McIntosh and Wydick (2005) argue thatcompetition between providers can be harmful for the poorestborrowers. Here, we focus on the impact of competition on theNGOs themselves, not the local beneficiaries. The latter mayin turn be affected if, in the event of difficulty to market theirprofile on Kiva, their future requests are turned down by theNGO.

3. DATA AND EMPIRICAL METHOD

(a) The data

Kiva is a P2P microfinance platform connecting philan-thropic individuals with poor entrepreneurs around the world.On Kiva, prospective individual lenders can browse throughentrepreneurs’ profiles and contribute to funding projects. Inthe event that the loan is repaid, lenders can use the fundsto lend again or simply withdraw their balance. Kiva connectsindividual lenders and entrepreneurs through local field part-ners worldwide. The latter are established microfinanceorganizations and are in charge of posting their chosen entre-preneurs’ profiles online, and monitoring loan use and repay-ment. Overall, the data set covers 150 field partners from 50countries. 10 In practice, although the idea of Kiva is to pro-mote a P2P approach between individual lenders and entrepre-neurs, the money collected online is disbursed to the fieldpartner organizations. The latter usually “backfill” loans al-ready made to existing entrepreneurs before advertising themon Kiva. However, lenders still bear the risk specific to theirchosen project.

Our data for this paper consist of projects funded on Kivafrom April 17, 2006 to September 24, 2009. Kiva makes itssource database available through an application user inter-face. The original data set contained 133,700 loans. Afterremoving observations due to inconsistencies and missingdata, 11 our final working data set contains 132,495 observa-tions.

We are interested in estimating the effect of competition be-tween projects on their fundraising speed. Before turning tothe description of our key variables, we briefly present themain characteristics of the sample. Descriptive statistics are re-ported in Table 1.

The average loan size requested for a project is $694.32 andthe average loan term is 10.75 months. Only 11.2% of the sam-ple concerns group loans, and there are very few male ormixed groups. About 76.2% of loans in the sample were madeto individual women or groups of women. In the analysis be-low, we control for gender with a dummy variable, taking thevalue one for loans to women entrepreneurs (group and indi-vidual), and zero otherwise.

Kiva assigns a risk rating, from one to five stars, to each ofits field partners, taking into account financial sustainabilityand reliability. A five-star rating refers to the lowest risk.According to Kiva, the risk rating “is intended to providemore insight for those who are sensitive to repayment risk.”The majority of projects (70.2%) were posted by field partnerswith a rating of four or five stars.

Finally, projects are broken down into (mutually exclusive)sectors of activity, including agriculture, arts, clothing, con-struction, education, entertainment, food, health, housing,manufacturing, personal use, retail, services, transportation,and wholesale.

(b) Description of key variables

(i) Funding timeThroughout the analysis, our dependent variable is funding

time (in logarithm), defined as the number of minutes it tookfor each project to be fully funded from the moment it wasposted. A project is fully funded once the sum of individualcontributions by Kiva lenders covers the full loan amount re-quested. We use funding time as a measure of projects’ fund-raising success. Indeed, if some projects fund faster, it isreasonable to assume they were favored by lenders. Then, to

05

1015

2025

Perc

ent

180

360

540

720

1440

2160

2880

4320

5760

1008

0M

ore

Minutes to fund

Figure 2. Distribution of project funding times.

Table 1. Descriptive statistics

N = 13,2495 Mean (standard deviation) 25th, 50th, 75th percentiles Maximum

Funding time (in min) 3193 (6708) 176, 556, 2311 120,975

Competition variables

Total number of competing projects 372.93 (374.78) 61, 224, 634 1682Number of competing projects from the same region 101.19 (125.97) 13, 43, 157 760Number of competing projects advertising the same sector 69.72 (86.58) 9, 27, 107 451Number of competing projects posted by the same NGO partner. 14.88 (22.85) 3, 7, 17 260Total number of registered lenders as of day posted 296,646 (137,661) 241,719, 304,468, 419,094 490,833Number of new registered lenders on day posted 522.38 (432.13) 339, 420, 2290 5560

Project and entrepreneur characteristics

Loan amount (in United States $) 694.32 (586.47) 350, 550, 925 10,000Loan term (in months) 10.75 (4.26) 7, 11, 13 38Percentage of loans to women entrepreneurs (individuals and groups) 76.2%Percentage of group loans 11.2%Percentage of projects from field partners rated 4 or 5 stars 70.2%


estimate the effect of competition on fundraising success, wecontrol for project and entrepreneur characteristics through-out.

Figure 2 shows the distribution of funding times. The aver-age in the sample is 3193 min, with a standard deviation of6708 min. The majority of loans, 68%, get funded in less than24 h after being posted, and another 10% are funded withinthe following 24 h.

During our sample time frame, although most projects werefully funded, a few expired and, as a result, did not appear inthe database. Since then, there have been more cases of expi-ration, 12 and Kiva added the possibility of sorting by “expir-ing soon,” so that interested lenders can actively try to preventexpiration. In our analysis, we show that although most pro-jects fund very fast, others take much longer, thereby bringingthem closer to expiration. The magnitude of the effect of com-petition suggests that a one standard deviation increase incompetitors could push more projects toward expiration.Thus, if the database were significantly scaled up, some pro-jects may have increasing difficulty to get funded.

Furthermore, although advertised entrepreneurs were usu-ally pre-funded by the field partners, the latter advance theloans with their own funds and count on zero-interest capitalon Kiva to “backfill” the loans. Delayed or lack of access tozero interest capital implies seeking possibly more expensivesources of capital or depleting savings. In fact, Kiva CEO

Flannery has expressed concerns about the number of compet-ing projects, the risks expiration represents for field partners,and mentioned the need for restricting entry (Flannery, 2009).

(ii) CompetitionKey to the analysis is the measure of competition faced by

each project. There is considerable variation in the size ofthe fundraising pool over time. It is reasonable to expect pro-jects facing more competition to fund slower, as lenders havemore options to choose from. Our base measure is the numberof fundraising projects at the time each project was posted. Inother words, for each project i, we counted the number of pro-jects that fulfilled two conditions: first, they were posted beforeor at the same time (exact minute) as i, and second, they werenot yet fully funded by Kiva lenders. 13 This gives us a straight-forward measure of the size of the competing pool that eachproject faced when entering the market. Then, we refine thismeasure of competition to examine whether a project’s fund-ing time is more highly impacted by projects of a similar type(defined by region, sector, or field partner) than by others.Table 1 reports descriptive statistics for our competition mea-sures.

On average, a project i was posted facing 372 competitors.Among these, 101 projects were from the same region as i,and 69 were advertising the same sector of activity. Comparedto a given project, those from the same region or sector may beperceived by lenders as more substitutable and thus have ahigher impact on its funding time. We examine this possibilitybelow. Since lenders can sort project listings by region or sec-tor, projects of a given type are more likely to appear on thesame list and thus compete directly with one another.

Although it would take more clicks for lenders to examinethe list of projects of a given field partner (there is no sortingoption by this criterion), projects posted by a given organiza-tion look remarkably similar and may thus be perceived asmore substitutable from the perspective of lenders. Indeed,each partner NGO provides a biographical sketch of the entre-preneur and tends to follow the same structure, tone, and evenpicture setting across all its project descriptions. On average, aproject faced 14 competing ones posted by the same field part-ner.

(iii) The lender side of the marketWe expect an expansion of the pool of prospective lenders to

have two effects of importance for our analysis. First, it shouldreduce funding times overall. Second, it may also relax the im-pact of the number of competing projects on funding time.


In the first part of the analysis, we control for the size of thepool of registered Kiva lenders as of the day each project wasposted. 14 On average, each project was posted on a day when522 new Kiva lenders registered, and faced a total of 296,646registered Kiva lenders.

Then, we focus on two specific events that increased trafficto the website in order to examine the effect of a sudden andlarge increase in the number of lenders. In each case, we focuson a subset of the data, starting one month prior to the eventand ending one month after the event. Comparing the twoevents allows us to discuss how the effect of media exposuremay depend on its content.

First, on September 4, 2007, Kiva founders appeared onthe Oprah Winfrey Show. This event raised awareness ofKiva at a time when it was not yet widely known. On thatday, website visits quadrupled compared to the previousday, and 4,388 new lenders registered (Flannery, 2007).Our focus on the two-month period surrounding the eventleads to a sub sample with 2743 projects in the month beforethe show and 1901 projects after. During the month beforethe event, the average number of new registered Kiva lend-ers per day was 286. By contrast, in the month after, anaverage of 815 new lenders registered each day. In fact, thisnumber first rose to 1806 new users per day during the weekimmediately following the show. New membership sloweddown in the following three weeks but remained higher thanduring the previous month, with 513 lenders per day. Mean-while, during the previous month, funding time was1218 min on average, and fell to 463 min in the month afterOprah.

Much later, on June 10, 2009, Kiva advertised its first loansto entrepreneurs based in the United States. The event waswidely publicized in the media, and like in the case of Oprah,drew increased attention to Kiva, thereby increasing the pro-spective lender side of the market. On that day, 2095 new lend-ers registered, more than five times the average prevailing overthe previous 30 days (402). Over the week following the event,an average of 1042 new lenders per day signed up.

In this case, 7211 projects were posted during the monthprior to the event, when average funding time was 4821 min.Then, surprisingly, for the 6446 projects posted in the monththat followed, average funding time increased to 5485 min.This is counterintuitive because the event did attract a signif-icant new pool of lenders.

Figure 3 shows the distribution of funding times before andafter each media event. In the month following Oprah, fund-ing times become more concentrated toward the lower endof the distribution, and the high end of the distribution isgreatly reduced. However, in the case of the US loans launch,there is no noticeable shift of the distribution toward the lowerend. In fact, after the event, there is even a slightly lower per-centage of projects in the lower end, and a higher percentage inthe higher end, compared to before.

The empirical analysis below will make sense of this. Fornow, we note two interesting differences between the twoevents. First, the Oprah feature took place at a time whenKiva was relatively unknown, whereas by June 2009, it was al-ready a widely acclaimed organization. Second, while Oprahtaught a new wide audience about the existence of Kiva andits general mission, the US loans feature drew attention to aspecific subset of entrepreneurs, the American ones. In fact,on average, funding time greatly improved for projects fromNorth America (from 9196 to 5852 min) but worsened forthose from the rest of the world (from 4668 to 5539 min).We discuss this further when we interpret our results inSection 4.

(c) Empirical method

We explore the issue of competition on Kiva from two com-plementary angles. First, we estimate the effect of the size ofthe competing pool on funding time, using our four measuresof competition. Second, we investigate the impact of the twomedia events described above. Throughout the analysis, thedependent variable remains log(y), the logarithm of fundingtime y measured in minutes.

Let Ni denote the number of competing projects facing pro-ject i when it was posted. Then, Xi denotes the set of charac-teristics of project i, including loan amount, loan term, agender dummy, a group dummy (also interacted with the num-ber of borrowers), and a dummy variable for NGOs’ risk rat-ing. Xi also includes sector and country dummies.

The prospective lender side of the market plays an impor-tant role. Indeed, a larger pool of potential lenders is likelyto reduce funding time. Let Li be the total number of regis-tered lenders as of the day project i was posted. Note that un-like our measure of competition, which is taken at the exactminute each project was posted, data on the number of regis-tered Kiva lenders are only available for each day. Thus, wecan control for day to day changes in the number of prospec-tive lenders, but not for changes throughout a given day. Thisis still a good way to capture the effect of the lender side of themarket. Later in the paper, in our analysis of media events, wewill investigate the impact of large increases in the number ofprospective lenders around the specific events.

Let yi be the funding time of project i. In order to estimatethe effect of competition on funding time, we estimate the fol-lowing equation using OLS:

logðyiÞ ¼ b0 þ b1N i þ b2Li þ b3N i � Li þ cX i þ ui ð1ÞThe effect of competition on funding time is given by

b1 + b3Li. Since the dependent variable is the log of fundingtime, this represents the percentage change in funding timeassociated with an increase in competition by one unit, for agiven size of the lender pool Li. We expect this to be positive,with a positive b1 but a negative b3 since a larger pool of lend-ers may relax the effect of competition. Furthermore, we ex-pect the effect of the prospective lender pool, b2 + b3Ni, tobe negative. Then, a negative b3 would also mean that the ef-fect of the prospective lender pool is stronger the higher num-ber of competitors.Then, we turn to the impact of differenttypes of competition by breaking down Ni into two parts, ni

and mi, in three different ways. Let mi be the number of com-peting projects from a given “type,” at the moment project iwas posted, and ni be the rest of the competing pool, that is,Ni = mi + ni. We estimate the following equation:

logðyiÞ ¼ b0 þ b1mi þ b2ni þ b3Li þ b4mi � Li

þ b5ni � Li þ cX i þ ui ð2ÞFirst, Eqn. (2) is estimated with mi as the number of compet-

ing projects from the same region at the time project i wasposted, leaving ni as the number of competing projects froma region different from that of project i. Then, we estimate(2) with mi as the number of competing projects in the samesector of activity at the time project i was posted, leaving ni

as the number of competitors whose sectors differ from thatof project i. Finally, we take mi as the number of competingprojects posted by the same NGO field partner at the time pro-ject i was posted, leaving ni as the number of competing pro-jects posted by other field partners.

We expect b1 + b4Li and b2 + b5Li to be positive: anincrease in the number of competing projects should be

Before Oprah After Oprah

Before US loans launch After US loans launch

05

1015

20Pe

rcen

t

180

360

540

720

1440

2160

2880

4320

5760

1008

0M

ore

Minutes to fund

010

2030

4050

Perc

ent

180

360

540

720

1440

2160

2880

4320

5760

1008

0M

ore

Minutes to fund

05

1015

2025

Perc

ent

180

360

540

720

1440

2160

2880

4320

5760

1008

0M

ore

Minutes to fund

05

1015

2025

Perc

ent

180

360

540

720

1440

2160

2880

4320

5760

1008

0M

ore

Minutes to fund

Figure 3. Distribution of funding times before and after media events.


associated with slower funding times. Moreover, we expectthat b1 + b4Li to be greater than b2 + b5Li, meaning that pro-ject i’s funding time may be more sharply affected by the pres-ence of competitors with more similarities (same region,activity or field partner). In other words, the effect of compe-tition may be stronger among closer substitutes.

Next, we turn to an additional account of the role of the len-der side of the market. Instead of using the number of lendersdirectly, we use the two media events to investigate whetherthe content of media exposure matters. First, we examine asub sample including one month before Kiva founders’appearance on Oprah and one month after. For this, we createa dummy variable, AfterOprah, taking the value one for pro-jects posted on or after the day of the show, and zero for thoseposted earlier. We estimate the following equation, using OLS:

logðyiÞ ¼ b0 þ b1Ni þ b2AfterOprahþ b3AfterOprah

� Ni þ cX i þ ui ð3Þ

In Eqn. (3), since the Oprah dummy aims to capture thelarge increase in registered lenders due to exposure on theshow, we do not include the number of registered lenders di-rectly since the latter is the outcome of interest of the mediaevent. The average percentage change in funding time beforeand after Oprah, controlling for other factors, is given byb2 + b3Ni. We expect it to be negative, that is, the large in-crease in the number of prospective lenders following Oprahshould have reduced funding time. 15 The coefficient on theinteraction term, b3, allows us to estimate and compare the ef-fect of competition before and after Oprah. Indeed, we expectthe increased publicity generated by the show to reduce the ad-verse impact of competition on funding time due to the in-crease in the lender side of the market. To sum up, weexpect both b2 and b3 to be negative.

Furthermore, following the idea behind Eqn. (2), we esti-mate the following equation to capture the effect of differenttypes of competition during the two-month period aroundOprah:


logðyiÞ ¼ b0 þ b1mi þ b2ni þ b3AfterOprahþ b4mi

� AfterOprahþ b5ni � AfterOprahþ cX i þ ui ð4Þ

Consistent with the interpretation of Eqn. (2), we expect mi

to have a greater effect than ni, that is, b1 + b4 AfterOprah >b2 + b5 AfterOprah. Further, Oprah may have relaxed the ef-fect of both types of competition, that is, we expect negative b4

and b5.Finally, we examine a sub sample over the two-month per-

iod surrounding the launch of loans to entrepreneurs basedin the United States. The event was widely publicized, andgenerated an increase in traffic to the website. However, inSection 3 (b), we showed that average funding time actuallyincreased in the month following the event. Our methodol-ogy will help make sense of this seemingly counter-intuitivefact.

First, we perform the same exercise as in the case of Oprah.We include a dummy variable, AfterUSA, taking the valueone for projects posted on or after the day of the event, andzero for those posted earlier. We estimate the following equa-tion, using OLS:

logðyiÞ ¼ b0 þ b1Ni þ b2AfterUSAþ b3AfterUSA

� Ni þ cX i þ ui ð5Þ

The average percentage change in funding time before andafter the launch of US loans, given by b2 + b3Ni, is expectedto be negative. Moreover, if the event helped relax the effectof competition, b3 should be negative. However, consistentwith the descriptive analysis, we will show in Section 4 thatthe estimated b2 is positive, so that at least for some valuesof Ni (below a certain threshold), funding times increasedoverall after the US loans event, even after controlling forother characteristics.

Next, to make sense of the overall increase in funding time,we replace the country dummies by two broader categories:USA takes the value one for loans to American entrepreneurs,zero otherwise; ROW takes the value one for loans from therest of the world excluding North America. These two dummyvariables are included in the following model, leaving the baseregion to be North America excluding the United States. Weestimate the following equation by OLS:

logðyiÞ ¼ b0 þ b1Ni þ b2AfterUSAþ b3AfterUSA � N i

þ b4USAþ b5ROW þ b6ROW � AfterUSAþ cX i þ ui

ð6Þ

Now, controlling for other characteristics, we can comparethe average difference in funding speed before and after theUS loans event, distinguishing between the rest of the worldand North America. For the rest of the world (ROW = 1), thisdifference is given by b2 + b3Ni + b6. For North America(excluding the United States), it is given by b2 + b3Ni. We dis-cuss the estimates in Section 4 (c) and interpret this compari-son in light of the specificity of the US loans event, comparedto the Oprah feature.

Finally, we check that the results regarding different types ofcompetition (by region, sector and field partner) hold in thiscase, by estimating:

logðyiÞ ¼ b0 þ b1mi þ b2ni þ b3AfterUSAþ b4mi

� AfterUSAþ b5ni � AfterUSAþ b4USA

þ b5ROW þ b6ROW � AfterUSAþ cX i þ ui ð7Þ

4. RESULTS

(a) Competition and funding time

Table 2 reports the results from estimating Eqn. (1). Sincethe impact of competition is our primary focus, for clarity,part (A) reports the coefficients on the competition and lenderpool variables, and part (B) reports the coefficients on projectand entrepreneur characteristics, which will be described inSection 4 (b).

Column (1) shows that a one standard deviation increase inthe number of competing projects present at the time a projectwas posted (374) is associated with an increase in funding timeby 59%. In column (2), we control for other factors that mayaffect funding time, including the number of registered lenders,loan amount, loan term, gender, group loans and number ofborrowers, and the field partner’s risk rating. 16 Then, esti-mated at the mean number of registered lenders (296,646), aone standard deviation increase in the number of competingprojects (374) is associated with an increase in funding timeby 84%. For the average project, this would represent an in-crease by 2682 min. This is a very large effect, given that50% of projects in the sample were funded in less than600 min. In other words, as expected, projects posted withina larger pool of competitors fund slower and the effect is quan-titatively important.

Meanwhile, as expected, an increase in the number of regis-tered lenders helps reduce funding time and relaxes the effectof competition. Estimated at the average competition (372),a one standard deviation increase in the pool of lenders(137,661) is associated with a 28% reduction in funding time.

Furthermore, column (3) differentiates between the numberof new lenders registered on the day a project was posted andthe size of the lender pool as of the day before. Estimated atthe average size of competition (372), an increase in the num-ber of new registered lenders on the day project i was postedby one standard deviation (432) is associated with a decreasein funding time by 19%. As expected, an increase in the lenderside of the market helps reduce funding time and relaxes theeffect of competition.

The results are consistent with the standard economic intu-ition that entry on a market erodes profits for participants.Applied to the case of NGOs, Aldashev and Verdier (2010)show theoretically that entry of new NGOs on the marketfor donations decreases the marginal benefit of fundraising.Here, using funding time as a measure of fundraising success,we provide empirical evidence of this negative externality ofentry on the Kiva credit market. Moreover, the magnitudeof the effect is large. Theoretically, this may imply an oversup-ply of projects on Kiva. For example, assume that from thepoint of view of lenders, there is substitutability across com-peting projects on Kiva, which implies that an influx of newprojects decreases the visibility of existing ones (a “businessstealing” effect). Further, assume that each NGO only takesinto account its own fundraising success when posting its pro-jects on Kiva, thus ignoring the negative externality it exertson other NGOs. Then, the equilibrium number of projects willbe greater than the optimal number (that which maximizes thecollective welfare of all NGOs).

Our results illustrate empirically Aldashev and Verdier’s(2010) theoretical finding that a higher number of projects“triggers tougher competition for funds” (Aldashev & Verdier,2010, p. 53). Their model suggests this has important implica-tions for the actual impact of development projects. In partic-ular, they show that as increased entry triggers fiercercompetition for funds, NGOs shift their use of time more

Table 2. The effect of competition on funding time

(A) Competition and funding timeDependent variable: log of funding timeOLS estimatesRobust standard errors (clustered at field partner level) reported in parentheses

(1) (2) (3)

Competition at time posted (Ni) 0.0016*** (0.0001) 0.0028*** (0.0001) 0.0029*** (0.0001)Total number of registered lenders as of day posted (Li) �1.43e-06*** (3.76e-07)Competition � Lenders (LiNi) �1.81e-09*** (6.60e-10)Number of new registered lenders on day posted �0.0003*** (0.0003)Competition � new lenders �3.82e-07*** (1.01e-07)Total pool of registered lenders up to day before loan was posted �1.62e-06** (3.61e-07)Competition � pool lenders �1.60e-09** (6.16e�10)Controls for loan and borrower characteristics (see below) No Yes YesSector dummies No Yes YesCountry dummies No Yes YesN 132,495 132,495 132,495Adjusted R2 .1178 .4129 .4260

(B) Other determinants of funding time (project and entrepreneur characteristics)Robust standard errors (clustered at Field Partner level) reported in parentheses

(2) (3)Loan amount 0.0013*** (0.0001) 0.0013*** (0.0001)Loan term 0.0658*** (0.0124) 0.0666*** (0.0119)Female �0.3836*** (0.0506) �0.3841*** (0.0510)Group 1.1102*** (0.1448) 1.0997*** (0.1452)Group � number of borrowers �0.1564*** (0.0291) �0.1556*** (0.0289)Low risk 0.0879 (0.1102) 0.0993 (0.1126)

* Denote significance at 10%.** Denote significance at 5%.

*** Denote significance at 1%.


toward fundraising and less toward project implementation.Moreover, the authors note that a change in information tech-nology that eases NGOs’ access to private donors, a key aspectof P2P charitable markets, may contribute to excessive entry.

Furthermore, Table 3 reports the results of estimating Eqn.(2) with three different measures of competition. Column (1)compares the effect of competitors from the same region onproject i’s funding time to that of competitors from differentregions. 17 An increase in the number of competing projectsby one standard deviation (126) from the same region is asso-ciated with a 42% increase in funding time (estimated at theaverage number of lenders, 296,646). By contrast, a similar in-crease in the number of competitors from a different regionleads to a lower increase in funding time, by 22%. The nullhypothesis that the two effects are similar (H0: b1 + 296,646b4 = b2 + 296,646b5) can be rejected at the 1% level(F = 11.14). This suggests that a project’s funding time ismore strongly affected by competitors in the same region thanby those from other regions.

Column (2) compares the effect of competing projects in thesame sector of activity to that of projects in different sectors.While a one standard deviation increase (86) in competitorsfrom the same sector leads to a 35% increase in funding time(again, at the average number of lenders), a similar increasein competitors in different sectors is associated with a 15% in-crease in funding time, and the difference between the two ef-fects in statistically significant (F = 31.00).

Finally, column (3) shows that competitors posted by thesame NGO have a larger effect on funding time than compet-itors posted by different NGOs.

Overall, the results in Table 3 show that competition has astronger effect on a project’s funding time when it involves

projects with similarities, such as those from the same region,sector, or field partner. This suggests that lenders may perceivea higher degree of substitutability between projects of a giventype. The idea that competition may be fiercer between closersubstitutes is consistent with economic intuition. Our result isrelated to Aldashev and Verdier’s (2010) theoretical point thatdonations to an NGO depend negatively not only on the num-ber of competitors overall, but particularly on the fundraisingefforts of NGOs with similar projects. In their model, theauthors introduce a parameter of congruence between eachdonor’s preferred type of project and NGOs’ field of interven-tion. Then, a higher variety of NGOs on the market helps in-crease donors’ utility from giving, since they can find projectsclose to their ideals more easily.

Combined with the theoretical insights of Aldashev andVerdier (2010), our results have clear policy implications forinventory management on Kiva and similar platforms. First,platform designers need to take into account the negativeexternality associated with the entry of new competing pro-jects. Then, limiting the number of projects fundraising atthe same time may be desirable. Second, since the negativeexternality generated by entry is especially strong among pro-jects with similarities, ensuring a wide variety of types of com-peting projects across regions and sectors may help improveall projects’ funding time. These insights may apply to otherP2P development finance platforms, such as Global Giving,and our analysis could be extended to other interesting cases.

(b) Project and entrepreneur characteristics

Part (B) of Table 2 reports the coefficients on project andentrepreneur characteristics, including loan amount, loan

Table 3. The effect of competition from closer substitutes

(A) Competition and funding timeDependent variable: log of funding timeOLS estimatesRobust standard errors (clustered at field partner level) reported in parentheses

(1)a (2)b (3)c

Competition from same category at time posted (mi) 0.0046*** (0.0009) 0.0060*** (0.0009) 0.0055 (0.0086)Other competition (ni) 0.0020*** (0.0004) 0.0021*** (0.0003) 0.0027*** (0.0003)Total number registered lenders as of day posted (Li) �1.41e-06*** (2.76e�09) �1.41e-06*** (3.79e�07) �1.78e-06*** (3.44e�07)mi � Li �4.02e-09 (3.70e-07) �6.28e-09 (2.33e-09) �5.22e-08*** (2.02e-08)ni � Li �8.56e-10 (1.02e-09) �8.55e-10 (7.33e-10) �8.56e-10** (1.02e-09)Controls for loan and borrower characteristics (see below) Yes Yes YesSector dummies Yes Yes YesCountry dummies Yes Yes YesN 132,495 132,495 132,495Adjusted R2 .4148 .4142 .4264

(B) Other determinants of funding time (project and entrepreneur characteristics)Robust standard errors (clustered at field partner level) reported in parentheses

(1)a (2)b (3)c

Loan amount 0.0013*** (0.0001) 0.0013*** (0.0001) 0.0013*** (0.0001)Loan term 0.0665*** (0.0122) 0.0661*** (0.0123) 0.0644*** (0.0115)Female �0.3847*** (0.0501) �0.3830*** (0.0506) �0.3830*** (0.0506)Group 1.0982*** (0.1417) 1.1105*** (0.1442) 1.1052*** (0.1380)Group � number of borrowers �0.1555*** (0.0285) �0.1568*** (0.0290) �0.1641*** (0.0271)Low risk 0.1007 (0.1132) 0.0889 (0.1105) 0.1059 (0.1126)

a mi is the number of fundraising projects from the same region at the time a project was posted.b mi is the number of fundraising projects advertising the same sector at the time a project was posted.c mi is the number of fundraising projects posted by the same NGO at the time a project was posted.




term, gender, the number of borrowers, and a dummy variablefor NGOs’ risk rating. The coefficients remain similar in mag-nitude and significance across in Tables 2 and 3. In what fol-lows, we use the estimates reported in column (2) of Table 2.However, since our main focus in this paper is the impact ofthe size of the competitive pool, we only comment briefly onother characteristics for the sake of completeness.

First, not surprisingly, an increase in loan size by one stan-dard deviation is associated with an increase in funding timeby 76%. Second, longer loan terms are associated with slowerfunding times by 26%, which may indicate some impatience onthe part of lenders. After all, for charitable lenders, the possi-bility of repayment and being able to use the same funds to domore good constitutes one of the key benefits of microlendingon Kiva compared to traditional charitable giving.

Furthermore, women fund 38% faster than men on average.Thus, there is a clear gender preference on Kiva, which maycome from a willingness to target the most vulnerable borrow-ers as well as the idea that women have higher repaymentrates. This is in line with the widespread practice of womentargeting by microfinance NGOs (Morduch, 1999).

In addition, evaluated at the mean number of borrowers(1.7), requests for group loans were funded 84% slower thanindividual loans. However, groups of seven borrowers or morefund faster than individuals. Given that group solidarity canimprove repayment rates, this could indicate that lenders arerisk averse. Alternatively, lenders may consider that grouploans allow them to participate in helping more beneficiariesat once.

Finally, the coefficient on risk rating is not statistically sig-nificant, showing no evidence that safer loans fund faster thanriskier ones, at least based on the star rating provided.

This brief note on project and entrepreneur characteristicssuggests that Kiva lenders’ selection criteria may include tar-geting poor and vulnerable borrowers. Cull et al. (2009) sug-gest that donors and social investors will continue tosubsidize microfinance organizations only if the latter consis-tently demonstrate their economic and social impact. If so,NGOs will be able to keep competing for funds on websitessuch as Kiva. Then, it is all the more important to understandhow competition between projects affects their fundraisingsuccess.

(c) Competition and media exposure

(i) The Oprah effectTable 4 reports the results from estimating Eqns. (3) and (4).

As explained above, the sample is restricted to a period start-ing one month prior to the feature on Oprah to one monthafter. We control for project and entrepreneur characteristics,and include sector and country dummies throughout.

Column (1) reports results for Eqn. (3). Estimated at theaverage number of competitors over the two months underconsideration (52), projects funded 49% faster in the monthfollowing Kiva’s exposure on the show. Although the individ-ual coefficient on the AfterOprah dummy is not statisticallysignificant, one must account for the interaction term to assessthe effect (Ni is never equal to zero in the sample). The nullhypothesis that the Oprah effect, b2 + b3Ni is equal to zerofor Ni = 52 can be rejected at the 1% level (F = 23.19). Thisshows that the large increase in prospective lenders generatedby the event allowed all projects to be funded faster. This effectis intuitive and confirms that media advertising can be apowerful tool for NGOs. Moreover, the coefficient on the

Table 4. The Oprah effect

Dependent variable: log of funding timeOLS estimatesRobust standard errors (clustered at field partner level) reported in parentheses

(1) (2)a (3)b (4)c

Competition at time posted (Ni) 0.0217*** (0.0052)Competition from same category at time posted (mi) 0.0341*** (0.0059) 0.0369*** (0.0068) 0.0396*** (0.0052)Other competition (ni) 0.0139** (0.0062) 0.0177*** (0.0050) 0.0179*** (0.0052)AfterOprah = 1 after the Oprah feature, 0 otherwise �0.0139 (0.3682) �0.1507 (0.4018) �0.0639 (0.3624) �0.1161 (0.3680)AfterOprah � Ni �0.0092 (0.0056)AfterOprah � mi �0.0091 (0.0055) �0.0126* (0.0072) �0.0023 (0.0111)AfterOprah � ni �0.0070 (0.0070) �0.0076 (0.0055) �0.0081 (0.0056)Controls for project and entrepreneur characteristics Yes Yes Yes YesSector dummies Yes Yes Yes YesCountry dummies Yes Yes Yes YesN 4644 4644 4644 4644Adjusted R2 .5860 .5958 .5884 .5952





interaction term is negative, suggesting that the publicityhelped relax the effect of competition, though this is not statis-tically significant.

Then, columns (2), (3), and (4) report estimations of Eqn. (4)with our three measures of competition by type. First, theOprah effect is statistically significant and of similar magni-tude as before. Estimated at the mean of the competition vari-ables, in all three columns, the Oprah effect representsabout a 50% decrease in funding time compared to beforethe show.

Finally, consistent with the results presented in Section 4 (a),a project’s funding time seem to be affected more by the pres-ence of competitors from the same type (whether type is de-fined by region, sector or field partner) than by those of adifferent type.

(ii) The effect of the launch of US loansTable 5 reports the results from estimating Eqns. (5)–(7).

Like in the previous subsection, the sample is restricted to aperiod starting one month prior to the event to one monthafter. We control for project and entrepreneur characteristics,and include sector dummies throughout.

Column (1) reports estimates for Eqn. (5), with countrydummies. The coefficients on competition and on the interac-tion term remain consistent with our previous intuition. Onaverage, evaluated at the mean of the number of competingprojects in the subsample (716), projects funded 17% fasterafter the event compared to before. Moreover, the eventhelped relax the effect of competition. However, since the coef-ficient on AfterUSA is positive, evaluated at a number of com-petitors lower than 561, the event is found to have increasedfunding time.

Next, in column (2), instead of using country dummies, weinclude dummy variables to differentiate between the newcountry featured in the event, the United States, and the restof the world, excluding North America. The base region, then,is North America, excluding the United States. This corre-sponds to Eqn. (6).

First, we note that controlling for other factors, the projectsof entrepreneurs from the United States were funded over five

times faster than others. The coefficient is practically verylarge. However, it is worth pointing out that during the monthfollowing their introduction, only 44 US loans were posted,representing only 0.68% of Kiva loans during that month.

Second, the effect of the event on funding time is now bro-ken down into two cases. For North American projects (i.e.,ROW = 0, excluding United States ones), estimated at theaverage of competition over the period (716), funding timesare faster after the event, by 96%. This is in line with economicintuition: the increase in prospective lender traffic generated bythe media exposure led to a reduction in funding time. How-ever, for projects from the rest of the world, whether the eventincreased or decreased funding time depends on the size ofcompetition. For example, for projects facing fewer than 588competitors, the event had no effect or even increased fundingtime. This suggests that the increase in lenders following by theevent did not benefit all projects. Moreover, before the event,projects from outside North America funded 100% faster thanNorth American ones. After the event, this advantage in fund-ing speed was reduced to 19%.

This result may come from a specificity of the US loansevent. By contrast with the Oprah event, which advertisedKiva in general, attention was drawn to a specific subset ofprojects, those from the United States. Our results show thatnot only US loans were popular, but their introductionpointed attention to all projects in North America, therebyimproving their funding times. Meanwhile, the advantage ofprojects from the rest of the world was greatly reduced andsome were even hurt. This could be explained by the fact thatprospective Kiva lenders can sort projects by region. When theUS pilot was publicized, prospective lenders visiting the web-site were more likely to look at North American projects, ifonly out of curiosity to find out about the US loans. Theresulting list included all loans from North America, not justthe United States. Then, all loans on that list improved theirfunding time.

In other words, unlike Oprah in 2007, the publicity gener-ated by the launch of US loans did not benefit the fundingtimes of projects overall, but only a specific subset and evenreduced the existing advantage of the rest. This is especially

Table 5. The effect of the US loans launch

Dependent variable: log of funding timeOLS estimatesRobust standard errors (clustered at field partner level) reported in parentheses

(1) (2) (3)a (4)b (5)c

Competition at time posted (Ni) 0.0020*** (0.0001) 0.0020*** (0.0001)Competition from same category attime posted (mi)

0.0018*** (0.0006) 0.0038*** (0.0007) 0.0262*** (0.0064)

Other competition (ni) 0.0020*** (0.0003) 0.0016*** (0.0002) 0.0016*** (0.0001)AfterUSA = 1 after the UnitedStates loans launch, 0 otherwise

0.6175*** (0.1636) �0.1044 (0.3204) �0.0391 (0.3444) �0.1102 (0.3210) 0.0191 (0.3223)

AfterUSA � Ni �0.0011*** (0.0001) �0.0012*** (0.0001)AfterUSA � mi �0.0009** (0.0004) �0.0015*** (0.0005) �0.0017 (0.0051)AfterUSA � ni �0.0013*** (0.0002) �0.0011*** (0.0002) �0.0012*** (0.0001)USA �5.5435*** (1.2723) �5.5324*** (1.2686) �5.4968*** (1.2739) �5.4342*** (1.2362)Rest of World (excluding NorthAmerica)

�1.0085*** (0.2437) �0.9732*** (0.2977) �1.0251*** (0.2361) �0.9496*** (0.2751)

Rest of World�AfterUSA 0.8108*** (0.2993) 0.7427*** (0.3134) 0.8101*** (0.2986) 0.4844 (0.3049)Controls for project and entrepreneurcharacteristics

Yes Yes Yes Yes Yes

Sector dummies Yes Yes Yes Yes YesCountry dummies Yes No No No NoN 13,657 13,657 13,657 13,657 13,657Adjusted R2 .4587 .3566 .3567 .3578 .4022





striking given that requests from outside North America rep-resented 98% of all projects in the sub sample under consider-ation, and there were only 44 US loans in that period. In otherwords, the content of media exposure had an impact.

Finally, in columns (3), (4), and (5), we estimate Eqn. (7),distinguishing between our three different types of competi-tion. All the main results described above hold. First, compe-tition increases funding time and this effect is reduced bymedia exposure. Second, competing projects of a similar type(region, sector, or field partner) have a higher impact on fund-ing time. For example, a project’s funding time is more ad-versely affected by competitors from the same region, thanby those of a different region. Third, the advantage of non-North American projects over North American ones is re-duced after the event, confirming the difference between theeffect of the US loans launch and that of Oprah.

To sum up, the results regarding US loans confirm the intu-ition that competitive pressure is relaxed by increases in thelender side of the market generated by media exposure. How-ever, our analysis provides further evidence of the power ofmedia exposure. In this case, advertising the existence of afew dozen US projects out of thousands of others had a dra-matically different impact than the 2007 appearance on Oprah.Such forces may be harnessed for the benefit of NGOs. In par-ticular, if some types of projects are inherently harder to sell todonors, our results suggest that well targeted advertising canbe used to compensate this disadvantage.

5. CONCLUSION

This paper has provided empirical evidence of the effect ofcompetition between microfinance NGO projects on their abil-ity to raise subsidized capital from individual social investors.

Using data from the P2P microfinance platform Kiva, we findthat increases in the number of competitors have a sizable ad-verse impact on projects’ funding speed. Moreover, the effectof competition is stronger between projects with similarities,which may be perceived as close substitutes by charitable lend-ers. Meanwhile, an increase in the individual lender side of themarket helps reduce funding time and relax competitive pres-sure. The policy implications are clear. Given the existence of anegative externality of entry on the P2P development financemarket, regulating entry may improve welfare. The resultsdocument empirically several key findings in the theoretical lit-erature on NGO competition.

Then, we analyze the effect of media exposure. First, Kivafounders’ appearance on The Oprah Winfrey Show in 2007was found to improve funding speed and relax the effect ofcompetition. Second, we examine the effect of publicity gener-ated by the launch of loans to entrepreneurs based in the Uni-ted States in 2009. In this case, while the marginal impact ofcompetition was also relaxed, the improvement in averagefunding times only applies unambiguously to requests fromNorth America. We argue that this is because the event fo-cused the public’s attention on loans from this region.

These findings shed light on the way media exposure affectsNGOs’ fundraising success. Not only does publicity have asignificant impact, but its content matters. In this case, a groupof projects that initially funded slower than the rest of theworld saw its performance improve dramatically. This impliesthat for projects that donors tend to find less appealing, pub-licity can compensate this initial disadvantage.

P2P microfinance has been an innovative approach to pro-vide microfinance NGOs with capital at low cost. Our resultsshed light on the importance of competition. When NGOs facemore competitive pressure on the market for scarce subsidiesfrom donors or social investors, their incentives may be


distorted in ways detrimental to their mission of poverty alle-viation. P2P platforms have the advantage that their managerscan easily regulate entry and the mix of projects at different

points in time. Combined with the platforms’ role of promo-tion benefiting all their members, this may help offset someof the negative consequences of competition.

NOTES

1. See Morduch (1999) for a survey of the microfinance literature and adiscussion of the “microfinance promise.”

2. For more on Kiva, see http://www.kiva.org/.

3. Using a randomized controlled trial (in partnership with the NGOFINCA, one of Kiva’s field partners in Peru), Karlan and Valdivia (2011)show that business training for microfinance clients can help improveproject outcomes.

4. For surveys of the economic literature on nonprofits, see Rose-Ackerman (1996) and Andreoni (2006).

5. Ghosh and Van Tassel (2011) argue that access to cheap externalfunding allows microfinance providers to offer lower interest rates toentrepreneurs, although it may also attract inefficient providers.

6. There have been related findings and discussions in the case of for-profit firms. For example, Rosenbaum (1993) finds that profits respondpositively to entry barriers. Mankiw and Whinston (1986) derive condi-tions for inefficiently high entry of firms in some markets. Armstrong andSappington (2006) discuss the costs and benefits of policies that enhancecompetition. However, the implications of competition differ in the case ofNGOs, due to the tension they face between mission and marketobjectives. For an overview of this issue, see Young (2005).

7. A related literature discusses the effects of mergers between nonprofits(Vita & Sacher, 2001; Prufer, 2011).

8. Another related finding is in Yoruk (2009), who finds that the nationalfundraising campaign Give Five led to an increase in volunteering but nota significant increase in giving.

9. Ashta and Assadi (2010) challenge the idea that P2P developmentfunding has permitted genuine peer-to-peer contact.

10. Information about field partners is available at http://www.kiva.org/partners. The vast majority of field partners are NGOs by legal status.Others are cooperatives or non-bank financial institutions, but many ofthose are either owned by an NGO, or formerly registered as NGOs (somehave changed status due to new regulations of microfinance in theircountry).

11. Some timestamps were missing or indicated a project’s funded dateearlier than its date of posting on the database. Such inconsistencies were

present mostly among projects posted early in 2006. Some projects werestill fundraising on the last day of our sample, and thus their funding timewas not yet available.

12. A project expires if it has not been fully funded within 30 days.Funds are returned to Kiva lenders who had contributed. As of May 2011,167 projects had expired on Kiva, the vast majority after September 2009(the end of our sample). Expired projects can be found separately throughthe API but are not searchable on the website.

13. Of course, a project sees the pool of competitors grow over timeon the database. In additional estimations (available upon request), weincluded the growth rate of the pool of competitors in addition to thenumber at entry separately. It is positively associated with funding timeand more importantly, this does not change our other results.However, since our dependent variable is funding time, including thegrowth of competitors over time in the regressions introduces endo-geneity, making the coefficient on that variable difficult to interpret.Thus, for the sake of clarity, our focus is on the effect of competitionat the time of entry.

14. As in the case of our competition measure, projects that remain onthe data base longer experience a growing pool of lenders. In additionalestimations (available upon request), we checked that adding the growthin the number of lenders over a project’s fundraising time does not impactour other results. The coefficient on this additional variable is positive andlarge, which suggests endogeneity.

15. Note that the number of registered lenders would have increased evenin the absence of the Oprah feature. In additional estimations (availableupon request), we checked that the effect was stronger during the weekimmediately following the show than in the remainder of the month,which suggests that our Oprah dummy does not merely capture whatwould have happened had the feature not occurred. A similar exercise wasperformed for the US loans launch, with the same result. Of course, sinceall projects are exposed to each event after it happens, we lack a naturalexperiment setting to infer the causal effects more clearly. The comparisonbefore and after remains informative, however, especially when comparingthe two events.

16. For risk rating, the dummy variable “low risk” takes the value 1 forratings of 4 and 5 stars (the safest) and zero otherwise.

17. Prospective lenders can sort by region: Africa, Asia, NorthAmerica, Central America, South America, Eastern Europe, and theMiddle East.

REFERENCES

Aldashev, G., & Verdier, T. (2010). Goodwill bazaar: NGO competitionand giving to development. Journal of Development Economics, 91(1),48–63.

Andreoni, J. (2006). Philanthropy. In S.-C. Kolm, & J. Mercier Ythier(Eds.). Handbook on the economics of giving, reciprocity and altruism(Vol. 2). Elsevier.

Armstrong, M., & Sappington, D. E. M. (2006). Regulation, competition,and liberalization. Journal of Economic Literature, 44(2), 325–366.

Ashta, A., & Assadi, D. (2010). An analysis of European online micro-lending websites. Innovative Marketing, 6(2), 7–17.

Bishop, M., & Green, M. (2009). Philanthrocapitalism: How giving cansave the world (2nd ed.). Bloomsbury Press.

http://www.kiva.org/

http://www.kiva.org/partners

http://www.kiva.org/partners


Castaneda, M. A., Garen, J., & Thornton, J. (2008). Competition,contractibility, and the market for donors to nonprofits. Journal ofLaw, Economics and Organization, 24(1), 215–246.

Cooley, A., & Ron, J. (2002). The NGO scramble: Organizationalinsecurity and the political economy of transnational action. Interna-tional Security, 27(1), 5–39.

Cull, R., Demirguc-Kunt, A., & Morduch, J. (2007). Financial perfor-mance and outreach: A global analysis of leading microbanks.Economic Journal, 117(517), F107–F133.

Cull, R., Demirguc-Kunt, A., & Morduch, J. (2009). Microfinance meetsthe market. Journal of Economic Perspectives, 23(1), 167–192.

Desai, R. M., & Kharas, H. (2009). Do philanthropic citizens behave likegovernments? Internet-based platforms and the diffusion of internationalprivate aid. Wolfensohn Center for Development at Brookings,Working paper.

Economides, N., & Rose-Ackerman, S. (1993). Differentiated publicgoods: Privatization and optimality. In H. Ohta, & J. F. Thisse (Eds.),Does economic space matter? (pp. 111–132). London: St. Martin’sPress.

Feigenbaum, S. (1987). Competition and performance in the nonprofitsector: The case of US medical research charities. Journal of IndustrialEconomics, 35(3), 241–253.

Flannery, M. (2007). A Guinness for Kiva. Available from <http://www.socialedge.org/blogs/kiva-chronicles/archive/2007/09/06/a-gui-ness-for-kiva>.

Flannery, M. (2009). Kiva at four. Innovations: Technology, Governance,Globalization, 4(2), 31–49.

Ghosh, S., & Van Tassel, E. (2011). Microfinance and competition forexternal funding. Economics Letters, 112(2), 168–170.

Karlan, D., & Valdivia, M. (2011). Teaching entrepreneurship: Impact ofbusiness training on microfinance clients and institutions. The Reviewof Economics and Statistics, 93(2), 510–527.

Mankiw, G. N., & Whinston, M. D. (1986). Free entry and socialinefficiency. RAND Journal of Economics, 17(1), 48–58.

McIntosh, C., & Wydick, B. (2005). Competition and microfinance.Journal of Development Economics, 78(2), 271–298.

Morduch, J. (1999). The microfinance promise. Journal of EconomicLiterature, 37(4), 1569–1614.

Morduch, J. (2000). The microfinance schism. World Development, 28(4),617–629.

Nunnenkamp, P., & Ohler, H. (2010). Funding, competition and theefficiency of NGOs: An empirical analysis of non-charitable expenditureof US NGOs engaged in foreign aid. Kiel Institute for the WorldEconomy. Kiel working papers 1640.

Pestieau, P., & Sato, M. (2006). Limiting the number of charities.Universite catholique de Louvain, Center for Operations Researchand Econometrics, CORE discussion papers.

Polgreen, L., & Bajaj, V. (2010). India microcredit sector faces collapsefrom defaults. The New York Times.

Prufer, J. (2011). Competition and mergers among nonprofits. Journal ofCompetition Law and Economics, 7(1), 69–92.

Rose-Ackerman, S. (1982). Charitable giving and “excessive” fundraising.The Quarterly Journal of Economics, 97(2), 193–212.

Rose-Ackerman, S. (1996). Altruism, nonprofits, and economic theory.Journal of Economic Literature, 34(2), 701–728.

Rosenbaum, D. I. (1993). Profit, entry and changes in concentration.International Journal of Industrial Organization, 11(2), 185–203.

The Lancet. (2010). Growth of aid and the decline of humanitarianism.The Lancet, 375(9711), 253.

Van Diepen, M. (2008). Dynamics and competition in charitable giving.ERIM PhD Series in Research in Management, 159, ErasmusUniversity Rotterdam.

Vita, M. G., & Sacher, S. (2001). The competitive effects of not-for-profithospital mergers: A case study. The Journal of Industrial Economics,49(1), 63–84.

Yoruk, B. K. (2009). The effect of media on charitable giving andvolunteering: Evidence from the “Give Five” campaign. Discussionpapers 09-02, University at Albany, SUNY, Department of Econom-ics.

Young, D. (2005). Mission-market tension in managing nonprofit organi-zations. Andrew Young School of Policy Studies Research Paper SeriesNo. 06-26, Georgia State University.

competition between microfinance ngos: evidence from kiva

Documents