Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the...

46
Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara 1 Associate Professor of Finance Department of Finance Nova Southeastern University Emre Kuvvet Associate Professor of Finance Department of Finance Nova Southeastern University 1 Corresponding author. Email: [email protected], Phone: 1-800-541-6682

Transcript of Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the...

Page 1: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Peer-to-Peer Lending Across the Rural-Urban Divide

Pankaj K. Maskara1

Associate Professor of Finance

Department of Finance

Nova Southeastern University

Emre Kuvvet

Associate Professor of Finance

Department of Finance

Nova Southeastern University

1 Corresponding author. Email: [email protected], Phone: 1-800-541-6682

Page 2: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Peer-to-Peer Lending Across the Rural-Urban Divide

Abstract

This paper investigates whether P2P lending has been able to meet the needs of borrowers who

were otherwise likely left out of the market earlier. We find that localities in urban areas with lower

density of alternative lenders like pawnshops seem to have benefitted from P2P lending. We also

show that residents of localities in rural America with no or fewer credit unions participate more

in P2P lending. Overall, our results suggest that, by providing an alternative avenue to those with

limited options to borrow money, P2P lending industry appears to have completed the market for

financial intermediation.

Keywords: Peer-to-Peer Lending; Pawn Lending; Credit Unions; Rural; Urban; Access to

Credit; FinTech

JEL Classification: D53; G21; G23; G28; O16; O33; R2

Page 3: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

I. Introduction

Income inequality and opportunities gap in rural and urban areas in the U.S. have grown

larger over the years and the digital revolution appears to have exacerbated this gap. As a result,

the country now appears to be more segregated socially, economically, and politically across the

rural-urban frontier than at any other time in recent memory. The benefits of the new digital

economy have not always trickled down to those living in the countryside. In this study, we

investigate whether one aspect of the new economy, peer-to-peer lending (P2P), has benefitted

these rural areas.

Over the last 10 to 12 years, a concept called P2P lending has taken hold in the U.S. and

other countries (for example, China). In P2P lending, individuals make loans to other individuals

on a digital platform like Lending Club and Prosper. These platforms help facilitate the process

and charge the lender or the borrower (or both) for its services. In this paper, we show that P2P

lending has offered a financing alternative to at least some rural communities.

In pre-P2P lending era, borrowers had the option to borrow from financial institutions like

banks and credit unions. Those living in the urban areas often had ready access to these options.

The introduction of P2P lending in the options mix improved the choices for such people.

However, for those living in the rural areas, in the pre-P2P era, options were clearly inferior or, in

some cases, non-existent. They had to rely on banks and financial institutions located far from

their residence, if one existed. Often, such people would have to contend with borrowing from

informal channels like friends, family, and neighbors. In this paper, we show that P2P lending has

been able to meet the needs of borrowers who were otherwise likely left out of the market earlier.

In urban areas, localities with lower density of alternate lenders like pawnshops and payday lenders

seem to have benefitted from P2P lending. In rural America, localities without credit unions seem

Page 4: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

to have benefitted more from P2P lending. Either way, by providing an alternate avenue to those

with limited options to borrow money, the P2P lending industry appears to have completed the

market for financial intermediation. Our findings provide confirming evidence for some of the

conclusions of prior studies while providing fresh insight into the efficacy of P2P lending in

providing an additional financing alternative to the likely underserved sections of the society.

Studies on financial intermediation often use data on loans and other completed

transactions. This presents the classic problem of simultaneity. In studies on traditional financial

institutions like banks, credit unions, and payday lenders, aggregate lending data is observable

only at the equilibrium. Data are often limited to transactions where a loan is made or an item is

pawned. Information on loans that were attempted but not approved is often not available.

However, in our study, we use data on loan requests that include both successful and failed

attempts. In fact, 269,579 loan requests, i.e., over 40% of the loan requests in our data, went

unfulfilled. As a result of this unique feature of our dataset, we can draw stronger inferences about

the role of P2P lending in completing the market.2 P2P lending does not suffer from the locational

constraints faced by traditional financial institutions. Any individual, 18 years of age or older, with

a social security number and a bank/credit union account can apply for a loan on Prosper if he is

not a resident of Iowa, Maine, or North Dakota. Therefore, when an individual makes a loan

request on a P2P platform, it is likely to be the result of two reasons: 1) the need for funds is not

being met by existing institutions, or 2) the individual is shopping for better terms.3 We provide

2 Havrylchyk et al. (2018) use the same data but limit their analysis to 88,988 funded loans instead of 662,768 loans

requests we use in our analysis. This is due to the nature of their hypotheses that focuses more on lender decisions

(instead of borrower decisions in our case) and the choice of sample period (from January 2007 to December 2013

compared to our sample period of April 2008 to December. 2015). Wolfe and Yoo (2018) also use the same data but

limit their analysis to funded loans for similar reasons. 3 Cyree and Spurlin (2012) show that rural banking customers in the U.S. actually pay higher fees and loan rates when

a big bank competes against a rural bank in the market. Hence, it is likely that customers would resort to P2P platforms

as bank density increases in the rural areas.

Page 5: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

evidence in our study that our observed results are primarily attributable to the first reason rather

than the second. This allows us to argue that higher instances of loan request in areas with a smaller

presence of credit unions and pawnshops are a reflection of P2P lending platforms offering a fresh

alternative to an otherwise underserved market.

The paper proceeds as follows. Section II explains the background of the loan process for

the peer-to-peer lending platform, Prosper; Section III develops the research hypotheses; Section

IV explains data sources and variables; Section V presents the empirical evidence; and Section VI

discusses our conclusions.

II. Background

This study is based on data from Prosper, the first platform of its kind in the U.S. The loan

process for Prosper has evolved since the company’s platform launched in February 2006.

Originally the process was closer to the spirit of P2P lending. Prosper operated an eBay-style

auction marketplace where lenders and borrowers determined loan rates in a Dutch auction-like

system. Borrowers listed the loan amount they were requesting, between $1,000 and $35,000, and

the highest interest rate they were willing to pay.4 The interest rate was determined based on the

lowest interest rate bid that would result in full funding of the loan request. The loan was not

necessarily funded by one lender but by many lenders, each contributing an amount of choice at

or above the minimum. Listings that did not receive full funding in 14 days expired.

4 The minimum loan amount was increased to $2,000 in December 2010. Prosper allows a borrower to have a

maximum of two outstanding loans on its platform at any point of time and the aggregate outstanding principal on

such loans cannot exceed the maximum loan amount. In April 2015, Prosper became closely integrated partners with

OnDeck, a P2P platform for small businesses to raise anywhere from $5,000 to $250,000.

Page 6: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Members organized into groups based on a specific purpose or interests. Each group had a

leader who administered the group, including granting or denying access. The leader had the right

to request verification of information provided by a user and to seek additional information in the

process of vetting for a loan request. Many group leaders helped potential borrowers in writing

and designing the listing. They also had the option of endorsing the borrower and bidding on

listings they considered to be trustworthy. Due to their position in the group and primarily because

of their own efforts, group leaders often had an informational advantage that they could exploit

when making loans in the group. In order to incentivize the group leaders to create a reliable

lending platform, Prosper allowed them to charge a fee on loans closed in their group. It even

added group ratings to measure the performance of the group relative to historical default rates.

However, there is mixed evidence regarding the efficacy of group leaders. Berger and Gleisner

(2009) find that group leaders significantly improve borrower’s credit conditions by reducing

information asymmetries, especially for those with less attractive risk profile. Maier (2014) also

finds that members of groups providing verification services have higher odds of getting their loan

funded and at lower rates. On the other hand, Hildebrand, Puri, and Rocholl (2017) show that

among loans with origination fees, bids by the group leader are perceived to be a signal of good

loan resulting in lower interest rates for the borrower. However, such loans actually have higher

default rates.

Prior to July 2009, even the legal structure of lending on Prosper was closer to the spirit of

P2P lending where the lenders purchased and took assignment of the borrower loan directly while

Prosper only kept the right to service the loan. Now Prosper lenders get “member payment

dependent” notes when they make a loan on its platform. These notes are not obligations of the

borrower but of Prosper Marketplace and provide lenders limited recourse. In the event Prosper

Page 7: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

were to declare bankruptcy or become insolvent, investors in these notes may lose their investment

even if the underlying borrower continues to pay.

In December 2010, Prosper filed a new prospectus with the Securities and Exchange

Commission (SEC) and changed its business model. It started using pre-set rates instead of letting

lenders discover a price through an auction. Today, Prosper determines the rate on a loan based on

a proprietary formula evaluating each prospective borrower’s credit risk based on the borrower’s

credit report and self-reported information on her employment and annual income. Prosper

analyzes information provided by borrowers for irregularities and randomly verifies some, but not

all, information. Unlike in the past when Prosper assigned credit grade based on a borrower’s credit

score from Experian, it now assigns each borrower a grade based on two scores: a score from

consumer reporting agency and an in-house custom score calculated using historical performance

of previous borrower loans with similar characteristics. Lenders simply choose the amount of loan

they wish to fund at the rate determined by Prosper. Unlike before, borrowers have the option of

partial funding where a loan may be originated if there are enough bids to fund at least 70% of the

loan amount requested.

Prosper allocates loans to two different pools: fractional pools and whole loan pools. A

fractional pool allows retail investors with their small pool of funds to participate on the platform

alongside institutional investors with much larger asset bases. Earlier, for a two-year period (from

July 2009 to July 2011), lenders had the option of creating an automated plan that allowed them

to automatically bid on loans that met predefined criteria. Lenders could create automated plans

based on characteristics like loan amount, credit ratings, employment status, and debt-to-income

ratio, among others. During this time period, Prosper allocated loan requests to three funding

channels: 1) the active channel (much like the whole loan pool today) allowed institutional

Page 8: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

investors and high net worth individuals (HNI) to purchase 100% of the loan directly from Prosper,

2) the passive channel catered to investors with automated plans, and 3) the notes channel was for

retail investors, much like the fractional pools today.

Wei and Lin (2017) examine the effect of market mechanisms, auctions vs. posted prices,

on the probability of funding, interest rates, and defaults. They find that pre-set interest rates are

higher than contract interest rates in auctions and that loans funded via posted rates are more likely

to default. However, they also find that loans have a higher probability of being funded under the

posted prices framework (much like ‘Buy it Now’ option on eBay) than the auction framework.

P2P lending behaves much like traditional lending in many aspects. Lenders on P2P

platforms respond to economic events much like financial institutions. For example, Ramcharan

and Crowe (2013) find that when borrowing on P2P platform, homeowners in states experiencing

declining home prices, experience greater credit rationing and higher interest rates. The default

rates on P2P platforms are also correlated with factors shown to predict delinquency at traditional

financial institutions like gender, race, age, work experience, loan amount, debt to income ratio,

delinquency history, and revolving line utilization (Emekter, Tu, Jirasakuldech, & Lu, 2015; Lin,

Li, & Zheng, 2017; Pope & Sydnor, 2011; Tao, Dong, & Lin, 2017).

However, there are two components of P2P lending that are arguably unique: the social

network aspect and herding behavior of lenders. Borrowers can disclose their social media

accounts, provide friends’ endorsements, and voluntarily disclose other information on their

listings. Soft and non-standard information can especially help in the evaluation of lower-quality

borrowers (Iyer, Khwaja, Luttmer, & Shue, 2015). Voluntary disclosure of unverifiable

information has significant effect on reducing the cost of debt for the borrower (Michels, 2012),

even though its relationship with loan funding and ex-post loan quality is non-monotonic in nature

Page 9: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

(Caldieraro, Zhang, Cunha, & Shulman, 2018). Social media activities can predict borrowers’

default probability on P2P loans (Ge, Feng, Gu, & Zhang, 2017), and online borrowers are

significantly influenced by default events in their social networks. A friend’s default more than

doubles a user’s default rate (Lu, Gu, Ye, & Sheng, 2012).

Unless a loan is channeled as a whole loan, several lenders have to participate to make a

loan viable. If a request does not attract enough bids, those who bid are left with nothing but wasted

effort. Therefore, lenders have greater likelihood of bidding on an auction with more bids

(Berkovich, 2011; Herzenstein, Dholakia, & Andrews, 2011). Zhang and Liu (2012) show that

herding in P2P lending is rational and lenders learn about the creditworthiness of a borrower from

the decisions of other lenders in a sophisticated way. Ai, Chen, Chen, Mei, and Phillips (2016)

find that lenders who join a lending team benefit from the team’s recommendations and contribute

more. This invariably results in herding. Consequently, increasing market share of P2P lending

over time has accentuated investors’ herding behavior (Jiang, Ho, Yan, & Tan, 2018). Today, even

banks rely on certification by online lenders when deciding to increase access to credit for

consumers (Balyuk, 2018).

P2P lending is not with its own set of limitations. Existing literature has highlighted quite

a few of them. First and foremost, P2P lending allows for the possibility of making lending

decisions based on physical appearance, personal tastes, and stereotypes. For example, beautiful

applicants have higher odds of securing a loan on P2P platform and receive better interest rates

(Ravina, 2012) though the reverse may hold true when the borrower and lender are of the same

gender (Gonzalez & Loureiro, 2014). Jin, Fan, Dai, and Ma (2017) find that P2P lenders are more

tolerant toward attractive borrowers’ dishonest behavior. Lenders on P2P platforms make

investment decisions based on their assessment of the trustworthiness of the borrower based on

Page 10: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

photographs (Duarte, Siegel, & Young, 2012), attractiveness and positive word use (Ciuchta &

O’Toole, 2016), personal characteristics (Kgoroeadira, Burke, & Van Stel, 2018), and race (Pope

& Syndor, 2011). Evidence suggests that linguistic style, soft factors, spelling errors, text length,

positive emotion-evoking keywords, and identity claims in the text of the loan application affect

the probability of being funded even though they fail to predict default probabilities (see

Herzenstein, Sonenshein, & Dholakia (2011); Gao, Lin, & Sias (2016); Dorfleitner et al. (2016);

Nowak, Ross, & Yencha (2018)). They prefer borrowers who are culturally similar and

geographically proximate (see Burtch, Ghose, & Wattal (2014); Senney (2016); Lin &

Viswanathan (2016)).

Additionally, requesting a loan on a P2P platform requires that the requester have an

account with a bank or a credit union. This limits the capability of P2P platforms to meet the needs

of a significant portion of the society, especially the rural communities. According to FDIC report

of 2015, 26.9% of all households are unbanked or underbanked. Those in rural areas are twice as

likely to be unbanked than those in the urban areas. Many studies suggest that traditional financial

institutions, such as banks and credit unions, cannot adequately meet the borrowing needs of rural

communities. Lopez and Winker (2018) highlight the challenge of rural financial inclusion. They

argue that expanding financial inclusion in rural areas is more difficult as higher transaction risk

and contract design costs hamper the ability of rural microfinance institutions to take advantage of

economies of scale and productivity effects that support microfinance institution sustainability in

urban areas.

Lack of access to formal banking institutions in rural communities allows for alternative

financial service providers like pawnshops to fill the void. Given the vulnerabilities of borrowers

from such providers, governments, both at the federal and the state levels, have imposed significant

Page 11: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

regulatory constraints to protect the customer. Thirty-eight states have set interest rate ceilings and

10 state require pawnshops to return excess proceeds from the sale of collateral items (Shackman

& Tenney, 2006).

In addition, leading P2P lending platforms have raised the minimum credit standards

required to borrow on their marketplace. Prosper raised its minimum FICO score from 540 to 620

in July 2009 and again in 2016 to 640. Such limitations disproportionately constrain those who are

most likely to have the highest marginal utility from access to financing. At the same time, it has

become easier for lenders to participate in the P2P lending market in search of higher returns.

Prosper has even introduced secondary market for its loans so that small investors can cash out

their investments in case of unforeseen need for liquidity. Investors can also invest from their IRA

accounts and earn annual return north of 10%. In order to allow for diversification across multiple

loans, Prosper allows investors to invest as little as $25 per loan. However, not everybody can

benefit as an investor by lending on Prosper platform. Since July 2009, only residents of some

states (27 states in 2009, 32 by 2015) and Washington, D.C. were permitted to lend on the platform

even though all qualified residents of the United States, except for Iowa, Maine, and North Dakota,

are permitted to borrow.

III. Hypothesis Development

Given that borrowing on a P2P platform like Prosper requires borrowers to have a bank

account, one would expect loan requests to increase as the density of bank branches increases in a

locality. Goodstein and Rhine (2017) show that households with reasonable geographic access are

more likely to have a bank account and effects are strongest for households on the margin of bank

Page 12: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

account ownership. In addition, P2P platforms claim, and prior studies have shown, that borrowers

can get better rates on P2P platforms in comparison to banks (Bachmann et al., 2011; Demyanyk

& Kolliner, 2014). Butler, Cornaggia, and Gurun (2017) find that borrowers who reside in areas

with good access to bank financing request loans with lower interest from an online P2P lending

intermediary.5 Cyree and Spurlin (2012) show that rural banking customers in the U.S. actually

pay higher fees and loan rates when a big bank competes against a rural bank in the market. Hence,

it is likely that as bank density increases, even rural customers would resort to P2P platforms.

However, some of the recent studies have provided evidence to the contrary. For example,

Havrylchyk et al. (2018) suggest that high market concentration and high branch density may deter

the entry and expansion of P2P lending offerings. They also find that P2P lending platforms have

partly substituted for banks in counties that were affected more by the financial crisis. Wolfe and

Yoo (2018) show that high-risk online P2P loans substitute for bank loans, while low-risk loans

may be credit expansionary. Jagtiani and Lemieux (2018) find that P2P consumer loans have

penetrated areas that may be underserved by traditional banks, such as in highly concentrated

markets and areas with fewer bank branches per capital. These findings suggest that the

relationship between traditional banking and P2P lending is complex and requires investigations

across different dimensions. We will therefore tread deeper into two aspects of these markets:

other competitors in this space and population density (rural vs. urban).

For the purposes of eligibility to apply for a P2P loan, there is no difference between having

a bank account or a credit union account. Therefore, based on this factor alone, we would expect

a positive relationship between CU density and P2P loan requests. However, prior studies have

5 However, this observation does not extend to business loans. Mach, Carter, and Slattery (2014) find that small

business loans from traditional financial venues have almost half the interest rate compared to business loans on P2P

platforms. We note here that Prosper offers personal loans and not business loans.

Page 13: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

shown that credit unions charge lower interest rates on loans as compared to banks (Depken,

Hollans, & Swidler, 2010). This could mean that in localities with credit unions, the residents

would have little incentive to make P2P loan requests even if they were eligible. This would imply

a negative relationship between CU density and P2P loan requests. However, ex ante, it is not clear

if the rates offered by credit unions, though lower than those offered by banks, are also lower than

those available on P2P platforms. Pana, Vitzthum, and Willis (2013) look at the interest rates

offered by transactional website and those offered by CUs and do not find any meaningful

difference between the two. Therefore, we cannot make a general prediction about the relationship

between P2P loan requests and CU density.

However, small credit unions, especially the ones with occupational or associational single

common bond requirements are perceived to offer good service and advice to its members. Even

though Maskara and Neymotin (2018) contend that, on average, underserved individuals are no

more likely to be served by CUs than by banks. They concede that their study primarily captures

the behavior of big credit unions. Small credit unions are likely to stick to their mandate of serving

the underserved (Mohanty, 2006). The proportion of such credit unions is likely to be higher in the

rural areas. This implies that the presence of credit unions, especially in the rural communities,

where the sense of community is likely to be stronger, requests for P2P lending is likely to be

lower as compared to urban communities with similar CU density.

H1: Compared to the relationship between CU density and P2P loan requests in urban

areas, CU density is likely to impact P2P loan requests from rural area more negatively.

Page 14: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Additionally, we expect a negative relationship between pawnshop density and P2P

lending. The density of AFS providers per capita in a locality depends on demographics, labor

market conditions, creditworthiness of the people, and state-level regulatory requirements (Bos,

Carter, & Skiba, 2012; Caskey, 1991; Murhem, 2016; Prager, 2014; Shackman & Tenney, 2006).

Prior studies suggest that users of alternative financing services (AFS) from payday lenders and

pawnbrokers have a tendency to roll over (Carter, 2015). The demand for such services is price

inelastic. Avery and Samolyk (2011) show that pawnshop usage is higher in states with the highest

fee ceiling. Users of such services are habitual. They do not generally seek cheaper financing

alternatives. McKernan, Ratcliffe, and Kuehn (2013) look at the relation between state-level

alternative financial service (AFS) policies (prohibitions, price caps, and disclosures), and the use

of alternative financial service products like payday loans, auto title loans, pawn broker loans, and

refund anticipation loans. They find that prohibitions and price caps on one alternative financial

service product do not lead consumers to use other alternative financial services products. Users

of AFS are distinct from those borrowing from banks and credit unions. The purpose of loans from

AFS is also measurably different. Finally, P2P loans are unsecured, whereas collateral secures

loans from pawnshops.

Unlike banks and credit unions that help customers meet one of the requirements for

participation in the P2P market (i.e., a bank account), the pawnshops offer no such benefit.

Additionally, as the above discussion establishes, users of pawnshops do not seek funds at better

terms elsewhere. Therefore, pawnshop density is not expected to have a positive relationship with

number of P2P loan requests. On the contrary, in the absence of AFS providers, P2P provides an

alternative to seek financing. Therefore, we expect a negative relationship between the density of

pawnshops and P2P loan requests from an area.

Page 15: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

H2: People from locality with few or no pawnshops/payday lenders are more likely to

request P2P loans than otherwise.

We note that, unlike us, Havrylchyk et al. (2018) hypothesize a positive relationship

between P2P lending and payday establishments and state, “the evidence that P2P lending is

diffused in counties with a higher number of payday loan establishments is not robust across

specifications and need to be explored further (pg. 16)”.

IV. Data and Variables

Our P2P lending loan data comes from Prosper, the second-largest online P2P lending

platform in the U.S. after Lending Club. Mariotto (2016) has documented that Prosper lends to

riskier clients at higher interest rates, but lower average amounts. Given the nature of our

hypotheses that involve rural borrowers and those using services of pawnshops, Prosper loans lend

themselves better for such analysis than those from Lending Club. Our sample data are from April

15, 2008 until December 31, 2015.6 We include loan requests from all credit rating grades, income-

levels, maturities, and loan-purposes for our analysis. We eliminate observations from Iowa, North

Dakota, and Maine since, after July 2009, borrowers residing in these states are not permitted to

borrow through Prosper.

Data available from Prosper does not have information about a borrower’s exact address

or zip code. However, it has self-reported information on the borrower’s city. We use data from

unitedstateszipcodes.org (hereafter referred to as USPS data) to ascertain the zip code(s) for the

6 On April 15, 2008, Prosper became exempt from most state usury laws.

Page 16: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

borrower’s city. Some cities have 25 or more zip codes, while many small cities share a zip code

with another relatively bigger city and are therefore not reported in the USPS data.

There are over 27,500 unique values in the cities field of Prosper database for our sample

period. Given that values in this field are typed by the users, it has numerous spelling errors and

inconsistencies. Many of the unique values are logical variations of the same city like “Colorado

spgs” for “Colorado Springs.” We standardize city names to account for all logical variations.

However, we find that there still remain over 3,000 unique values that are either too unique,

typographical errors, or non-standard truncations. We analyze all the observations manually to

account for these variations and errors. We finally reconcile all observations to about 8,200 unique

values for a city name. When in doubt, we use Google to identify if a city of such a name actually

exists in a state. Even when the spellings are incorrect and illogical to us, Google can ascertain the

right city based on its vast database. As long as there are five or more observations in Prosper’s

listing data for a given city, we use Google to find all the zip codes for the city in question, even

when the USPS database does not have the required information.

Zip code information is used in our study to account for the demographic environment and

the availability of financial institutions and pawnshops in a borrower’s area of residence, among

others. In order to capture the most relevant information, we find the zip code for the smallest area

unit based on the information available. For example, if the borrower reports living in Jamaica

(Jamaica is in Queens, NYC), then we use all the zip codes for Jamaica only. However, if we only

have information that the borrower lives in Queens Borough, we use all the zip codes of Queens

to average the demographic and other variables for the larger area, in this case, the borough. When

the only information available is that the borrower lives in NYC, we average all variables at that

Page 17: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

level. We standardize all values to ensure that the use of the different units of area does not impact

the results of our study.

We remove decommissioned zip codes from the USPS database. Also excluded are zip

codes for specific entities that receive a lot of mail (for example, IRS). We exclude zip codes for

military bases because they are not free-market systems. The government provides services at

military bases that are not available in the general marketplace.

Finally, we exclude zip codes for P. O. boxes. If we were to use both the standard codes

and the P.O. box codes, then areas with boxes would be overrepresented in the analysis. Inclusion

of boxes would effectively be double counting some of the areas in the city. However, for about

100 small cities, there is no standard code available, likely because of the remoteness of the area

in question. In such cases, we use P. O. box codes.7

Our dependent variable is LoansPerMillion, the total number of loans per year for each

locality divided by its population measured in millions. Each observation in our analysis represents

a locality for a particular year. Given that our data spans from 2008 to 2015, we have up to eight

observations for each locality.

Stores, Banks, and CUs measure the total number of pawnshops, FDIC-insured bank

branches and credit unions in the locality, respectively. We get the data for the number of

pawnshops and credit unions from Infogroup. Infogroup is a big data analytics firm that provides

the data to populate geographic information systems (GIS). The data for bank branches comes

from the FDIC. StoresDensity, BanksDensity and CUsDensity measure the density of pawnshops,

7 We also perform our analysis, excluding the P. O. Box zip codes for the small cities. Exclusion of such zip codes

does not impact our results because such observations represent extremely small percentage of our data.

Page 18: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

banks, and CUs in each locality per square mile, respectively. We use lagged value of these

variables in our analysis to avoid reverse causality.

We calculate the density of population for each locality and use this information to define

rural, suburban, and urban locales. Following the U.S. Department of Defense guidelines,8 we

define Urban as a locality with more than 3,000 people per square mile. Suburban is defined as a

locality with between 1,000 and 3,000 people per square mile. Rural is defined as a locality with

fewer than 1,000 people per square mile.

We get IncomePerCapita in the borrower’s county from the U.S. Bureau of Economic

Analysis. UnemploymentRate in borrower’s county is from the Bureau of Labor Statistics and

PovertyRate, the percentage of the population living below the poverty line in the borrower county

of residence, is from the U.S. Census Bureau. We also gather the county-level mortgage

delinquency rate, MortgagesPercent90PlusLate, from the Consumer Financial Protection Bureau.

We collect information on median home values (measured in millions), percentage of

foreign-born population, and percentage of female population in the borrower’s locality. We

measure educational attainment of people living in the borrower’s locality as the percentage of

people with a high school education or less, those with a college education but no degree, and

those with a Bachelor’s Degree or higher. We also measure the percentage of people in the

borrower’s locality with ages between 15 and 29 years, and the percentage of people ages 60 or

higher. All the above-mentioned variables, including the race variables (White, Black, Asian, and

OtherRace), come from the U.S. Census Bureau. All of our analyses include state and year dummy

variables.

8 https://policy.tti.tamu.edu/theres-no-such-thing-as-the-suburbs/

Page 19: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

We get data on broadband availability from the website of the Federal Communications

Commission (FCC). This data is based on Form 477 that service providers file with the FCC.

Providers of broadband connections are required to file Form 477 if they have at least one

broadband connection in service to an end user. We collect yearly data from 2008 onwards for

number of residential broadband connections per 1,000 households in the county with a minimum

speed of 200 kbps in any direction. The variable Internet index takes a value of zero when

broadband is not available, 1 when less than 20% of the households are connected, and 2, 3, 4, and

5 when at least 20%, 40%, 60%, and 80%, respectively, of the households are connected.

Ex ante expectations for control variables

There are significant learning costs associated with P2P lending. These include search

costs, cognitive efforts, emotional costs, psychological risks, and social risks associated with the

understanding of a new business model and building trust in it (Havrylchyk et al., 2018). P2P

lending involves the use of technology and requires that users create a profile on the lending

platform and file a loan request, something that is done by a loan officer at a financial institution.

Also a significant percentage of the population does not know of the existence of this mechanism.

The diffusion of this technology depends on the network effect. As more people use this

technology, they introduce their friends and family to it thereby facilitating its use. Finally, P2P

loan requests depend on the demand for credit much like other forms of lending.

Therefore, ex ante, for the education-related variables, we expect a positive relationship

between P2P lending and education reflecting higher learning costs for those with lower levels of

educational attainment. However, given that those with advanced degrees may have relatively

Page 20: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

lower demand for credit, the relationship between education variables and loan requests may not

be monotonic. Similarly, young people have lower learning costs, but it is not apparent whether

they would have lower or higher demand for credit. We, therefore, do not have any prediction for

the sign of its coefficient ex ante. However, older people have higher learning costs and are likely

to have lower demand for credit compared to the working population. Therefore, we expect a

negative relationship between the percentage of older people in an area and the number of loan

requests from such an area. Given the positive network effect associated with population density,

we expect it to have a positive relationship with P2P loan requests.

Foreign-born people have tendency to look for alternatives and try new avenues. They

therefore face lower learning costs and are also likely to have a higher demand for credit. We

expect a positive relationship between loan requests and the percentage of the foreign-born

population in an area. We expect people living in areas with high home values to have lower

demand for the relatively small consumer credit on offer on P2P platforms. Such people can

generally borrow against their home equity line of credit, if needed. On the other hand, areas where

a larger percentage of homeowners are late on their mortgages are likely to have a higher need for

P2P-type credit.

The case for unemployment rates, per capita income, and poverty rates in an area is

relatively less clear. These variables are less transitory in nature compared to delinquency rates on

mortgages. The existing literature does not provide any theoretical priors for the sign of these

coefficients. Similarly, we do not have any expectations for the relationship between loan requests

and gender or race variables. Earlier, Prosper allowed borrowers to post pictures with their listings,

which allowed for the identification of gender and race of the applicant. But after several studies

documented discrimination against Blacks and Hispanics on its platform, Prosper disallowed

Page 21: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

posting of such identifying information. It is possible that after this change, racial minorities turned

to P2P lending in higher numbers to avoid the discrimination they face at traditional financial

institutions. Finally, given that reliable access to the Internet is a requirement for the use of the

P2P lending platform, we expect a positive relationship between P2P loan requests and broadband

availability.

V. Empirical Evidence

We present the descriptive statistics of our sample in Table 1. Our sample includes both

sparsely populated localities with population density as low as 0.003 persons per square mile and

densely populated ones with 904 persons per square mile. Sixty-four percent of the localities in

our sample are rural, while 15% are urban, and the rest are classified as suburban. The median and

average numbers of bank branches per square mile are about 0.1 and 0.5, respectively, though

some areas have as many as 32 branches per square mile. Over 75% of the localities in our data

sample do not have any pawnshops. The average income per capita is about $45,000, but some

localities have per capita income above $150,000 and a median home value of $2.0 million.

[INSERT TABLE 1 HERE]

In Table 2, we present the breakdown of the total numbers of loan requests across rural,

suburban, and urban localities. We note that the total number of loan requests are almost evenly

distributed across different types of localities, even though, as shown in Table 1, urban

observations constitute less than 15% of the 65,956 observations in the sample, while the rural

localities constitute 64% of the sample. This difference is because almost half (49%) of rural

locations do not have any P2P loan requests in a given year. Even when there are P2P loan requests

Page 22: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

from rural areas, the number of requests per location-year is just 9.3 compared to 32.6 for urban

locations. However, when we account for lower density of population in rural areas, we find the

P2P loan requests per capita in such areas is actually higher than those of suburban and urban

areas. As expected, the densities of banks, credit unions, and pawnshops are higher in more

populated areas and the average value of the Internet availability index is slightly lower for

observations without any P2P loan requests.

[INSERT TABLE 2 HERE]

In Table 3, we present the Pearson correlation coefficient for the different variables used

in the study. We find a high correlation between population density and the variables measuring

the density of financial institutions (bank density, CU density, and pawnshop density). This is

expected. We also observe a high correlation between the following pairs: mortgage delinquency

and unemployment rate; income per capita and poverty; median home value and income per capita;

Internet availability and income per capita; and poverty unemployment rate. All of these

relationships are as expected.

[INSERT TABLE 3 HERE]

To begin with, we estimate the model for P2P loan requests (measured per million

residents) as a function of population density using the OLS regression in Model 1 of Table 4. We

include demographic variables (age, education, race, and gender), economic indicators (income,

poverty, unemployment, home value, and delinquency rate), and the Internet density variable as

control variables. For control variable with theoretical priors, we find the sign and significance of

the coefficient to be consistent with our expectations. However, we did not find a significant

coefficient for the population density variable.

Page 23: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

In Model 2, we include bank density and CU density variables in the equation to observe

the impact of the presence of traditional financial institutions on P2P loan requests. We find

insignificant coefficients for both of the newly added variables. The signs of all the other control

variables remain the same. We continue to find an insignificant coefficient for population density.

In Model 3, we add a variable for pawnshop density and continue to find the same results.

These results suggest that the presence of other financial institutions and population density have

no impact on the number of P2P loan requests from an area. This is contrary to our hypotheses. In

Model 4, we substitute population density, a continuous variable with two categorical variables,

rural and suburban. We find a highly significant positive coefficient for the rural variable, while

all the other coefficients continue to maintain their signs and significance.

If taken at face value, these results suggest that, all else being equal, residents of rural areas

make more P2P loan requests and the presence of alternative sources of financing in these areas

does not impact the number of P2P loan requests. This belies our expectations.

We note, however, that by including only the areas in our data sample that had P2P loan

requests in a given year, we bias our sample towards those areas that made P2P loan requests. We

therefore reconstitute our data sample to include observations with zero P2P loans for a location-

year. As a result, our sample size increases from 38,328 location-years to 65,956.

We perform the same analysis as before with the new sample dataset. In unreported results,

we find that the rural variable loses its significance while bank density and pawnshop density

variables become highly significant. The sign and significance of control variables continue to

meet expectations. We note that use of OLS regression on the new dataset is highly questionable

Page 24: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

due to censoring of the dependent variable at zero. To address this concern we next estimate a

Tobit model.

[INSERT TABLE 4 HERE]

We present the results of our Tobit model estimation in Model 1 of Table 5. We find highly

significant positive coefficient for the population density variable as expected. We also find a

highly significant positive coefficient for the CU density variable, suggesting that number of P2P

loan requests from an area increases as the density of credit unions increases. The coefficient for

the bank density variable is positive but insignificant. On the other hand, the pawnshop density

variable has a highly significant negative coefficient, suggesting that people living in areas with

fewer pawnshops are more likely to apply for P2P loans. This provides support for our second

hypothesis.

To understand the different mechanisms at play affecting the number of P2P loan requests

from an area, we dissect our loan sample into rural, suburban, and urban subsamples. In Models

2, 3, and 4 of Table 5, we limit our sample to rural, urban, and suburban areas only, respectively.

The estimates of our Tobit model provide confirming evidence for our hypotheses. We find a

highly significant negative coefficient for the CU density variable in the rural subsample, but a

highly significant positive coefficient for the urban subsample and an insignificant negative

coefficient for suburban subsample. In addition, we find a negative coefficient for the pawnshop

density variable in all the subsamples, even though it is highly significant only in the urban

subsample.

The results suggest that when a rural area has a credit union, its residents feel a lesser need

to apply for a loan on P2P lending platforms. However, in urban areas, the credit unions seem to

Page 25: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

behave much like banks as Maskara and Neymotin (2018) contend. Our results suggest that an

addition of one credit union in an urban area would increase the number of loan requests in the

one square mile area of the credit union by about 10%. The mean number of P2P loan requests for

urban areas is 421 requests per million residents (see Table 2) and the coefficient of CU density

per square mile variable is about 42 (Model 3, Table 5).

On the contrary, an addition of one CU in a rural area decreases P2P loan requests from

the residents in one-square-mile areas of the CU by about 60%. The numbers presented herewith

are just provided for perspective and to contrast the differential impact of CU density on P2P loans

in rural and urban areas in support of our hypothesis. We realize that, in reality, the impact of the

presence of a CU would not be limited to one square mile, but would spread over a much larger

area.

[INSERT TABLE 5 HERE]

We have argued that people make P2P loan requests either because they do not have other

alternatives to obtain financing or because they are hunting for better terms. If our contention is

true, the number of requests that actually get funded would have to differ across areas that have

significant number of banks and those that do not. If some of the applicants for P2P loans are just

shopping for better terms, we should see fewer of these loan requests funded from those areas.

Also, pawnshop density in an area should not impact the percentage of loan requests completed

for an area. We have contended that users of pawnshops are materially different from the general

population and such users are unlikely to make P2P loan requests in search of better terms.

To test the validity of our arguments, we next estimate completed loans as a percentage of

total loan requests from an area. In Model 1 of Table 6, we present the results of the Tobit model

Page 26: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

censored from below at zero and from above at one. We find significant negative coefficients for

the bank density variable. This suggests that fewer loan requests made from areas with high bank

density get completed. In addition, we find an insignificant coefficient for pawnshop density.

These results are in line with our assertions.

[INSERT TABLE 6 HERE]

There is a possibility that our findings about the different impacts of CU density and

pawnshop density across rural and urban areas is because the credit risk of people requesting P2P

loans is materially different across this dimension. To check for this possibility, we sort the loan

requests across different risk categories for rural, urban, and suburban subsamples. As shown in

Table 7, we find that the distribution of loan requests across risk levels is similar for rural, urban,

and suburban areas. The benefit of better interests on loans made on Prosper platform is very

limited for low-risk borrowers. Prosper loans typically have small dollar value and maturity period

of no more than 3 years. It would be relatively more beneficial for high-risk borrowers to shop for

better terms on P2P platforms. This is because the compounded total benefit from lower interest

rates will be more material for such borrowers.

[INSERT TABLE 7 HERE]

In Model 2 of Table 6, we estimate the ratio of completed loans to loan requests using the

Tobit model for high-risk loans only. These loans are defined as those with a credit grade of D, E,

HR, or none. We continue to find a highly significant negative coefficient for banks’ density

variable and an insignificant coefficient for pawnshop density. The results remain the same when

we limit our analysis to the rural subsample or the urban subsample (Models 3 and 4 in Table 6).

In unreported results we test the robustness our findings using varying definition of high risk

Page 27: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

borrower. We find that our results continue to be the same after the exclusion of observations with

no credit grade from the high-risk subsample or the inclusion of C credit grade loans in the

subsample.

VI. Conclusion

In this paper, we provide fresh insight into the efficacy of P2P lending in providing

additional financing alternative to likely underserved sections of the society. We suggest that

higher instances of loan request in areas with smaller presence of credit unions and pawnshops is

a reflection of P2P lending platforms offering a fresh alternative to otherwise underserved market.

However, there is a caveat for our finding that P2P lending is expanding the financial inclusion in

underserved communities. Even though P2P lending is unique in many ways, it is still about a

person lending money to another person. People have been borrowing money from friends and

family for ages, long before the Internet and P2P platforms came into existence. Hence, it is still

possible that we are capturing some of the borrowing and lending activity that now takes place in

a P2P setting but, if P2P did not exist, would have still taken place among friends. There is some

evidence to suggest that friends of borrowers are more likely to offer loans and at better terms than

strangers in an online P2P lending website (Lin, Prabhala, & Viswanathan, 2013; De, Brass, Lu,

& Chen, 2015), even though it is not clear whether existing friendships are related positively or

negatively with defaults. Though Lin et al. (2013) find fewer defaults among friends, Freedman

and Jin (2017) find the opposite.

We also recognize the limits of P2P lending. First, almost 11% of Americans still do not

use the Internet (Anderson, Perrin, & Jiang, 2018), especially those living in the rural areas who

Page 28: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

are twice as likely to be non-users of the Internet as are their urban counterparts. This implies that

P2P lending cannot reach a substantial portion of the population that suffers from a lack of access

to financial institutions. Second, P2P lending marketplaces have been increasing their credit

standards over time. Today, one has to have a minimum credit rating of 640 to apply for a loan on

the Prosper platform. Such eligibility standards effectively preclude a significant portion of the

population from participating in this market.

Page 29: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

References

Ai, W., Chen, R., Chen, Y., Mei, Q., & Phillips, W. (2016). Recommending teams promotes

prosocial lending in online microfinance. Proceedings of the National Academy of Sciences

of the United States of America, 113(52), pp. 14944-14948.

Anderson, M., Perrin, A., & J. Jiang (2018). 11% of Americans don’t use the Internet. Who are

they? Pew Research Center, Fact Tank News in the Numbers, March 5, 2018.

Avery, R.B., & Samolyk, K.A. (2011). Payday loans versus pawns shops: The effects of loan fee

limits on household use. Working Paper.

Bachmann, A., Becker, A., Buerckner, D., Hilker, M., Kock, F., Lehmann, M., & Funk, B. (2011).

Online peer-to-peer lending: A literature review. Journal of Internet Banking and

Commerce, 16(2), pp. 2-18.

Balyuk, T. (2018). Financial innovation and borrowers: Evidence from peer-to-peer lending.

Rotman School of Management, Working Paper No. 2802220.

Berger, S. C., & Gleisner, F. (2009). Emergence of financial intermediaries in electronic markets:

The case of online P2P lending. Verband der Hochschullehrer fur Betriebswirtschaft, 2(1),

pp. 39-65.

Berkovich, E. (2011). Search and herding effects in peer-to-peer lending: Evidence from

Prosper.com. Annals of Finance, 7(3), pp. 389-405.

Bos, M., Carter, S. P., & Skiba, P.M. (2012). The pawn industry and its customers: The United

States and Europe. Vanderbilt University Law School Working Paper.

Burtch, G., Ghose, A., & Wattal, S. (2014). Cultural differences and geography as determinants of

online prosocial lending. MIS Quarterly: Management Information Systems, 38(3), pp.

773-794.

Page 30: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Butler, A. W., Cornaggia, J., & Gurun, U. G. (2017). Do local capital market conditions affect

consumers’ borrowing decisions? Management Science, 63(17), pp. 4175-4187.

Caldieraro, F., Zhang, J. Z., Cunha, M., & Shulman, J. D. (2018). Strategic information

transmission in peer-to-peer lending markets. Journal of Marketing, 82(2), pp. 42-63.

Carter, S. P. (2015). Payday loan and pawnshop usage: The impact of allowing payday loan

rollovers. Journal of Consumer Affairs, 49, 436-456.

Caskey, J. P. (1991). Pawnbroking in America: The economics of a forgotten credit market.

Journal of Money, Credit, and Banking, 23(1), pp. 85-99.

Ciuchta, M. P., & O’Toole, J. (2016). Looks and linguistics: Impression formation in online

exchange marketplaces. Journal of Social Psychology, 156(6), pp. 648-663.

Cyree, K. B., & Spurlin, W. P. (2012). The effect of big-bank presence on the profit efficiency of

small banks in rural markets. Journal of Banking and Finance, 36(9), pp. 2593-2603.

De, L., Brass, D. J., Lu, Y., and Chen, D. (2015). Friendship in online peer-to-peer lending: Pipes,

prisms, and relational herding. MIS Quarterly: Management Information Systems, 39(3),

pp. 729-742.

Demyanyk, Y. & Kolliner, D., 2014. Peer-to-peer lending is posted to grow, Federal Reserve Bank

of Cleveland Working Paper.

Depken, C. A., Hollans, H., & Swidler, S. (2010). Do tax benefits conferred to Sub-S banks affect

their deposit or loan rates? Financial Research Letters, 7(4), pp. 238-245.

Dorfleitner, G., Priberny, C., Schuster, S., Stoiber, J., Weber, M., Castro, I., & Kammler, J. (2016).

Description-text related soft information in peer-to-peer lending: Evidence from two

leading European platforms. Journal of Banking and Finance, 64, pp. 169-187.

Page 31: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Duarte, J., Siegel, S., & Young, L. (2012). Trust and credit: The role of appearance in peer-to-peer

lending. Review of Financial Studies, 25(8), pp. 2455-2483.

Emekter, R., Tu, Y., Jirasakuldech, B., & Lu, M. (2015). Evaluating credit risk and loan

performance in online Peer-to-Peer (P2P) lending. Applied Economics, 47(1), 54-70.

Freedman, S., & Jin, G. Z. (2017). The information value of online social networks: Lessons from

peer-to-peer lending. International Journal of Industrial Organization, 51(C), 185-222.

Herzenstein, M., Dholakia, U. M., & Andrews, R. L. (2011). Strategic herding behavior in peer-

to-peer loan auctions. Journal of Interactive Marketing, 25, pp. 27-36.

Herzenstein, M., Sonenshein, S., & Dholakia, U. M. (2011). Tell me a good story and I may lend

you money: The role of narratives in peer-to-peer lending decisions. Journal of Marketing

Research, 48, pp. S138-S149.

Hildebrand, T., Puri, M., & Rocholl, J. (2017). Adverse incentives in crowdfunding. Management

Science, 63(3), 587-608.

Gao, Q., Lin, M., & Sias, R. W. (2016). Words matter: The role of texts in online credit markets.

Working Paper.

Ge, R., Feng, J., Gu, B., & Zhang, P. (2017). Predicting and deterring default with social media

information in peer-to-peer lending. Journal of Management Information Systems, 34(2),

pp. 401-424.

Goenner, C. F. (2016). The policy impact of new rules for loan participation on credit union

returns. Journal of Banking and Finance, 73, pp. 198-210.

Gonzalez, L., & Loureiro, Y. K. (2014). When can a photo increase credit? The impact of lender

and borrower profiles on online peer-to-peer loans. Journal of Behavioral and

Experimental Finance, 2, pp. 44-58.

Page 32: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Goodstein R. M., & Rhine, S. L. W. (2017). The effects of bank and nonbank provider locations

on household use of financial transaction services. Journal of Banking and Finance, 78,

91-107.

Havrylchyk, O., Mariotto, C., Rahim, T., & Verdier, M. (2018). What has driven the expansion of

the peer-to-peer lending? Working Paper.

Iyer, R., Kwaja, A. I., Luttmer, E. F. P., & Shue, K. (2015). Screening peers softly: Inferring the

quality of small borrowers. Management Science, 62(6), pp. 1554-1577.

Jagtiani, J., & Lemieux, C. (2018). Do fintech lenders penetrate areas that are underserved by

traditional banks? Journal of Economics and Business, Forthcoming.

Jiang, Y., Ho, Y. C., Yan, X., & Tan, Y. (2018). Investors platform choice: Herding, platform

attributes, and regulations. Journal of Management Information Systems, 35(1), pp. 86-

116.

Jin, J., Fan, B., Dai, S., & Ma, Q. (2017). Beauty premium: Event-related potentials evidence of

how physical attractiveness matters in online peer-to-peer lending. Neuroscience Letters,

640, pp. 130-135.

Kgoroeadira, R., Burke, A., & Van Stel, A. (2018). Small business online loan crowdfunding: Who

gets funded and what determines the rate of interest? Small Business Economics,

Forthcoming.

Lin, M., Prabhala, N. R., & Viswanathan, S. (2013). Judging borrowers by the company they keep:

Friendship networks and information asymmetry in online peer-to-peer lending.

Management Science, 59(1), 17-35.

Lin, M., Viswanathan, S. (2016). Home bias in online investments: An empirical study of an online

crowdfunding market. Management Science, 62(5), pp. 1393-1414.

Page 33: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Lin, X., Li, X., & Zheng, Z. (2017). Evaluating borrower’s default risk in peer-to-peer lending:

Evidence from a lending platform in China. Applied Economics, 49(35), pp. 3538-3545.

Lopez, T., & Winker, A. (2018). The challenge of rural financial inclusion-evidence from

microfinance. Applied Economics, 50(14), pp. 1555-1577.

Lu, Y., Gu, B., Ye, Q., & Sheng, Z. (2012). Social influence and defaults in peer-to-peer lending

networks. Working Paper.

Mach, T. L., Carter, C. M., & Slattery, C. R. (2014). Peer-to-peer lending to small businesses.

Federal Reserve Working Paper.

Maier, M. S. (2014). Lending to strangers: Does verification matter? University of Alberta

Working Paper.

Mariotto, C. (2016). Competition for lending in the internet era: The case of peer-to-peer lending

marketplaces in the USA. Working Paper.

Maskara, P., & Neymotin, F. (2018). Do credit unions serve the underserved: Evidence from the

consumer finance survey, Working paper.

McKernan, S. M., Ratcliffe, C., & Kuehn, D. (2013). Prohibitions, price caps, and disclosures: A

look at state policies and alternative financial product use. Journal of Economic Behavior

and Organization, 95, pp. 207-223.

Michels, J. (2012). Do unverifiable disclosures matter? Evidence from peer-to-peer lending.

Accounting Review, 87(4), pp. 1385-1413.

Mohanty, S. K. (2006). Comparing credit unions with commercial banks: Implications for public

policy. Journal of Commercial Banking and Finance, 5(2), pp. 97-113.

Murhem, S. (2016). Credit for the poor: The decline of pawnbroking 1880-1930. European Review

of Economic History, 20(2), pp. 198-214.

Page 34: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Nowak, A., Ross, A., & Yencha, C. (2018). Small business borrowing and peer-to-peer lending:

Evidence from Lending Club. Contemporary Economic Policy, 36(2), pp. 318-336.

Pana, E., Vitzthum, S., & Willis, D. (2013). The impact of internet-based services on credit unions:

A propensity score matching approach. Review of Quantitative Finance and Accounting,

44(2), pp. 329-352.

Pope, D. G., & Sydnor, J. R. (2011). What is in a picture? Evidence of discrimination from

Prosper.com. Journal of Human Resources, 46(1), pp. 53-92.

Prager, R. A. (2014). Determinants of the locations of alternative financial service providers.

Review of Industrial Organization, 45(1), pp. 21-38.

Ramcharan, R., & Crowe, C. (2013). The impact of house prices on consumer credit: Evidence

from an internet bank. Journal of Money, Credit, and Banking, 45(6), pp. 1085-1115.

Ravina, E. (2012). “Love & loans: The effect of beauty and personal characteristics in credit

markets.” Working Paper.

Senney, G. T. (2016). The geography of bidder behavior in peer-to-peer lending markets. Working

Paper.

Shackman, J. D., & Tenney, G. (2006). The effects of government regulations on the supply of

pawn loans: Evidence from 51 jurisdictions in the U.S. Journal of Financial Services

Research, 30, pp. 69-91.

Tao, Q., Dong, Y., & Lin, Z. (2017). Who can get money? Evidence from the Chinese peer-to-

peer lending platform. Information Systems Frontiers, 19(3), pp. 425-441.

Wei, Z., & Lin, M. (2017). Market mechanisms in online peer-to-peer lending. Management

Science, 66(12), pp. 4236-4257.

Page 35: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Wolfe, B. & Yoo, W. (2018). Crowding out banks: Credit substitution by peer-to-peer lending.

Working Paper.

Wolken, J. D., & Navratil, F. J. (1980). Economies of scale in credit unions: Further evidence.

Journal of Finance, 35(3), pp. 769-775.

Zhang, J., & Liu, P. (2012). Rational herding in microloan markets. Management Science, 58(5),

pp. 892-912.

Page 36: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Appendix – Variable definitions

Variable Definition

Population density The number of people per square mile for each locality

Rural The locality with fewer than 1,000 people per square mile

Suburban The locality with between 1,000 and 3,000 people per square mile

Urban The locality with more than 3,000 people per square mile

Pawnshop density The density of pawnshops in each locality per square mile

Banks density The density of banks in each locality per square mile

CU density The density of credit unions in each locality per square mile

Foreign-born population (%) The percentage of foreign-born population in each locality

Female population (%) The percentage of female population in each locality

Age 15-29 population (%) The percentage of people with age between 15 and 29 in each locality

Age >60 population (%) The percentage of people with age 60 or higher in each locality

White population (%) The percentage of Whites in each locality

Hispanic population (%) The percentage of Hispanics in each locality

Black population (%) The percentage of Blacks in each locality

Asian population (%) The percentage of Asians in each locality

Other races population (%) The percentage of Other races in each locality

Late on mortgage (%) The percentage of delinquency rate in each locality

Income per capita (‘000) The income per capita in each locality in thousands

Unemployment rate (%) The percentage of people unemployed in each locality

Poverty rate (%)

The percentage of the population living below the poverty line in each

locality

Median home value ($ mil) The median home value in millions

Internet index

The index variable that takes a value of zero when broadband is not

available, 1 when less than 20% of the households are connected, 2, 3, 4,

and 5, respectively, when at least 20%, 40%, 60%, and 80% of the

households are connected.

Not high school graduate The percentage of people with no high school education in each locality

High school or some college

The percentage of people with high school education or some college in

each locality

College graduate or higher The percentage of people with bachelor’s degree or higher in each locality

P2P loan density P2P loan requests per million residents in each locality

Page 37: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Table 1 - Descriptive statistics

Table 1 presents the descriptive statistics of our sample. Variable definitions are provided in Appendix.

Variable N Mean Minimum 10th Pctl 25th Pctl Median 75th Pctl 95th Pctl Maximum

Population density 65956 15.780 0.003 0.481 1.272 4.773 17.930 62.033 903.597

Rural 65956 0.640 – – – 1.000 1.000 1.000 1.000

Suburban 65956 0.211 – – – – – 1.000 1.000

Urban 65956 0.149 – – – – – 1.000 1.000

Pawnshop density 65956 0.031 – – – – – 0.165 7.692

Banks density 65956 0.464 – – 0.017 0.113 0.519 1.956 32.318

CU density 65956 0.077 – – – – 0.069 0.394 5.854

Foreign-born population (%) 65956 9.534 – 1.000 2.500 6.000 12.650 32.600 72.900

Female population (%) 65956 50.561 17.800 48.800 49.800 50.700 51.600 53.200 65.300

Age 15-29 population (%) 65956 18.294 0.800 13.800 15.700 17.600 20.000 25.300 76.550

Age >60 population (%) 65956 20.105 0.100 13.100 16.200 19.400 22.800 31.300 88.400

White population (%) 65956 74.845 0.166 38.911 64.580 83.113 92.340 97.538 100.000

Hispanic population (%) 65956 11.434 — 0.909 2.190 5.283 13.122 44.916 99.787

Black population (%) 65956 7.371 — — 0.548 2.152 7.521 33.413 96.947

Asian population (%) 65956 3.612 — — 0.366 1.438 3.861 15.028 65.557

Other races population (%) 65956 2.737 — 0.523 1.176 2.021 3.134 6.420 93.761

Late on mortgage (%) 65956 3.022 0.100 1.200 1.750 2.600 3.800 6.600 13.000

Income per capita (‘000) 65956 45.080 20.616 32.914 36.872 42.368 50.537 69.114 153.500

Unemployment rate (%) 65956 7.491 2.400 4.700 5.700 7.200 8.800 11.900 25.500

Poverty rate (%) 65956 13.045 3.100 6.900 9.500 12.900 16.100 20.500 37.300

Median home value ($ mil) 65956 0.258 0.010 0.106 0.143 0.205 0.315 0.612 2.000

Internet index 65956 4.079 1.000 3.000 4.000 4.000 5.000 5.000 5.000

Not high school graduate 65956 10.827 — 3.400 5.500 8.800 13.700 25.800 74.600

High school or some college 65956 58.490 4.700 39.900 51.100 60.800 68.000 75.500 100.000

College graduate or higher 65956 30.683 — 12.800 18.000 27.300 40.500 63.000 93.900

P2P loan density 65956 288.025 — — — 68.122 320.596 1,369 50,000

Page 38: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Table 2 - Descriptive statistics across rural, suburban, and urban locations

Table 2 shows the breakdown of total number of loan requests across rural, suburban, and urban localities. Variable

definitions are provided in Appendix.

Number of Loans Urban Suburban Rural Total

N (Location-yrs) 7,147 9,632 21,549 38,328

Mean 32.6 23.8 9.3 17.3

Total (=N * mean) 233,317 229,129 200,322 662,768

% of Total Loans 35% 35% 30% 100%

Zero P2P loans N (Location-yrs) 2,701 4,286 20,641 27,628

Percent of Total Location-yrs 27% 31% 49% 42%

Urban Suburban Rural

P2P per million residents Non-Zero Location-yr 421 423 553

Zero P2P Lending Location-yr – – –

Pawnshop density

Non-Zero Location-yr 0.138 0.035 0.007

Zero P2P lending location-yr 0.142 0.030 0.002

Banks density

Non-Zero Location-yr 1.867 0.628 0.104

Zero P2P Lending Location-yr 1.682 0.713 0.066

CU density

Non-Zero Location-yr 0.282 0.123 0.019

Zero P2P lending location-yr 0.273 0.122 0.011

Internet

Non-Zero Location-yr 4.41 4.36 4.04

Zero P2P Lending Location-yr 4.24 4.24 3.82

Page 39: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Table 3 - Correlation matrix

Table 3 presents the Pearson correlation coefficient for the different variables used in the study. Variables are defined in Appendix. All correlations in bold are

significant at the 1% level.

1 2 3 4 5 6 7 8 9 10

1 Population density 1.00

2 Pawnshop density 0.49 1.00

3 Banks density 0.81 0.38 1.00

4 CU density 0.45 0.23 0.46 1.00

5 Late on mortgage 0.21 0.12 0.18 0.08 1.00

6 Income per capita 0.21 0.02 0.28 0.15 -0.09 1.00

7 Unemployment 0.01 0.01 -0.02 -0.02 0.56 -0.39 1.00

8 Poverty 0.01 0.08 -0.06 0.01 0.17 -0.57 0.44 1.00

9 Median home value 0.20 0.00 0.26 0.08 0.09 0.58 -0.09 -0.34 1.00

10 Internet 0.13 0.03 0.16 0.10 0.03 0.58 -0.32 -0.47 0.38 1.00

11 P2P loans per million 0.02 0.01 0.02 0.01 -0.17 0.16 -0.29 -0.05 0.03 0.22

Page 40: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Table 4 - OLS regression: Dependent variable - number of P2P Loans per million residents

Table 4 estimates OLS models for P2P loan requests. The dependent variable is the number of P2P loan requests per million residents. Our sample size in Table 4

is 38,328 locations years, as we include only such areas in the table that had P2P loan requests in a given year. Variables are defined in Appendix. All of our

analysis include state and year dummy variables. Coefficients are included with standard errors in the next column. ***, **, * denote significance at the 1%, 5%,

and 10% level, respectively.

Model 1 Model 2 Model 3 Model 4

Est. Std. Err. Est. Std. Err. Est. Std. Err. Est. Std. Err.

Intercept 905.454 227.663 *** 901.808 228.189 *** 906.153 228.223 *** 674.880 231.175 ***

Population density -0.038 0.093 -0.091 0.150 -0.042 0.156

Rural 52.374 11.670 ***

Suburban -0.312 10.380

Internet density 14.337 6.341 ** 14.610 6.344 ** 14.831 6.347 ** 20.133 6.395 ***

Pawnshop density -23.634 21.602 -20.825 20.711

Banks density 3.899 4.884 4.144 4.889 5.327 3.542

CU density -21.361 16.254 -21.270 16.254 -2.278 16.642

Income per capita '000) 0.830 0.412 ** 0.848 0.417 ** 0.835 0.417 ** 1.177 0.418 ***

Unemployment rate (%) -2.800 3.009 -2.858 3.011 -2.836 3.011 -3.092 3.011

Poverty rate (%) 2.649 1.106 ** 2.697 1.109 ** 2.667 1.109 ** 2.860 1.110 ***

Late on mortgage (%) 1.950 2.844 1.918 2.845 1.912 2.845 3.986 2.874

Median home value ($ mil) -13.448 31.180 -13.757 31.201 -14.234 31.204 -25.783 31.179

Hispanic population (%) -1.116 0.474 ** -1.114 0.474 ** -1.108 0.474 ** -0.949 0.475 **

Black population (%) -0.502 0.277 * -0.444 0.282 -0.438 0.282 -0.175 0.283

Asian population (%) -1.740 0.734 ** -1.712 0.734 ** -1.758 0.736 ** -1.378 0.738 *

Other races population (%) 1.994 1.236 2.039 1.236 * 2.027 1.237 2.188 1.236 *

High school or some college 14.240 0.844 *** 14.211 0.845 *** 14.187 0.846 *** 14.461 0.846 ***

College graduate or higher 8.754 0.710 *** 8.697 0.714 *** 8.669 0.715 *** 9.109 0.716 ***

Foreign-born population (%) 3.675 0.746 *** 3.674 0.746 *** 3.684 0.746 *** 4.041 0.743 ***

Female population (%) -16.997 1.613 *** -16.926 1.615 *** -16.944 1.615 *** -15.128 1.639 ***

Age 15-29 population (%) -10.041 0.725 *** -9.960 0.730 *** -9.970 0.730 *** -9.533 0.734 ***

Age >60 population (%) -2.147 0.600 *** -2.170 0.604 *** -2.179 0.605 *** -2.139 0.602 ***

N 38,328 38,328 38,328 38,328

Page 41: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

F-value 279.09 *** 271.47 ***

267.82 *** 265.11 ***

Adjusted R-sq 0.34 0.34 0.34 0.3407

Page 42: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Table 5 - Tobit model: Dependent variable – number of P2P loans per million residents

Table 5 estimates Tobit models for P2P loan requests. The dependent variable is the number of P2P loan requests per million residents. Our data sample in the

table includes observations with zero P2P loans for a location-year. Thus, our sample size increases from 38,328 locations years in Table 4 to 65,956 in Table 5.

In Models 2, 3, and 4 of Table 5, we limit our sample to rural, urban, and suburban areas only, respectively. Variables are defined in Appendix. All of our analysis

include state and year dummy variables. Coefficients are included with standard errors in the next column. ***, **, * denote significance at the 1%, 5%, and 10%

level, respectively.

Model 1 Model 2 Model 3 Model 4

Est. Std. Err. Est. Std. Err. Est. Std. Err. Est. Std. Err.

Intercept -1030.116 237.370 *** -643.733 301.412 ** 188.366 243.652 354.904 221.435

Population density 0.699 0.159 *** 39.061 2.513 *** 0.329 0.146 ** -1.223 0.910

Internet density 21.334 6.161 *** 8.600 8.964 -14.025 13.567 -16.759 11.186

Pawnshop density -113.493 21.646 *** -62.441 281.146 -74.399 16.911 *** -101.424 68.537

Banks density 3.101 4.956 48.552 42.018 5.489 4.407 4.363 11.005

CU density 50.588 16.450 *** -325.134 114.230 *** 42.469 14.114 *** -44.807 30.373

Income per capita (000) -0.196 0.423 -1.055 0.777 -0.250 0.637 -0.385 0.602

Unemployment rate (%) 2.439 2.816 -2.999 3.978 -5.678 6.752 6.335 5.498

Poverty rate (%) -9.855 1.086 *** -8.825 1.694 *** -5.822 2.118 *** -9.026 1.778 ***

Late on mortgage (%) 22.482 2.728 *** 19.477 4.538 *** 26.018 4.578 *** 1.070 4.518

Median home value ($ mil) -224.351 30.208 *** -252.336 53.900 *** -172.276 50.346 *** -292.593 44.342 ***

Hispanic population (%) 1.067 0.458 ** -0.350 0.770 -1.146 0.704 -0.076 0.876

Black population (%) 0.525 0.277 * 1.412 0.584 ** -1.291 0.403 *** 1.607 0.418 ***

Asian population (%) 2.041 0.756 *** -4.151 2.045 ** -2.002 0.853 ** 3.383 1.271 ***

Other races population (%) -2.487 0.994 ** -4.950 1.299 *** 3.700 3.312 -1.117 3.116

High school or some college 16.570 0.773 *** 13.808 1.149 *** 19.336 1.468 *** 16.035 1.670 ***

College graduate or higher 14.998 0.661 *** 13.551 0.994 *** 14.451 1.300 *** 13.121 1.433 ***

Foreign-born population (%) 5.920 0.749 *** 5.915 1.645 *** 7.137 0.877 *** 4.643 1.372 ***

Female population (%) 15.546 1.528 *** 12.266 2.201 *** 0.520 3.545 -4.158 2.978

Age 15-29 population (%) -0.525 0.732 0.634 1.212 -7.107 1.366 *** -4.297 1.075 ***

Age >60 population (%) -9.042 0.586 *** -9.131 0.901 *** -10.827 1.412 *** -6.025 0.897 ***

Sigma 651.355 2.440 *** 764.385 3.857 *** 465.226 3.952 *** 480.639 3.546 ***

N 65,956 42,190 9,848 13,918

Page 43: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Log likelihood (317,895) (183,887) (55,677) (75,833)

AIC 635,942 367,924 111,480 151,810

Limited to Rural Urban Suburban

Page 44: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Table 6 - Tobit model: Dependent variable – number of completed loans divided by number of all loan requests

Table 6 estimates Tobit models. The dependent variable is the completed loans as a percentage of total loan requests from an area. In Model 1 of Table 6, we

present the results of Tobit model censored from below at zero and from above at one. In Model 2 of Table 6, we estimate ratio of the completed loans to loan

requests using Tobit model for high risk loans only. High risk loans are defined as those with credit grade of D, E, HR, or no credit grade. In Models 3 and 4, we

limit our analysis to rural subsample and urban subsample, respectively. Variables are defined in Appendix. All of our analysis include state and year dummy

variables. Coefficients are included with standard errors in the next column. ***, **, * denote significance at the 1%, 5%, and 10% level, respectively.

Model 1 Model 2 Model 3 Model 4

Est. Std. Err. Est. Std. Err. Est. Std. Err. Est. Std. Err.

Intercept 54.273 17.814 *** 62.45 28.561 ** 29.537 24.063 -8.766 22.067

Population density 0.028 0.012 ** 0.049 0.02 ** 0.537 0.24 ** 0.038 0.013 ***

Internet density -0.487 0.5 -1.509 0.844 * -0.548 0.811 -1.996 1.233

Pawnshop density 2.09 1.66 0.62 2.696 34.157 24.526 0.167 1.559

Banks density -0.926 0.38 ** -1.568 0.613 ** -10.657 4.787 ** -1.173 0.403 ***

CU density 1.93 1.255 2.141 2.032 7.791 12.109 2.076 1.268

Income per capita (000) 0.079 0.032 ** 0.076 0.054 -0.031 0.068 0.131 0.055 **

Unemployment rate (%) -0.343 0.237 -0.407 0.403 -0.186 0.375 0.474 0.619

Poverty rate (%) 0.023 0.087 0.001 0.146 -0.208 0.154 -0.009 0.189

Late on mortgage (%) 0.143 0.224 1.115 0.374 *** 0.022 0.425 0.477 0.421

Median home value ($ mil) -10.093 2.431 *** -4.644 4.399 1.202 5.753 -3.251 4.579

Hispanic population (%) 0.002 0.037 0.105 0.063 * 0.057 0.078 0.143 0.061 **

Black population (%) -0.009 0.022 -0.007 0.036 0.115 0.052 ** -0.024 0.036

Asian population (%) 0.09 0.057 0.103 0.093 0.131 0.193 0.105 0.072

Other races population (%) 0.193 0.1 * 0.193 0.181 0.134 0.152 0.151 0.293

High school or some college 0.118 0.067 * 0.255 0.119 ** 0.31 0.125 ** 0.376 0.13 ***

College graduate or higher 0.154 0.056 *** 0.297 0.1 *** 0.289 0.104 *** 0.359 0.115 ***

Foreign-born population (%) -0.161 0.058 *** -0.294 0.095 *** 0.089 0.169 -0.18 0.076 **

Female population (%) -0.016 0.128 0.02 0.228 0.354 0.217 0.845 0.321 ***

Age 15-29 population (%) -0.126 0.057 ** 0.051 0.1 0.072 0.108 0.248 0.128 *

Age >60 population (%) -0.137 0.047 *** -0.134 0.083 -0.189 0.085 ** -0.097 0.13

Sigma 41.389 0.215 *** 58.181 0.421 *** 44.242 0.338 *** 32.802 0.362 ***

N 38,328 29,321 15,212 6,124

Page 45: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Log likelihood -133,316 -83,457 -55,328 -23,377 AIC 266,785 167,065 110,807 46,881 Limited to High Risk High Risk Rural High Risk Urban

Page 46: Peer-to-Peer Lending Across the Rural-Urban Divide Paper 120… · Peer-to-Peer Lending Across the Rural-Urban Divide Pankaj K. Maskara1 Associate Professor of Finance Department

Table 7 - Descriptive statistics: Number of loan requests made across different risk categories by borrowers residing in different types of locations

Table 7 shows the loan requests across different risk categories for rural, urban, and suburban subsample.

Prosper Ratings Rural Urban Suburban Total

AA 15,468 19,521 18,449 53,438

A 36,788 44,308 41,806 122,902

B 41,194 47,139 45,649 133,982

C 43,209 48,280 47,470 138,959

D 24,368 27,632 27,209 79,209

E 12,496 14,798 13,828 41,122

HR 10,208 11,807 11,864 33,879

N/A 18,238 19,554 21,403 59,195

Total 201,969 233,039 227,678 662,686

AA 7.7% 8.4% 8.1% 8.1%

A 18.2% 19.0% 18.4% 18.5%

B 20.4% 20.2% 20.0% 20.2%

C 21.4% 20.7% 20.8% 21.0%

D 12.1% 11.9% 12.0% 12.0%

E 6.2% 6.4% 6.1% 6.2%

HR 5.1% 5.1% 5.2% 5.1%

N/A 9.0% 8.4% 9.4% 8.9%

Total 100.0% 100.0% 100.0% 100.0%