Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Keywords: Despite the explosive growth of peer-to-peer lending in China, information asymmetry remains a
Voluntary information disclosure critical issue and is likely to be amplified in such an evolving credit market compared to a
Manipulation traditional credit market. This paper studies how investors screen the nonstandard and often
Information asymmetry
unverifiable information disclosed voluntarily by the borrowers to make their investment de
P2P lending
cisions. Using data from Renrendai, one of the leading P2P lending platforms in China, we find
that the amount of information disclosed voluntarily by the borrowers can significantly improve
the funding probability. The impact is even more remarkable for the borrowers with lower credit
rating. However, the loan default probability increases with the amount of disclosure, indicating
the possibility of information manipulation by the borrowers. Further investigation shows the
puzzle that lenders remain attracted by such loan listings can be explained by the higher prof
itability offered by the borrowers. These findings imply the necessity of regulation on the in
formation disclosure in the P2P lending
1. Introduction
The online peer-to-peer (P2P) lending platforms have emerged as an alternative to traditional lending institutions around the world
(Sorenson et al., 2016). Bypassing banks by capitalizing on the advance of digital technology, the online P2P lending is a particular
type of credit market in which individuals engage in lending practices. The lenders provide microloans to borrowers without collateral
and the mediation of financial intermediaries (Lin et al., 2013), facilitating access to credit for small borrowers (Paravisini et al., 2017),
while earning a higher rate of return (Duarte et al., 2012). Information asymmetry persists to be a critical and somewhat magnified
issue in such an evolving market, relative to the traditional credit market (Herzenstein et al., 2011). In the later, financial in
termediaries’ role is to evaluate and monitor borrowers’ creditworthiness and accordingly, make professional lending decisions, while
in the former the platforms act as the matchmaker refraining from conducting any function that implies financial intermediation. The
lenders make the investment decision mainly based on the standard financial information as well as nonstandard information
voluntarily disclosed by the borrowers (Iyer et al., 2016). There is a sizable literature that has extensively investigated the role of
disclosure, particularly the mandated and audited financial reports, in mitigating information asymmetry in the financial markets
* Corresponding author at: Asian Development Bank, 6 ADB Avenue, Mandaluyong City 1550, Metro Manila, Philippines.
E-mail addresses: chenxiao427@163.com (X. Chen), bihuang@adb.org (B. Huang), ms272@leicester.ac.uk (M. Shaban).
1
Present address: Asian Development Bank, 6 ADB Avenue, Mandaluyong City 1550, Metro Manila, Philippines.
https://doi.org/10.1016/j.jcorpfin.2020.101805
Received 4 October 2019; Received in revised form 2 November 2020; Accepted 19 November 2020
Available online 4 December 2020
0929-1199/© 2020 Elsevier B.V. All rights reserved.
Please cite this article as: Xiao Chen, Journal of Corporate Finance, https://doi.org/10.1016/j.jcorpfin.2020.101805
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
(Brockman et al., 2008; Brockman et al., 2010; Zhao et al., 2013; Balakrishnan et al., 2014; Chung et al., 2015; Goldstein and Yang,
2019). Nonetheless, little is known about the role of information disclosure by individuals in a peer-to-peer context. Disclosure in such
information opaque market is likely to have a remarkable impact on the investment decisions of peer lenders and indeed may shape the
future of such new but rapidly growing fintech market.
This study fills the gap in the literature by capitalizing on the opulence of the Chinese P2P lending market. We use unique data from
Renrendai, one of the leading P2P lending platforms in China, to study the voluntary disclosures by the borrowers and their impact on
market efficiency. China has developed the fastest-growing market for online P2P lending whose volume of transaction exceeded 2.8
trillion yuan (US$ 403 billion) in 2016, with an increase of 138% from a year earlier.2 Despite the explosive growth of Fintech, like in
many emerging economies, the social credit system remains underdeveloped in China. As of 2014, the People’s Bank of China
maintained credit histories for around 350 million citizens, less than one-third of the adult population while in America 89% of adults
have credit scores.3 The well-established credit system in high-income countries can provide hard and solid information to support P2P
lending. For example, Smava, the P2P platform in Germany only allows loan applications by borrowers with a specific minimum credit
score (Dorfleitner et al., 2016). On such platforms, investors rely heavily on hard information like credit scores while the effect of soft
information on the funding success and the default rate is limited. At the inception, most of the Chinese platforms did not have credit
scores for borrowers. The voluntary disclosure by the borrowers is the primary information source for investors to infer the credit
quality and make investment decisions. Under such conditions, the information asymmetry in the P2P lending market is amplified.
Therefore, it is essential to explore the various mechanisms through which the information asymmetry between borrowers and lenders
could be moderated (Strausz, 2017).
Analyzing 604,885 loan listings posted on Renrendai, we find that voluntary disclosure plays a significant role in forming lenders’
investment decision. A single item of voluntary information disclosure enhances the funding success rate by 23.6%. The impact is even
more remarkable for the borrowers with lower credit rating. We further compare the influence of verified and unverified information
on the probability of funding. Our findings show that borrowers with more verified information are more likely to get a loan, con
firming that the importance of voluntary information disclosure in shaping investment decisions.
Yet, the probability of default is also positively associated with information disclosure. An additional item of information disclosure
increases the default probability by 11.7%, suggesting the possibility of information manipulation by the borrowers. In P2P lending
market, in order to get loan request funded, the borrowers might self-select to disclose more unverified or even false information to
mimic the good-quality borrowers, while conceal the information not beneficial to them. In other words, our findings reveal a dark side
of P2P lending, confirming borrowers’ moral hazard. In a market where information asymmetry between borrowers and lenders
cannot be alleviated through financial intermediaries and hard information like credit scores are not widely available, poor-quality
borrowers tend to disclose more information to obtain funding, however probably with a premeditated intention to default. Such
manipulation of disclosure exacerbates market inefficiency arising from information asymmetry.
There are a couple of important questions that seems to impose itself in our study. Are investors sophisticated enough to infer the
real credit quality of borrowers, considering the amount and quality of information voluntarily disclosed by borrowers? Why investors
are still willing to invest in such loan requests if they are aware of the potential risks not reflected by information disclosure? The
empirical results show that it takes longer time for the loan listings with a higher level of default risk to be funded, suggesting that
investors are probably aware of real credit quality of borrowers not fully captured by voluntary disclosure. Further analysis suggests
that the investors are willing to invest in the loan requests with more voluntary information disclosure because such loan listings can
yield higher rate of return despite the possibility of information manipulation.
To infer the causal impact of disclosure on investment choice, there are a number of important endogeneity concerns need to be
addressed. First, as default depends on success, we can only observe the defaults among the borrowers who have successfully get their
loan requests funded but cannot observe defaults by those who fail to raise the fund. Hence our estimation on the default might be
susceptible to the sample selection bias. Besides, some unobservable or omitted variables may contaminate our estimation results. For
example, social network (Chen et al., 2020) and investor sentiment may change the funding success rate. We employ several empirical
strategies to address these challenges, including the Heckman selection model and 2SLS model. In particular, we employ the infor
mation disclosed by a borrower in the previous borrowing as the instrument for information disclosure in the current borrowing. The
empirical results show that our conclusions are robust across different estimations after controlling for endogeneity.
We remark that this paper contributes not only to the growing stream of research on the importance of information disclosure to
financial markets but also enriches the financial technology literature (Fuster et al., 2019; Tang, 2019; Vallée and Zeng, 2019; Berg
et al., 2020; Franks et al., 2020) and, in particular, the fintech literature strand of peer to peer lending.
Disclosure plays a vital role in improving the efficiency of financial markets. It is closely associated with stock market performance,
bid-ask spreads, cost of capital, analyst coverage and institutional ownership. Imposing minimum disclosure requirements could
attenuate the information asymmetry between informed and uninformed investors (Hirshleifer and Teoh, 2003; Ball et al., 2012;
Bertomeu and Magee, 2015). While traditional theory argues that unverifiable disclosures should be irrelevant, more and more
research acknowledge the influence of unverifiable information on investment decision (Bertrand and Morse, 2011). Nonetheless, the
role of voluntary and unverifiable information in screening credit quality of borrowers and its impact on investment decisions is still
ambiguous (Bernardo et al., 2004). Without intermediation from the financial institutions, P2P lending platforms provide a
2
http://www.chinadaily.com.cn/business/2017-01/05/content_27866083.htm
3
https://www.economist.com/news/finance-and-economics/21710292-chinas-consumer-credit-rating-culture-evolving-fastand-
unconventionally-just
2
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
decentralized and market-based mechanism that facilitates investors to screen creditworthiness of borrowers by aggregating infor
mation disclosed by borrowers. Besides the standard and hard financial information commonly used by banks, such as the borrower’s
income and credit report, lenders can view nonstandard, unverifiable and less quantifiable information, such as the maximum interest
rate the borrower is willing to pay, a textual description of purpose of borrowing, and the borrowers’ personal information like age,
employment, marriage status, living place, etc. If investors are influenced by such information, the funding probability shall increase
with the amount of disclosures. Our study provides compelling evidence in this vein by showing that voluntary and unverifiable in
formation has a significant influence on lenders’ inference of borrowers’ creditworthiness even when standard financial information
like credit scores are not available.
Our study is an indispensable addition to the limited yet growing literature on P2P lending. The existing researches on P2P lending
have investigated a wide range of mechanisms to assess the creditworthiness of borrowers. For example, using data from Prosper.com,
the leading P2P lending platform in the US, Duarte et al. (2012) claim that borrowers appearing more trustworthy are more likely to
have their borrowing requests funded. Some research shows that a borrower’s network reflects his or her creditworthiness (Lin et al.,
2013; Chen et al., 2020). A growing amount of literature has documented the role of information regarding friends and groups in the
P2P lending market, in moderating the information asymmetry between borrowers and lenders (Pope and Sydnor, 2011; Lin et al.,
2013; Liu et al., 2015; Iyer et al., 2016; Hildebrand et al., 2017; Cumming and Hornuf, 2020). Our work is closely related to but
different from Michels (2012) who confirms that the borrowers who voluntarily disclose more information are more likely to borrow at
lower rates. His study focuses on the role of borrowers’ non-voluntary disclosure while our study investigates the role of both voluntary
disclosure and unverifiable information in shaping investors’ decision. We argue that both voluntary disclosure and unverifiable in
formation could be used by borrowers as a signal of creditability, affecting lenders’ investment choices. However, in the market where
the social credit system is underdeveloped, the signals delivered by borrowers is likely to lead to a higher probability of default,
destroying investors’ confidence and making the whole industry unsustainable.
The remainder of this paper is organized as follows. Section 2 establishes the theoretical framework of this research by reviewing
relevant literature; Section 3 describes our dataset and measurement of key variables; Section 4 reports the main results; Section 5
addresses the endogeneity concerns; Section 6 summarizes the various robustness checks; and Section 7 concludes the paper with
concrete policy suggestions.
Since the seminal contributions by Akerlof (1970), Spence (1973), and Stiglitz and Weiss (1981), the link between lemon theory
and disclosure has been widely investigated in the literature (Lewis, 2011; Tadelis and Zettelmeyer, 2015). Disclosure is seen as the
primary solution to the information asymmetry that impedes the efficient allocation of resources in the capital market. Converting
inside information into public information, financial reports can effectively moderate information asymmetry (Healy and Palepu
(2001). Moreover, high-quality public disclosure leads to increased price efficiency and decreased cost of capital when information
asymmetry is high (Barron and Qu (2014)).
Early research has typically assumed that disclosures must be made truthfully and signals are costless and verifiable (Bagnoli and
Watts, 2007). However, the seminal paper by Crawford and Sobel (1982) has induced more and more researchers to explore scenarios
where the disclosures are not necessarily truthful. In the “cheap-talk” games, disclosures can even be false. Agarwal & Hauswald
(2010) argue although most information accessible to investors in traditional lending markets is nonstandard, unverifiable or “soft”,
they are valuable signals of borrowers’ creditworthiness. Gigler (1994) shows that even when disclosures are unverifiable, the cost
associated with disclosure lends it credibility. Analyzing self-reported anticorruption efforts, Healy and Serafeim (2016) conclude that
on average, firms’ disclosures signal real efforts to combat corruption. Michels (2012) claims that voluntary information disclosure can
reduce the interest rate and default rate. This argument is consistent with the theory of cheap talk and behavioral economics that
people tend to believe whatever information they can get, and it is difficult to ignore the irrelevant information in decision-making. A
large number of studies have shown that corporate voluntary information disclosure can moderate the cost of capital. Balakrishnan
et al. (2014) find that voluntary disclosure is beneficial for a firm by improving its liquidity, increasing its market value and reducing
its capital cost. At the same time, voluntary information disclosure can reduce the cost of capital, increase stock liquidity and reduce
operation’s risks (Francis et al., 2006).
In the lending market, high-quality borrowers have good reasons to voluntarily disclose more information. Yet, the impact of
information disclosure on the funding success hasn’t reached consensus yet. Wittenberg-Moerman (2008) finds that the borrower’s
timely financial report does not have a significant impact on the bid-ask spread in loan transactions, implying that more financial
information does not necessarily enhance investors’ trust in borrowers and funding success rate because some particular information
may trigger discrimination against the borrowers. In P2P lending, disclosures of some specific information are found to have significant
impacts on funding success. For example, using data from Prosper.com, the leading P2P lending platform in the US, Pope and Sydnor
(2011) find loan listings by blacks are less likely to be funded than those of whites with similar credit profiles while the interest rate
paid by blacks is higher than that by comparable whites. Employing similar data, Duarte et al. (2012) show that borrowers appearing
more trustworthy are more likely to have their borrowing requests funded. Chen et al. (2020) discover a gender gap that discriminate
against female borrowers in a Chinese P2P lending platform. These studies imply that some information like gender, race or
appearance may trigger discrimination toward borrowers, resulting lower funding probability.
It is noted that the likelihood of information manipulation is higher when voluntary disclosure is closely related to the response of
capital market. Wen (2013) find that the management might disclose information beneficial to the company if voluntary disclosure
affects its stock price. Also, the risk faced by the company affects its voluntary disclosure of information. Companies facing high
3
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
litigation risk tend to improve the quality of voluntary disclosure . In the credit market, borrowers are aware of their ability and
willingness to repay the loan. They may manipulate the disclosure in order to win the trust of lenders and acquire loans. Such
manipulation, in turn, implies a potential positive relationship between default probability and the amount of voluntary disclosure.
In a word this research is closely related to several strands of literature, including information asymmetry, voluntary disclosure and
its impact on the financial markets.
We obtain the data for this study from Renrendai, one of the largest peer-to-peer lending platforms in China. Founded in 2010, it
now has over 1 million members located in more than 2000 cities or counties across the country. Moreover, the reputation of
Renrendai has been well recognized. In 2014 and 2015, it was awarded as AAA (highest level) online lending platform by Internet
Society of China and China Academy of Social Science. It ranked No. 53 in a list of China’s top 100 internet companies released by the
Internet Society of China and the Ministry of Industry and Information in 2015.
The transactions taking place at Renrendai are typical P2P lending. Its transaction module is comparable to Prosper, the largest
lending platform in the U.S. and the main data source for most of the existing research (Duarte et al., 2012; Michels, 2012; Zhang and
Liu, 2012; Lin et al., 2013; Iyer et al., 2016; Lin and Viswanathan, 2016; Hildebrand et al., 2017). On Renrendai, borrowers can post
loan requests or listings with the required information, including loan title, borrowing amount, interest rate, description of loan usage,
and monthly installment. Renrendai only provides basic verification on borrowers’ national identification cards, credit reports, and
addresses. It assigns a credit score to each borrower according to his or her borrowing/lending history and the number of verified
information. Akin to Prosper.com, Renrendai’s profit mainly comes from borrower’s closing fee and lender’s servicing fee. Since the
verification and credit rating provided by Renrendai is limited, it is hence of critical importance for the lenders to identify the
trustworthiness of the borrowers from the observable information disclosed at the platform. In particular, when creating the loan
listings, borrowers are encouraged to disclose additional information regarding the purpose of the loan and other personal information
in a freeform text called the loan description. Once a loan listing is posted online, lenders may place bids by stating the amount they
want to fund. With a minimum bid amount of RMB 50, a list typically requires dozens of bids to be fully funded. A list that achieves
100% funding is regarded as “successful”; otherwise, the borrower receives zero funding.
This study uses all loan listings posted on Renrendai between January 1, 2011, and December 31, 2015. We eliminate the loans
listed earlier and later than this period to avoid the initial launching period and to observe the status of loan repayments, respectively.
The original sample includes 795,110 listings. We eliminate 190,225 listings guaranteed by the platform because they are not typical
P2P lending. In addition, we winsorize the loan listings whose AMOUNT and AGE are at the top or bottom one percentile of their
respective distributions to eliminate outliers. As a result, our sample includes 604,885 loan listings, of which 27,112 were successfully
funded while the rest were not funded. We track the repayment of all successful loan listings. By the end of Sep 2017, there are 4094
defaulted loans and 414 in the progress of repayment. The rest have been successfully repaid.
To gauge the effects of voluntary information disclosure on loan outcome and loan performance, we construct an information
disclosure index. There are two kinds of information disclosure at Renrendai, compulsory disclosure and voluntary disclosure. The
compulsory information consists of three categories: (1) borrowing amount, interest rate and term; (2) a borrower’s age and assets like
ownership of houses or cars; (3) loan description, corresponding title and borrower’s nickname.
There are nine items of voluntary disclosures at Renrendai, including education, employment, income, marriage, living place,
purpose of borrowing, etc. We award a point for each of them to quantify the amount of voluntary disclosure. The detailed description
of all these eight items are as follows.
Education: a borrower’s education attainment. It is classified into four levels, including high school or below, junior college,
bachelor, and postgraduate and above.
Working experience: the length of time that a borrower has been worked. It is classified into four categories of 1 year or less, 1–3
years, 3 to 5 years, and more than 5 years.
Income: a borrower’s monthly income. It is classified into seven ranks of less than RMB 1000, 1001–2000, 2001–5000,
5001–10,000, 10,001–20,000, 20,001–50,000, and more than 50,000.
Marriage: the marital status of borrowers, including divorced, widowed, single or married.
Living place: the prefecture or district (of a municipality) that a borrower is living in.
Firm size: the size of the firm that a borrower is working for. It is classified into four categories of less than 10 employees, 10–100
employees, 100–500 employees and more than 500 employees.
Purpose of borrowing: the usage of fund described by the borrowers, including short-term turnover, personal consumption, auto
loans, mortgage, wedding planning, education or training, investment, medical expenditure, home renovation, etc.
Industry: the industry that a borrower is working for, including IT, restaurant/hotel, the real estate, public utilities, public welfare
organizations, computer systems, construction, transportation, education/training, finance, law, retail/wholesale, media/advertising,
energy, agriculture, sports/arts, medical/sanitation/health care, entertainment, government agencies and manufacturing, or others.
Position: the position that a borrower has in his or her working place, like clerk, manager, etc.
4
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
We denote the above-mentioned nine items of borrowers’ voluntary disclosure as Edu_Disclosure, Worktime_Disclosure, Income_Di
sclosure, Marry_Disclosure, City_Disclosure, Firmsize_Disclosure, Purpose_Disclosure, Ind_Disclosure and Position_Disclosure respectively.
We then construct an indicator to comprehensively measure the intensity of information disclosure: DSCORE. We give a point to each
item of information disclosed in a loan list. DSCORE is the sum of the points that a loan listing is awarded for all the information
voluntarily disclosed.
In addition to the disclosure, we include two categories of control variables in the regression. The first is the information related
with loan listings, including the term, interest rate, borrowing amount, etc. The second is related with the credit risk of borrowers,
including the credit score if any, mortgage, auto loans, etc. Table 1 summarizes the definition of all variables used in this study.
Panel A of Table 2 shows the summary statistics for information disclosure. Almost all borrowers disclose their borrowing purpose
and marital status, around 70% disclose their living places, the sizes of the firms, industries they are working for and their positions.
Overall, most borrowers are willing to disclose as much personal information as possible so as to get their loan requests funded.
Panel B of Table 2 tests the mean differences between funded and unfunded listings, as well as whether loan defaults or not. In terms
of information disclosure, the mean of DSCORE for funded loan lists is 8.75, or 1.74 point significantly higher than that of unfunded
loan lists. In addition, we also find that the mean of DSCORE for defaulted loans is 8.83, or 0.1 point significantly higher than that of a
loan repaid on time. Comparing with the loan lists that don’t disclose all information, the lists with full information disclosure on
average are 5% more likely to get loan request funded and are 6% more likely to default. Their interest rates are also 0.93% higher.
Table 3 presents descriptive statistics of all variables used in this study. The average funding success rate is about 4.48%, among
which 15.1% default, implying that the competition for funding is very tough while the credit risk is high on this platform. On average,
each loan request needs about 25 bids and takes about 123 min for it to be funded successfully. A typical loan has an expected return of
19.82 yuan.
Borrowers on average voluntarily disclose about 7.09 out of 9 items of information while only 2.5 items of information are verified
by the platform. The borrowers who voluntarily disclose all information account for 63.8% of all borrowers, implying that most
borrowers are willing to disclose as much information as possible so as to transmit signals of trustworthiness to investors. The average
borrowing rate is approximately 13.36% and the average borrowing amount is RMB 59,000 (around USD 10,000), indicating the role
of P2P lending market in facilitating microfinance. The credit ratings of borrowers are universally low with an average value of 1.083.
Most borrowers are young in P2P lending market with an average age of around 32 years old. Additionally, 30.7% of borrowers own
houses and 17.8% have cars. The average length of loan title contains 13.72 Chinese characters and punctuations while the average
length of loan description is 92.49 Chinese characteristics.
Table 1
Variables and definitions
Name Definition
5
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
Table 2
Summary Statistics of Information Disclosure.
Panel A Summary Statistics of Information Disclosure
Variable Mean Sd N
Table 3
Summary Statistics.
Variable Obs Mean Std.Dev. Min Max
4. Empirical results
We first examine the extent to which information disclosure affects funding probability, default and investment process. Given the
unexpected relationship between disclosure and default, we further investigate whether investors are aware of the risks not reflected
by disclosure. Finally, we focus on solving the puzzle that lenders remain attracted by loan listings with more disclosure but higher
probability of default by assessing the profitability of such listings.
This sub-section estimates the impact of information disclosure on funding probability and default by the logit model as follow:
yi = β0 + β1 Disclosurei + β2 Controli + εi (1)
where the dependent variable is funding probability (SUCCESS) or default (DEFAULT). Both are dummy variables. SUCCESS equals one
6
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
if a borrower’s loan request is funded, and zero otherwise. DEFAULT equals 1 if a borrower defaults after the loan request is suc
cessfully funded, and 0 otherwise. Disclosure, our main variable of interest, is measured by DSCORE, the sum of all points that a loan
listing is awarded for all the information voluntarily disclosed by the borrower. We control other variables that might affect funding
probability, including the characteristics of loan listing and borrowers. εi is random disturbance term.
4
At Renrendai, a borrower’s credit rating is classified in descending order from 1 to 7, with 1 as high-risk borrower and 7 as low-risk borrower.
HR denotes the high-risk borrowers whose credit rating is as low as 1.
7
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
Table 4
Voluntary Disclosure and Funding Success.
(1) (2) (3) (4) (5)
Note: (1) This table reports Logit regression results on funding success. The Dependent variable is SUCCESS, a dummy variable taking value of 1 if a
loan listing is fully funded, and 0 otherwise. The explanatory variables include: DSCORE – the borrower’s disclosure score (see Table 2 for definition);
POOR – a dummy variable taking value of 1 if a borrower’s Credit is high risk (HR), and 0 otherwise; lnAMOUNT – natural log of loan amount(in RMB)
requested by the borrower; INTEREST – the interest rate that a borrower pays on the loan; MONTHS – loan term(in months) requested by the
borrower; CREDIT – credit grade of the borrower at the time the listing was created; AGE – the age of a borrower expressed in years; HOUSE – a
dummy variable taking value of 1 if borrower is a homeowner, and 0 otherwise; CAR – a dummy variable taking value of 1 if a borrower owns a car,
and 0 otherwise; T_Length – the number of characters in a loan title; D_Length – the number of characters in a loan description; N_Length – the number of
characters in a borrower’s nick name; and Year – Year dummy.
(2)*,**,*** indicate significance at 10%, 5%, 1% levels respectively. Robust standard errors are used and Z-statistics are reported in parentheses. N is
number of observations. r2_p is pseudo R-square.
(3) Columns (3) and (5) in show the corresponding marginal effects.
For example, the well-educated borrowers may choose to disclose his or her degree while conceal other important information on their
real risks. On the other hand, the information disclosed by the borrowers is hard to be verified. The poor-quality borrowers may self-
select to disclose false information, to mimic the good-quality borrowers to acquire loans. Such manipulation of disclosure will
exaggerate market inefficiency arising from information asymmetry.
The findings presented in the previous subsection reveal that borrowers might manipulate disclosures to acquire loans. An
important question is hence whether investors are sophisticated enough to infer the real creditworthiness of the borrowers that might
8
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
Table 5
Voluntary Disclosure and Default.
(1) (2) (3) (4) (5)
Note: (1) This table reports Logit regression results on default. The Dependent variable is DEFAULT, a dummy variable taking value of 1 if the funded
loan has been default, and 0 otherwise. The explanatory variables include: DSCORE – the borrower’s disclosure score (see Table 2 for definition);
POOR – a dummy variable taking value of 1 if a borrower’s Credit is high risk (HR), and 0 otherwise; lnAMOUNT – natural log of loan amount(in RMB)
requested by the borrower; INTEREST – the interest rate that a borrower pays on the loan; MONTHS – loan term(in months) requested by the
borrower; CREDIT – credit grade of the borrower at the time the listing was created; AGE – the age of a borrower expressed in years; HOUSE – a
dummy variable taking value of 1 if borrower is a homeowner, and 0 otherwise; CAR – a dummy variable taking value of 1 if a borrower owns a car,
and 0 otherwise; T_Length – the number of characters in a loan title; D_Length – the number of characters in a loan description; N_Length – the number of
characters in a borrower’s nick name; and Year – Year dummy.
(2)*,**,*** indicate significance at 10%, 5%, 1% levels respectively. Robust standard errors are used and Z-statistics are reported in parentheses. N is
number of observations. r2_p is pseudo R-square.
(3) Columns (3), and (5) in show the corresponding marginal effects.
not be reflected by the information disclosure. To answer this question, we assume that the same amount of disclosure corresponds to
the same level of default risk if the market is fully efficient (Fama, 1970, 1991). In other words, investors can infer the default
probability of the borrowers by the amount of information voluntarily disclosed by the borrower. However, given that borrowers may
disclose their information strategically under the premise of cheap talk, the two loans with the same amount of voluntary disclosure
may contain different level of risks. Investors hence have to infer the credit quality using the information other than voluntary dis
closures. To measure the risk of default reflected by disclosures, we first estimate the following equation:
DEFAULT i = β0 + β1 Disclosurei + εi (3)
The coefficients estimated by eq. (3) are used to predict the default risk captured by disclosures, i.e., Pr(DEFAULTi| Disclosurei).
Similarly, using the coefficients estimated by eq. (1), we can measure the default risk captured by all information observable to the
investors as Pr(DEFAULTi| Disclosurei, Controli). Therefore, the default risk that is not reflected by voluntary information disclosure can
be computed as:
defalut riski ≡ Pr(DEFAULTi |Disclosurei , Controli ) − Pr(DEFAULTi |Disclosurei ) (4)
If lenders are sophisticated enough, they shall be able to infer the default risk not revealed by the voluntary information disclosure
and make rational investment choices. We assume that the loan listings with higher level of default risk shall need longer time to get
loan funded because smart investors shall be reluctant to invest in them. This assumption can be tested by the following equation:
FundTimei = β0 + β1 default riski + β2 Controli + εi (5)
9
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
Table 6
Voluntary Disclosure and Financing Time.
(1)
FundTime_M
DSCORE 5.272**
(2.02)
lnAMOUNT 105.158***
(15.76)
INTEREST − 8.918***
(− 3.25)
MONTHS 0.826*
(1.91)
CREDIT − 15.809***
(− 4.93)
AGE 0.670
(1.24)
HOUSE − 23.281***
(− 3.28)
CAR − 21.282***
(− 2.89)
T_Length − 0.702
(− 1.37)
D_Length 0.234***
(4.50)
N_Length 4.342***
(3.57)
_cons − 974.883***
(− 13.52)
Year YES
N 27,112
r2_p 0.153
The empirical results are shown in Table 7. Column (1) and Column (2) report the coefficients estimated for eq. (3) and (1)
respectively. The Pseudo R2 of the model (1) is just 0.2%, meaning that in the P2P lending market, voluntary information disclosure
only reflects the limited amount of default risk. In Column (2), we add variables other than disclosure, including the characteristics of
loan listing, borrowers’ age, financial assets, length of loan description, etc. Pseudo R2 of the model (2) increases to 28.2%, suggesting
that information other than disclosures are important to infer the credit quality.
We further estimated the impact of default_riski on FundTime. The empirical results reported in Columns (3) show that the influence
of default_riski on the funding time is positive and significant. For a successful loan listing, a 10% increase in default_riski raises the
funding time by 72 min. Our results confirm that lenders are aware of the risks not reflected by voluntary disclosures. It should be noted
that since we don’t have hard information to directly investigate whether lenders can correctly identify the credit quality of borrowers,
this method enables us to investigate this issue in an indirect way. In the future, if more data on the borrowers are available, we can
10
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
Table 7
Voluntary Disclosure and Risk Identification.
(1) (2) (3)
default_risk 721.161**
(2.02)
DSCORE 0.174*** 0.182***
(6.84) (6.84)
lnAMOUNT 0.319*** − 125.241
(11.19) (− 1.11)
INTEREST 0.119*** − 94.612**
(10.07) (− 2.20)
MONTHS 0.058*** − 40.808**
(23.28) (− 1.97)
CREDIT − 2.100*** 1498.271**
(− 27.62) (2.00)
AGE 0.038*** − 26.603**
(12.11) (− 1.96)
HOUSE − 0.204*** 123.664*
(− 4.90) (1.72)
CAR − 0.127*** 69.978
(− 2.83) (1.51)
T_Length 0.004 − 3.379**
(1.37) (− 2.50)
D_Length 0.001*** − 0.207
(2.83) (− 0.91)
N_Length 0.009 − 1.824
(1.16) (− 0.55)
_cons − 3.259*** − 6.729*** 1527.449
(− 14.46) (− 16.34) (1.26)
Year NO YES YES
N 27,112 27,112 27,112
r2_p/a 0.002 0.282 0.153
Note: (1) This table reports Logit regression results on default in columns (1) and (2), and OLS regression results on funding time
in columns (3). The dependent variables are (i) DEFAULT, a dummy variable taking value of 1 if the funded loan has been
default, and 0 otherwise; (ii) FundTime_M, the amount of time needed for a list to be successfully funded (minutes); The
explanatory variables include: DSCORE – the borrower’s disclosure score (see Table 2 for definition); default_risk – estimated
default risk defined by eq. (8); lnAMOUNT – natural log of loan amount(in RMB) requested by the borrower; INTEREST – the
interest rate that a borrower pays on the loan; MONTHS – loan term(in months) requested by the borrower; CREDIT – credit
grade of the borrower at the time the listing was created; AGE – the age of a borrower expressed in years; HOUSE – a dummy
variable taking value of 1 if borrower is a homeowner, and 0 otherwise; CAR – a dummy variable taking value of 1 if a borrower
owns a car, and 0 otherwise; T_Length – the number of characters in a loan title; D_Length – the number of characters in a loan
description; N_Length – the number of characters in a borrower’s nick name; and Year – Year dummy.
(2)*,**,*** indicate significance at 10%, 5%, 1% levels respectively. Robust standard errors are used and Z/T-statistics are
reported in parentheses. N is number of observations. r2_a/p is adjusted R-square(pseudo R-square).
further verify to what degree this method can predict the ability of lenders to screen the creditworthiness of borrowers.
The empirical results of this paper bring a logic dilemma. Despite the higher probability of default, the loan listings with more
information disclosure actually have higher funding success rate. A possible explanation for this puzzle is the higher profitability
offered by the borrowers. To test this hypothesis, we further examine the effect of voluntary disclosure on loan performance with the
following model:
Loan Performancei = β0 + β1 Disclosurei + β2 Controli + εi (6)
We use two indicators to comprehensively gauge Loan Performancei, including interest rate and expected profit (EP). Disclosure is the
indicator counting the number of information voluntarily disclosed by a borrower. We control other variables including the charac
teristics of loan listing, borrowers’ age, financial assets, length of loan description, etc. εi is random disturbance term.
Table 8 presents the estimation on the relationship between voluntary information disclosure and interest rate. Column (2) shows
that the coefficient of DSCORE is 0.02 and significant at the 1% confidence level, implying that one additional item of voluntarily
disclosed information enhances the interest rate on average by 0.14% (0.02/13.36). Contrary to Michels (2012) who finds that dis
closures lower the funding cost, our result reflects the potential disclosure manipulation by the borrowers. In P2P lending market, the
borrowing rate directly determines the rate of return on investment. Borrowers who tend to manipulate information disclosure are
more likely to set a higher interest rate to attract the lenders.
11
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
Table 8
Voluntary Disclosure and Loan Profitability.
(1) (2) (3)
INTEREST INTEREST EP
Note: (1) This table reports OLS regression results on interest rate. The Dependent variable are (i) INTEREST, the interest rate
that the borrower pays on the loan. (ii) EP, expected profit of a loan listing; The explanatory variables include: DSCORE – the
borrower’s disclosure score (see Table 2 for definition); lnAMOUNT – natural log of loan amount(in RMB) requested by the
borrower; INTEREST – the interest rate that a borrower pays on the loan; MONTHS – loan term(in months) requested by the
borrower; CREDIT – credit grade of the borrower at the time the listing was created; AGE – the age of a borrower expressed in
years; HOUSE – a dummy variable taking value of 1 if borrower is a homeowner, and 0 otherwise; CAR – a dummy variable
taking value of 1 if a borrower owns a car, and 0 otherwise; T_Length – the number of characters in a loan title; D_Length – the
number of characters in a loan description; N_Length – the number of characters in a borrower’s nick name; and Year – Year
dummy.
(2)*,**,*** indicate significance at 10%, 5%, 1% levels respectively. Robust standard errors are used and T-statistics are re
ported in parentheses. N is number of observations. r2_a is adjusted R-square.
In addition to the interest rate, we further estimate the association of disclosure with expected profit. Assume that each loan is for
$1, and if the borrower repays the loan, the lender receives (1 + r), where r is the interest rate. This means that the lender earns a net
profit of r if the borrower repays the loan, and loses the entire dollar if the borrower fails to repay the loan. If the default probability
(DP) is δ, a lender’s expected profit (EP) on a loan listing is E[π] = (1 − δ)r − δ. To get the DP, the likelihood that a borrower defaults,
we estimate the following equation by the Probit model:
Pr(DEFAULT i ) = α0 + α1 Xi + ui (7)
where the dependent variable indicates whether a loan listing i defaults after it is successfully funded. It equals 1 if the borrower
defaults, and 0 otherwise. Xi is a vector of control variables including loan characteristics, borrower characteristics, and year fixed
effect. ui refers to the error term. The coefficients estimated from eq. (8) is then used to predict the default probability of each loan
listing. With the default probability and interest rate, we are able to measure the expected profit for each loan listing.
Columns (3) of Table 8 presents the corresponding regression results. We find that DSCORE has a significant impact on expected
profit. The coefficient of DSCORE reported in column (1) is 0.003 and significant at the 1% confidence level, meaning that the more
voluntary information disclosures, the higher the expected profit. All these results suggest that although the loan listings with more
voluntary information disclosure are more likely to default, the higher expected profit offered by the borrowers can compensate for the
risk. Therefore, lenders are still willing to invest in loan listings with more voluntary information disclosure.
5. Endogeneity concerns
In evaluating the impact of voluntary disclosure on loan outcomes, there are a couple of important methodological challenges that
need to be addressed. First, as default depends on success, we can only observe the defaults among the borrowers who have successfully
get their loan requests funded but cannot observe defaults by those who fail to raise the fund, making the estimation on default
12
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
susceptible to the sample selection bias. Second, some unobservable or omitted variables may contaminate our estimation results. For
example, social network and investor sentiment may change funding success rate. We employ the Heckman Selection Model and 2SLS
model to address these two concerns respectively.
Following Heckman (1979)’s approach, we first estimate the selection model on the probability of funding success (SUCCESS). The
Probit model is then used to treat DEFAULT as the dependent variable and other information as the independent variables. A
convincing implementation of Heckman selection model is to identify from the first stage’s selection model at least one exogenous
independent variable that can be validly excluded from the vector of explanatory variables in the second stage regression. The
transactions at the P2P platform indicate that the higher the number of bids for a loan, the more likely for it to be successfully funded.
Table 9
Estimation of Heckman Selection Model.
(1) (2)
SUCCESS DEFAULT
BIDS 0.144***
(61.59)
DSCORE 0.147*** 0.189***
(58.70) (7.05)
IMR 0.100***
(2.74)
lnAMOUNT − 0.445*** 0.351***
(− 82.00) (11.35)
INTEREST − 0.052*** 0.112***
(− 26.09) (9.31)
MONTHS 0.010*** 0.058***
(16.89) (23.39)
CREDIT 0.407*** − 2.084***
(52.69) (− 27.21)
AGE 0.015*** 0.039***
(20.70) (12.34)
HOUSE 0.124*** − 0.201***
(12.37) (− 4.83)
CAR 0.104*** − 0.122***
(8.85) (− 2.71)
T_Length 0.010*** 0.004
(15.13) (1.56)
D_Length 0.000*** 0.001***
(6.16) (2.91)
N_Length − 0.051*** 0.005
(− 32.53) (0.73)
_cons 0.631*** − 7.130***
(11.15) (− 16.08)
Year YES YES
N 604,885 27,112
r2_p 0.713 0.283
Note: (1) This table reports the Heckman two-step regression results on the probability of
default. Columns (1) is estimated by Probit regression while Columns (2) is the Heckman
two-step regression. In Column (1), the dependent variable is SUCCESS, taking a value of
1 if a loan listing is fully funded, and 0 otherwise. In Column (2), the dependent variable
is DEFAULT dummy, taking a value of 1 if the funded loan has been defaulted, and
0 otherwise. DSCORE – the borrower’s disclosure score (see Table 2 for definition). BIDS
is the number of lenders in a loan. IMR is the inverse Mills ratio. Other explanatory
variables include: lnAMOUNT – natural log of loan amount(in RMB) requested by the
borrower; INTEREST – the interest rate that a borrower pays on the loan; MONTHS – loan
term(in months) requested by the borrower; CREDIT – credit grade of the borrower at the
time the listing was created; AGE – the age of a borrower expressed in years; HOUSE – a
dummy variable taking value of 1 if borrower is a homeowner, and 0 otherwise; CAR – a
dummy variable taking value of 1 if a borrower owns a car, and 0 otherwise; T_Length –
the number of characters in a loan title; D_Length – the number of characters in a loan
description; N_Length – the number of characters in a borrower’s nick name; and Year –
Year dummy.
(2)*,**,*** indicate significance at 10%, 5%, 1% levels respectively. Robust standard
errors are used and Z-statistics are reported in parentheses. N is number of observations.
r2_p is pseudo R-square.
13
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
Given that the chance of default largely depends on the a borrower’s income or cash flow after he or she successfully gets the loan, the
number of bids should not directly affect the probability of default. We hence employ the total number of bids for a loan (BIDS) as an
instrument for model identification.
The estimation results are shown in Table 9. Column (1) reports the first stage estimation on SUCCESS. The coefficient on BIDS is
positively significant, implying that the more active the bidding process, the higher chance for a borrower to get the loan request
funded. Column (2) presents the endogeneity-adjusted estimate on default where Inverse Mill’s Ratio (IMR) estimated by the first stage
is added. The coefficient on IMR is negative and significant, indicating the existence of sample selection bias that needs to be addressed.
The coefficient on DSCORE is 0.189, significant and similar in size to the baseline estimation, corroborating that our conclusions are
robust after controlling for the sample selection bias.
The other empirical challenge to our estimation is omitted variables. For example, the borrowers who need fund urgently may
choose to disclose as much information as possible. In this paper, we employ the instrumental variable (IV) regression to address these
concerns. A suitable instrument should not directly correlate with funding success, interest rate, and probability of default, but can
directly influence the voluntary disclosure by the borrowers. Considering that a person’s financial strategy usually is formed by his or
her previous experiences and can persist for some period of time, we utilize the amount of information that a borrower disclosed in the
previous round of borrowing as the instrument for the amount of disclosure in the current borrowing. This empirical strategy has been
well acknowledged in the finance literature. For example, Bhandari and Javakhadze (2017) use the initial level of a firm’s CSR score as
an instrument. Gande and Kalpathy (2017) use the previous year’s value of CEO delta (vega) as instruments for current year’s delta
(vega). Following these literatures, we assume that a borrower’s decision on current information disclosure is largely affected by what
he or she disclosed in the previous round of borrowing. He/she may choose to disclose more information to obtain the loan if such
strategy was successful in the past. Moreover, at Renrendai, the information that a borrower disclosed is saved automatically. When the
borrower submits a new loan application, the saved information will pop up automatically, facilitating the borrower to disclose them
again. Therefore, the quantity of information disclosed in the previous borrowing will affect that in the current borrowing, but should
not directly affect the funding success, interest rate and probability of default of current borrowing. We denote this variable as
Lag_DSCORE and use it as the instrument for 2SLS regression.
Among our sample, 94,896 people borrowed more than one time and posted 293,817 loan requests at Renrendai, generating
198,921 observations for our 2SLS estimation. Table 10 reports the estimation results. The first stage regression result shown in
Column (1) indicates the information disclosed in the previous borrowing is a strong predictor for that in the current borrowing.
According to Staiger and Stock (1994), the suggested critical F-value is 8.96 when the number of instruments is one. The F-statistic of
our estimation shown at the bottom of the table is 6254, enabling us to reject the null that the coefficient on the instrument is
insignificantly different from zero and exclude the concern of weak instrument. The F-value is large in this estimation mainly due to the
large size of our sample. The second-stage regression results shown in Columns (2)–(4) are in line with the baseline estimations. The
loan listings with more voluntary information disclosure tend to have a significantly higher funding success rate, higher probability of
default and higher interest rate.
6. Robustness check
In this section, we implement several robustness checks to further prove the validity of our findings reported in the previous
sessions. For the sake of brevity, we show all the corresponding results in the supplementary.
Regarding that Probit model is also suitable for binary variable estimation and has been widely used in economic research, we re-
estimate the impact of information disclosure on funding success and probability of default with Probit model. Moreover, as requested
by the regulator, the interest rate limit set by Renrendai is “no more than four times the benchmark interest rate set by the central bank
in the same period of time”. Therefore, we also use Tobit model to re-estimate the impact of information disclosure on interest rate
whose upper bound is set to be 24%. As shown in Table A2, after controlling for other factors, borrowers’ voluntary information
disclosure is still positively and significantly related with funding success, interest rate, and the probability of default.
In this subsection, we adjust the sample to exclude the loan listings with extreme value of borrowing amount and borrowing rate.
According to the regulation5 issued by China Banking Regulatory Commission in 2016, a person cannot borrow over 200,000 yuan at
the same P2P lending platform. Moreover, the highest annual borrowing rate cannot exceed four times of annual interest rate set by the
5
On 24 August 2016, the Chinese government released the Interim Measures on Administration of the Business Activities of Peer-to-Peer Lending
Information Intermediaries to crack down illegal fundraising activities through online finance so as to prevent financial risks and potential social
unrest.
14
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
Table 10
2SLS Estimation.
(1) (2) (3) (4)
Lag_DSCORE 0.833***
(12.00)
DSCORE 0.003*** 0.016*** 0.058***
(7.15) (3.87) (12.42)
lnAMOUNT − 0.008*** − 0.031*** 0.010*** − 0.004
(− 8.37) (− 58.38) (3.18) (− 0.63)
INTEREST − 0.002* − 0.009*** − 0.002
(− 1.71) (− 43.14) (− 1.55)
MONTHS 0.004*** 0.000 0.009*** 0.058***
(32.14) (0.98) (24.43) (82.01)
CREDIT − 0.006** 0.148*** − 0.049*** − 0.707***
(− 2.20) (192.41) (− 28.43) (− 88.55)
AGE 0.000 0.003*** 0.003*** 0.003***
(0.26) (31.62) (6.78) (3.29)
HOUSE 0.062*** 0.008*** − 0.024*** − 0.137***
(2.73) (6.91) (− 4.39) (− 11.12)
CAR 0.035** 0.027*** 0.007 − 0.276***
(2.40) (20.70) (1.25) (− 20.07)
T_Length 0.001 0.001*** 0.001** 0.021***
(1.34) (14.65) (2.49) (26.62)
D_Length 0.000*** 0.000*** 0.000*** 0.001***
(5.22) (14.26) (3.46) (11.29)
N_Length − 0.003*** − 0.005*** 0.000 − 0.034***
(− 5.56) (− 30.58) (0.14) (− 18.31)
_cons 1.462** 0.226*** − 0.181*** 11.544***
(2.36) (30.96) (− 3.63) (159.54)
N 198,921 198,921 16,283 198,921
F statistics 6254
r2_a 0.811
Note: (1) This table reports 2SLS regression results. Columns (1) is estimated by OLS regression; Columns (2), (3) and (4) is estimated by 2SLS
regression. Lag_DSCORE is the amount of information disclosed by the borrower in the last loan. DSCORE – the borrower’s disclosure score (see
Table 2 for definition). Other explanatory variables include: lnAMOUNT – natural log of loan amount(in RMB) requested by the borrower; INTEREST –
the interest rate that a borrower pays on the loan; MONTHS – loan term(in months) requested by the borrower; CREDIT – credit grade of the borrower
at the time the listing was created; AGE – the age of a borrower expressed in years; HOUSE – a dummy variable taking value of 1 if borrower is a
homeowner, and 0 otherwise; CAR – a dummy variable taking value of 1 if a borrower owns a car, and 0 otherwise; T_Length – the number of
characters in a loan title; D_Length – the number of characters in a loan description; N_Length – the number of characters in a borrower’s nick name; and
Year – Year dummy.
(2)*,**,*** indicate significance at 10%, 5%, 1% levels respectively. Robust standard errors are used and T/Z-statistics are reported in parentheses. N
is number of observations. r2_a is adjusted R-square.
commercial banks. According to our calculation, the maximum borrowing rate is 24% under the current regulation framework. Thus,
in robustness tests, we exclude the loan listings that don’t match the regulation requirements. Table A3 reports the estimation results
for the loan listings whose borrowing amount are less than 200,000 yuan and borrowing rate lower than 24%. We find that voluntary
information disclosures are still positively and significantly associated with funding success, interest rate, and probability of default.
In this paper, we assume that different information provided by borrowers shall affect the loan requesting results by different
degrees. If the impacts of different types of information disclosure are the same, the lenders would be indifferent in investing in the
listings with different degree of disclosure, while the borrowers would have no incentive to manipulate the information disclosure. To
test this assumption, we examine the differentiated impacts of each type of information on loan outcomes under the condition of full
disclosure of information. The results reported in table A4-A6 indicate that different types of information do have differentiated
impacts on the funding success, interest rate, and probability of default.
To avoid estimation bias and ensure the soundness of our conclusion, we do some additional tests, but don’t report the results due to
space constraint. First, we exclude loan listings that are under repayment or defaulted from the sample, and redo empirical analysis.
Second, we divide the samples into several subgroups according to whether the borrower owns housing property or car, whether
borrowing rate, term, borrowers’ age, the lengths of borrowing title and loan description are above or lower than the median value.
Third, we re-estimate the models using the Bootstrap (100 times) standard error. All are robust and consistent with the main results.
15
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
7. Conclusions
The information asymmetry and its amplified impacts on the evolving P2P lending market where the traditional financial inter
mediation is absent motivates us to explore the role of voluntary disclosures in moderating the information gap between the borrowers
and investors. Using data from Renrendai, one of the largest peer-to-peer lending platforms in China, we first investigate the impact of
voluntary information disclosure on the loan requesting results. We find that the voluntary disclosure significantly enhances the
funding success rate. However, the loan listings with more disclosure are more likely to default. These findings suggest that the
borrowers may strategically disclose information in favor of themselves, which further escalates the information disadvantages facing
the lenders. Further empirical analyses show that although lenders are aware of the potential default risks, loan listings with more
voluntary disclosure remain attractive to them due to the higher profits offered by the borrowers.
Identifying malpractice and misconduct is the top priority of financial risk management (Alexander and Cumming, 2020). Fintech
frauds have become an oft-repeated and pronounced concern of many regulators. Cumming et al. (2020) present the first ever evidence
that fraudsters in crowdfunding markets have specific characteristics: they are less likely to have engaged in prior crowdfunding
activities or have a social media presence, but are more likely to provide poorly worded and confusing campaign pitches with a greater
number of enticements through pledge categories. Our research also confirms that borrowers could manipulate the information
disclosure and carry out fraud activities in P2P online lending market.
Our study has strong implications for both policymakers and practitioners. Despite its considerable power in promoting financial
inclusion, P2P lending has raised increasing concerns on financial stability (Nemoto et al., 2019). This market shares all the risks
associated with traditional “brick and mortar” lending, including lending fraud, identity theft, money laundering, consumer privacy
and data protection violations. These risks are then married to and amplified due to the absence of financial intermediations between
borrowers and lenders. Lax regulation has helped the industry to prosper, but as it approaches a significant size that may impact the
stability of financial system, it would be wise for regulatiors to play a more active and influential role. Fintech’s promise for financial
inclusion can only be realized if the accompanying risks are managed to maintain trust in the system and avoid a build-up of risks that
could lead to financial instability.
Globally, the existing legal framework and regulations covering P2P lending are patchy at best. However, the degree of information
asymmetry is escalated in developing countries where the credit system is under-developed. In the absence of effective credit scoring
system, any voluntary information disclosure without verification could mislead the investment choices. Therefore, the urgent task is
to extend the coverage of credit scoring system, and integrate the consequences of personal false information disclosure and default
into the credit evaluation.
In addition to establishing the credit system covering the majority of the population, the platforms shall also take measures to
control information manipulation. Before posting the borrowers’ loan applications, the platforms should verify the basic information
disclosed by the borrowers and inform the investors of the possibility of information manipulation by the borrowers. The industry
associations for P2P lending can also play an important role to alleviate the information asymmetry. On the one hand, it can promote
the sharing of borrowers’ information across the platforms, so that the whole P2P online lending industry can quickly respond to the
illegal behaviors of borrowers. On the other hand, it shall educate investors and provide them with the knowledge and tools to make
rational financial decisions.
Acknowledgements
We are grateful to Douglas Cumming (the editor) and anonymous referees for their very insightful comments. We would also like to
thank the participants at IFABS Asia 2017 China Conference and Birmingham Business School Conference on Developments in
Alternative Finance 2019 for their comments. Xiao Chen would like to thank Professor Bohui Zhang, Professor Dezhu Ye, and Professor
Rui Shen for their careful guidance and insights. Xiao Chen would like to thank the China Postdoctoral Science Foundation for its
financial support (Grant No. 2020M671852).
References
Agarwal, S., Hauswald, R., 2010. Distance and private information in lending. Review of financial studies 23, 2757–2788.
Cumming, D.J., Hornuf, L., Karami, M., Schweizer, D., 2020. Disentangling crowdfunding from fraudfunding. Annual Conference 2017 (Vienna): Alternative
Structures for Money and Banking 168120. Verein für Socialpolitik / German Economic Association.
Akerlof, G.A., 1970. The market for “lemons”: quality uncertainty and the market mechanism. Q. J. Econ. 84, 488–500.
Alexander, C., Cumming, D., 2020. Corruption and Fraud in Financial Markets: Malpractice, Misconduct and Manipulation. John Wiley & Sons.
Bagnoli, M., Watts, S.G., 2007. Financial reporting and supplemental voluntary disclosures. J. Account. Res. 45, 885–913.
Balakrishnan, K., Billings, M.B., Kelly, B., Ljungqvist, A., 2014. Shaping liquidity: on the causal effects of voluntary disclosure. J. Financ. 69, 2237–2278.
Ball, R., Jayaraman, S., Shivakumar, L., 2012. Audited financial reporting and voluntary disclosure as complements: a test of the confirmation hypothesis. Journal of
Accounting & Economics 53, 136–166.
16
X. Chen et al. Journal of Corporate Finance xxx (xxxx) xxx
Barron, O.E., Qu, H., 2014. Information asymmetry and the ex ante impact of public disclosure quality on Price efficiency and the cost of capital: evidence from a
laboratory market. Account. Rev. 89, 1269–1297.
Berg, T., Burg, V., Gombovic, A., Puri, M., 2020. On the rise of FinTechs: credit scoring using digital footprints. Rev. Financ. Stud. 33, 2845–2897.
Bernardo, A.E., Cai, H.B., Luo, J.A., 2004. Capital budgeting in multidivision firms: information, agency, and incentives. Rev. Financ. Stud. 17, 739–767.
Bertomeu, J., Magee, R.P., 2015. Mandatory disclosure and asymmetry in financial reporting. Journal of Accounting & Economics 59, 284–299.
Bertrand, M., Morse, A., 2011. Information disclosure, cognitive biases, and payday borrowing. J. Financ. 66, 1865–1893.
Bhandari, A., Javakhadze, D., 2017. Corporate social responsibility and capital allocation efficiency. J. Corp. Finan. 43, 354–377.
Brockman, P., Khurana, I.K., Martin, X., 2008. Voluntary disclosures around share repurchases. J. Financ. Econ. 89, 175–191.
Brockman, P., Martin, X.M., Puckett, A., 2010. Voluntary disclosures and the exercise of CEO stock options. J. Corp. Finan. 16, 120–136.
Chen, X., Huang, B.H., Ye, D.Z., 2020. Gender gap in peer-to-peer lending: evidence from China. J. Bank. Financ. 112.
Chung, H.M., Judge, W.Q., Li, Y.H., 2015. Voluntary disclosure, excess executive compensation, and firm value. J. Corp. Finan. 32, 64–90.
Crawford, V.P., Sobel, J., 1982. Strategic information transmission. Econometrica 50, 1431–1451.
Cumming, D.J., Hornuf, L., 2020. Marketplace Lending of SMEs. Working Paper.
Dorfleitner, G., Priberny, C., Schuster, S., Stoiber, J., Weber, M., de Castro, I., Kammler, J., 2016. Description-text related soft information in peer-to-peer lending -
evidence from two leading European platforms. J. Bank. Financ. 64, 169–187.
Duarte, J., Siegel, S., Young, L., 2012. Trust and credit: the role of appearance in peer-to-peer lending. Rev. Financ. Stud. 25, 2455–2483.
Fama, E.F., 1970. Efficient capital markets - review of theory and empirical work. J. Financ. 25, 383–423.
Fama, E.F., 1991. Efficient Capital-Markets .2. J. Financ. 46, 1575–1617.
Francis, J., Nanda, D., Wang, X., 2006. Re-examining the effects of regulation fair disclosure using foreign listed firms to control for concurrent shocks. Journal of
Accounting & Economics 41, 271–292.
Franks, J.R., Serrano-Velarde, N.A.B., Sussman, O., 2020. Marketplace Lending, Information Aggregation, and Liquidity. Review of Financial Studies, Forthcoming.
Fuster, A., Plosser, M., Schnabl, P., Vickery, J., 2019. The role of Technology in Mortgage Lending. Rev. Financ. Stud. 32, 1854–1899.
Gande, A., Kalpathy, S., 2017. CEO compensation and risk-taking at financial firms: Evidence from US federal loan assistance. J. Corp. Finan. 47, 131–150.
Gigler, F., 1994. Self-enforcing voluntary disclosures. J. Account. Res. 32, 224–240.
Goldstein, I., Yang, L., 2019. Good disclosure, bad disclosure. J. Financ. Econ. 131, 118–138.
Healy, P.M., Palepu, K.G., 2001. Information asymmetry, corporate disclosure, and the capital markets: a review of the empirical disclosure literature. Journal of
Accounting & Economics 31, 405–440.
Healy, P.M., Serafeim, G., 2016. An analysis of Firms’ self -reported anticorruption efforts. Account. Rev. 91, 489–511.
Heckman, J.J., 1979. Sample selection Bias as a specification error. Econometrica 47, 153–161.
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.
J. Mark. Res. 48, S138–S149.
Hildebrand, T., Puri, M., Rocholld, J., 2017. Adverse incentives in crowdfunding. Manag. Sci. 63, 587–608.
Hirshleifer, D., Teoh, S.H., 2003. Limited attention, information disclosure, and financial reporting. J. Account. Econ. 36, 337–386.
Iyer, R., Khwaja, A.I., Luttmer, E.F.P., Shue, K., 2016. Screening peers softly: inferring the quality of small borrowers. Manag. Sci. 62, 1554–1577.
Lewis, G., 2011. Asymmetric information, adverse selection and online disclosure: the case of eBay motors. Am. Econ. Rev. 101, 1535–1546.
Lin, M.F., Viswanathan, S., 2016. Home Bias in online investments: an empirical study of an online crowdfunding market. Manag. Sci. 62, 1393–1414.
Lin, M.F., Prabhala, N.R., Viswanathan, S., 2013. Judging borrowers by the company they keep: friendship networks and information asymmetry in online peer-to-
peer lending. Manag. Sci. 59, 17–35.
Liu, D., Brass, D.J., Lu, Y., Chen, D.Y., 2015. Friendships in online peer-to-peer lending: pipes, prisms, and relational herding. MIS Q. 39, 729–742.
Michels, J., 2012. Do unverifiable disclosures matter? Evidence from peer-to-peer lending. Account. Rev. 87, 1385–1413.
Nemoto, N., Storey, D.J., Huang, B., 2019. Optimal Regulation of P2P Lending for Small and Medium-Sized Enterprises. ADBI Working Paper No. 912.
Paravisini, D., Rappoport, V., Ravina, E., 2017. Risk aversion and wealth: evidence from person-to-person lending portfolios. Manag. Sci. 63, 279–297.
Pope, D.G., Sydnor, J.R., 2011. What’s in a picture? Evidence of discrimination from Prosper.com. J. Hum. Resour. 46, 53–92.
Sorenson, O., Assenova, V., Li, G.C., Boada, J., Fleming, L., 2016. Expand innovation finance via crowdfunding. Science 354, 1526–1528.
Spence, M., 1973. Job Market Signaling. Q. J. Econ. 87, 355–374.
Staiger, D., Stock, J.H., 1994. Instrumental variables regression with weak instruments. Econometrica 65, 557–586.
Stiglitz, J.E., Weiss, A., 1981. Credit rationing in markets with imperfect information. Am. Econ. Rev. 71, 393–410.
Strausz, R., 2017. A theory of crowdfunding: a mechanism design approach with demand uncertainty and moral Hazard. Am. Econ. Rev. 107, 1430–1476.
Tadelis, S., Zettelmeyer, F., 2015. Information disclosure as a matching mechanism: theory and evidence from a field experiment. Am. Econ. Rev. 105, 886–905.
Tang, H., 2019. Peer-to-peer lenders versus banks: substitutes or complements? Rev. Financ. Stud. 32, 1900–1938.
Vallée, B., Zeng, Y., 2019. Marketplace lending: a new banking paradigm? Rev. Financ. Stud. 32, 1939–1982.
Wen, X.Y., 2013. Voluntary disclosure and investment. Contemp. Account. Res. 30, 677–696.
Wittenberg-Moerman, R., 2008. The role of information asymmetry and financial reporting quality in debt trading: evidence from the secondary loan market. Journal
of Accounting & Economics 46, 240–260.
Zhang, J.J., Liu, P., 2012. Rational herding in microloan markets. Manag. Sci. 58, 892–912.
Zhao, Y.J., Allen, A., Hasan, I., 2013. State antitakeover Laws and Voluntary disclosure. J. Financ. Quant. Anal. 48, 637–668.
17