The effect of successful borrowing times on behavior of investors: An empirical investigation of the P2P online lending market

Abstract

The development of online peer-to-peer lending is in full bloom due to the support of network technology and information technology. More and more people are keen to participate in them. Online P2P lending platforms provide services where investors lend money to borrowers without the involvement of traditional financial institutions. Due to its convenience, the platforms have attracted many investors and borrowers. However, these investors may suffer a significant loss if they cannot make better borrowing decisions based on prediction credit risk. The purpose of this article is to study the relationship between successful borrowing times (SBT) and successful loans and default loans. This article further explores SBT plays an importance of the role in online P2P lending. Data was collected from PaiPaiDai, an online P2P lending platform in China. Logistic modeling was employed to test the proposed hypotheses. The results show that the probability of default loan by borrowers with more SBT is low, and the probability of successful loan is higher. It shows that investors are more willing to choose borrowers with more SBT. The investors can effectively recognize the value of SBT in the transaction process, and can identify credit risk through SBT in the online P2P lending market. SBT can predict credit risk, which suggests the proposed method has favorable potential being implemented in real-world P2P lending platforms.

Keywords

Successful borrowing times online P2P lending behavior of investors identify credit risk P2P lending market

Wang Xinmin, is a Ph.D scholar at Beijing University of Posts and Telecommunications, Beijing China. He received master degree in management science from Lanzhou University of technology. During his study, he got honors Excellent Graduate Student in Beijing University of Posts and Telecommunications. His current research interest is P2P online lending.

Hui Peng is born in Hunan, China on Sep. 11th, 1969. Educational background: Ph.D of Finance, People’s Bank of China, Beijing, China, 1998. She is currently working as a Professor of Finance/ Director of Internet Finance Research Center, Beijing University of Posts and Telecommunications, Beijing, China. She was a visiting scholar in University of Michigan, Michigan, U.S.A. 2003–2005, she was post-doctorate researcher in Renmin University, Beijing, China. Current research interest: risk management; E-commerce. Professor Peng is a member of Global Association of Risk Professionals, and a member of International Association of Computer Science and Information Technology.

Umair Akram, He is doing post doctorate research at Guanghua School of Management, Peking University. He got Ph.D. degree in (Marketing) from School of Economics and Management, Beijing University of Posts and Telecommunications, Beijing, China. Before commencing his Doctoral research, he worked as a Lecturer of marketing management at Institute of Business and Management Sciences, UAF, Pakistan He is the author of several publications and international conference proceedings. He has published numerous papers in journals such as Human Systems Management, China Economic Review, Sustainability, Asia Pacific Journal of Marketing and Logistics, Asia Pacific Journal of Financial Studies, Chinese Management Studies and International Journal of Conflict Management etc. He is reviewer board member of numerous SSCI journals. He is editorial advisory board member of Management Decision (SSCI) and Journal of Electronic Commerce in organization (ESCI) journals. His Current areas of interest includes, Online Impulse Buying Behavior, E-Marketing, E-Commerce, Social Commerce and Consumer Psychology.

Yan, Mengling. Currently working as Assistant professor in School of Economics and Management, Beijing University of Posts and Telecommunication. She got doctoral degree from Guanghua school of management, Peking University. She has published several papers in SSCI/SCI indexed journals and participate in reputable international conferences. Her current research areas of interest include, Business value of IT, innovation management and digital transformation.

Dr. Saman Attiq is currently working as an Associate Professor of Marketing at Air University Islamabad. She has done her PhD from Mohammad Ali Jinnah University, Islamabad. She is recipient of many prestigious awards including HEC Awarded “Post-Doc Fellowship Phase-II” Scholarship, HEC Awarded “Start-Up Research Grant” and MAJU Awarded “Dean’s Honor Roll”. She is currently the Managing Editor of “Journal of Business & Economics at Air University Islamabad. She has been Editor-in-Chief_ “UW Journal of Management Sciences”.

1 Introduction

The rapid grow of P2P, which is a result of the remarkable advancement of the internet and technology, presents a new paradigm for the financial markets [1]. The massive development of China’s P2P market facilitates the economic development and make China the largest online P2P lending market in the world. Online P2P lending platforms perform as a third-party platform for connecting individual lenders and borrowers by providing micro-credit services [2]. P2P online lending plays a vital role in everyone, s life. In China, P2P lending platforms could provide services with lower charges, more flexible conditions, and quicker loan approvals than traditional lending institutions [3]. These advantages attract individuals and platform investors, promoting a significant growth on P2P online lending market. According to Wangdaizhijia (The website is: http://www.wdzj.com), the number of P2P platforms is 2307 in 2016, with an annual growth rate of 2.81%. Number of transaction has reached a staggering 28 trillion RMB, increasing137% since 2015 [4]. Compared with the traditional credit market, P2P online lending is easy to access [5]. P2P solves the problem of financing difficulties for micro-enterprises and low-income people. It provides a new investment channel, and improves the efficient utilization of social funds to borrowers [6]. P2P online lending is more profitable, easier to operate, although unsecure than traditional banks, makes it attractive to a large number of investors. In the P2P online loan market, the performance differs from investor to investor. Some investors have investment losses due to lack of relevant professional knowledge and adequate realize of P2P online loans, while others with better financial knowledge have achieved higher returns. In the P2P online lending market, many factors are considered to judge the credit risk. These include historical invested experience [7], educational level [8], social capital [9], etc. However, few studies have used the successful borrowing times (SBT) as the basis for credit risk judgment.

P2P online lending plays an important role in the financial sector. Therefore, it has received widespread attention and research from scholars. The successful factor of borrowing was studied, and it has been shown that the demographic characteristics of borrower had an impact on the successful loan, such as gender, age and marital status [10, 11]. Using the data of PaiPaiDai platform, the paper analyzes borrowing orders’ the basic attribute, borrowers’ the basic information, and the impact of social capital on the successful rate of the loan. At the same time, there is a herding effect in the P2P online lending market [12]. Borrowers have more trustworthy appearance are more likely to have their lending funded [13]. Borrowers’ financial information have an impact on successful borrowing rate [14]. There are more friends who can borrow money easily, the interest rate is lower, and the default rate is low [15]. The lending descriptive language is longer, the successful rate of the loan is higher, and the default rate is lower. Some positive words can improve the successful rate of loan [16]. Borrowing descriptions have an impact on investor’ behavior [17]. A lower interest rate decreases the borrowers’ chances of getting the loan funded, while a smaller lending amount increases the successful probability [18]. Studying the behavior of the herd, the investors can infer the credit status of the borrower from the behavior of other borrowers [19]. The behavior of the herd is more significant when the bidding is not full, and behavior of the herd disappears when the bidding is full [10]. The herd behavior reduces online lending rates and reduces investor’ returns [20]. There are discrimination in the online lending market, such as regional discrimination, marriage discrimination, gender discrimination [21 –23], etc. Research online lending related policies [24]. We aim to study that these factors impact on investor’ behavior.

Although a voluminous research can be found on the successful loans, these studies do not embrace the successful borrowing times i.e. SBT. Moreover, all of these studies have been conducted on influencing factors of investor’ behavior. The objective of this study is in threefold: (a) to examine the relationship between successful loan and SBT (b) to examine the relationship between default repayment and SBT (c) to judge credit risk with SBT.

This manuscript is organized as follows, initially a review of literature is discussed, followed by theoretical framework and corresponding hypotheses. Next, methodology section explains sampling procedure and measures. Finally, results discussed along with theoretical and managerial implications.

2 Theoretical background and hypotheses development

Assume that the borrowing rate and credit rating in the pacing lending list are not considered for SBT. According to theory of signal transmission by Spence [25], The lending costs and credit ratings are negatively correlated. The borrower’s SBT is higher, the risk rating is higher, and lending cost is lower. The borrower’s SBT is lower, the credit rating is lower, and lending costs is higher. Therefore, the investors can judge the credit risk of the borrowers according to SBT, and provide more borrowing opportunities for the borrowers if the SBT is higher. In this case, SBT has a role of signal transmission. Previous studies have confirmed the results of transmission theory. SBT is significantly positive for the successful loan, and the borrower’s previous successful lending records will attract investors [8]. Ding et al [26] found that the probability of successful borrowing has a positive relationship with SBT. Song & Jin [27] found that investors prefer borrowers with more SBT. The greater SBT, the greater the credit ratings for borrowers as well as greater the lending successful rate. This explains less expensive to borrow. In addition, the PaiPaiDai platform will blacklist the defaulters, and exposure of private information. But for the exposed borrowers have different SBT, the lending costs are different. The destruction of the credit rating makes borrowers increase the cost in the future borrowing process. Therefore, borrowers are more important to value their own credit and not to default as much as possible. This discussion explains that lower lending costs and higher default costs make borrowers with more SBT less likely to default.

At the same time, in the actual transaction. Using the SBT as a key reference factor to determine credit rating and interest rate. Borrowers with a higher SBT are considered to be borrowers with higher degree of credit in the PaiPaiDai platform. These borrowers have lower probability of default repayment and lower probability of failed borrowing. Thus, the platform will set the borrower’s higher credit rating, and set the lower lending rate. However, we believe that under the same conditions of credit rating and interest rate, the borrowers with more SBT have higher probability of default repayment, and the probability of borrowing failure is higher. Aforementioned discussion leads us to formulate the following hypotheses:

H1: Borrower with more SBT have lower probability of failed loan.

H2: Borrower with more SBT have higher probability of failed loan.

H3: Borrower with more SBT have lower probability of default repayment.

H4: Borrower with more SBT have higher probability of default repayment.

Importantly, SBT value can be recognized in the online lending market as well as in traditional financial market. However, there are relatively few empirical studies on how SBT affects the operation of the online loan market. Borrowers have more SBT in the previous period, the higher their positive reputation and the lower default rate of borrowers [17]. In the P2P online lending platforms, the borrowers with low credit rating accumulated SBT in order to be able to successfully loan, increase credit rating, and give investors the illusion. Therefore, under the same credit rating and borrowing rate, investors can use the disclosed information by the borrower on the platform to improve their ability to predict risks. If the borrowers with more the SBT have fewer defaults, then the investors should recognize the value of SBT, the SBT is favored by investors in P2P online lending.

In the PaiPaiDai lending platform, can investors effectively use SBT to identify credit risks? If the borrower’s SBT is higher, the probability of default is lower and the probability of successful loan is higher. Or the probability of default is higher and the probability of successful borrowing is lower. Or have no effect on the borrower’s probability of default repayment and probability of successful loan. Therefore, there is no deviation in identifying credit risk with SBT. But we believe that SBT have deviations to identify credit risk. We put forward two opposing hypotheses:

H5: There is no deviation when investors use the borrower’s SBT to judge credit risk.

H6: There is deviation when investors use the borrower’s SBT to judge credit risk.

3 Methods

3.1 Sample and data

PaiPaiDai online lending platform was established in 2007, it is the earliest P2P online loan platform established in China. Borrowing is generally a short-term loan, the longest loan period in our data set is 12 months. The lending process as follows: The borrowers will upload personal information and borrowing requests. The platform will set the credit level of the borrowing list based on the information uploaded by the borrowers, set the borrowing rate within a certain range, and form a list of borrowings available for investors to bidding. The investors will select the borrowers by browsing the information on the lending lists. Within the specified time, if the bidding amount reached the borrowing requests, the borrowing list will be closed and become a successful borrowing list, form a lending relationships. If the borrowing amount is not met and will become a list of failed loan. In list of successful loans, the article further collected the repayment status of the lending lists which may judge default behavior.

In order to accurately know the default lending of the information, we collected a total of 1057858 records of loans in 2017, and got their repayment information. In these lending lists, there are 228875 successful borrowing records. The sample of successful loans accounted for 21.6% of the total sample. Excluding incomplete records and missing records, only have 987754 records of loans. In these samples, there are 3547 records of default repayment and 758879 records of failed loan, the sample ratio of successful loans accounted for 23.2%. The difference between the two samples is small. Therefore, this paper will select the sample of the SBT as the research object.

3.2 Measures

Dependent variable. Research is carried out from two aspects: successful loan and default loan. First, we constructed two dummy variables: Success and Default. If successful loan then it donates with a value of 1, otherwise a value of 0.If default loan then it donates with a value of 1, otherwise it is 0. Whether the borrowers have default loans or successful loans, we test hypotheses H1, H2, H3, and H4 by studying the impact of SBT on successful loan and default loan. Whether investors use SBT as a method to judge the credit risk of borrowers. Whether the borrowers can get the trust of the investors, combined with the regression results of hypotheses H1, H2, H3, and H4. The investors can judge the deviation of the borrowers’ risk and the real risk. Thus, making a correct judgment on the hypotheses H5 and H6.

Independent variable. The SBT is the core variable of this paper, we set as independent variable. For reasons of robustness, the article set three different independent variables: (a) SBT set as continuous variable (b) SBT1 set as dummy variable, if SBT1 value represent 5 times or more than 5 times then it donates with a value of 1, otherwise a value of 0.(c) SBT2 set as dummy variable, if SBT2 value represent 10 times or more than 10 times then it donates with a value of 1, otherwise a value of 0.

Control variable. Control variable includes three categories: (a) borrower’s personal information related variables (b) borrowing list related variables (c) Variables extracted from the proof materials uploaded by the borrower. Variables related to borrowers include Gender; Age; Marital status; Car; House; Education; Success and Default. The variables associated with the loan list are: The credit rating is AAA, AA, A, B, C, D, E, F and so on eight levels then they donate with a value of 1,2,3,4,5,6,7,8. Borrowing interest rate; Borrowing amount (Yuan); Borrowing term (month); Borrowing purpose. The proof materials have two parts: a part is proof of the borrower’s personal information and lending listing information; another part is disclosed information about other aspects of the borrowers.

More than 60 kinds of materials were uploaded by borrowers in the PaiPaiDai platform. We refer to Michels (2012) for the way to simple classification [28]. First, a part of proof is about borrower’s information and lending list information, it includes a total of five kinds: Proof of marriage; Proof of education; Proof of buying house; Proof of buying car; Proof of the borrowing purpose. Second, the borrower’s other information, the article divides this information into six categories: Proof of credit; Proof of personal information; Proof of social relationships; Proof of consumption; Proof of debt; Proof of repayment ability. In the process of empirical analysis, setting these control variables as dumb variables. Table 1 shows descriptive statistical analysis.

Table 1
Descriptive statistical analysis

Variable All sample Default repayment Successful loan Failed loans

(987754) (3547) (228875) (758879)

SBT 1.52(16.32) 2.12(3.15) 6.45(30.21) 0.13(1.85)

Gender 0.75 0.85 0.81 0.87

Age 30.25(6.23) 33.16(6.11) 31.70(6.62) 28.22(6.55)

Education 3.10(1.22) 3.75(1.08) 3.55(1.13) 3.14(1.18)

Marriage 0.45 0.53 0.54 0.44

Car 0.24 0.27 0.31 0.24

House 0.25 0.31 0.39 0.27

Failed loans 0.39(1.31) 0.75(1.20) 0.72(2.45) 0.24(0.80)

Credit rating 6.11(2.35) 6.45(1.20) 2.76(1.95) 7.40(0.95)

Log (Quota) 9.07(0.35) 8.35(0.77) 8.54(0.45) 8.01(0.16)

Interest rate 0.11(0.02) 0.21(0.01) 0.16(0.04) 0.17(0.01)

Term 8.85(2.54) 8.13(3.49) 7.86(3.11) 8.67(2.63)

Loan purpose 0.85 0.63 0.53 0.87

Proof of marriage 0.02 0.11 0.16 0.24

Proof of educational 0.03 0.08 0.11 0.02

Proof of buying house 0.04 0.15 0.13 0.05

Proof of buying car 0.06 0.20 0.14 0.02

Proof of borrowing purpose 0.85 0.65 0.78 0.94

Proof of personal 0.41 0.09 0.85 0.23

Proof of social relations 0.07 0.24 0.18 0.06

Proof of consumption 0.12 0.41 0.36 0.07

Proof of repayment ability 0.21 0.57 0.42 0.18

Proof of debt 0.05 0.04 0.03 0.02

Variable	All sample	Default repayment	Successful loan	Failed loans
SBT	1.52(16.32)	2.12(3.15)	6.45(30.21)	0.13(1.85)
Gender	0.75	0.85	0.81	0.87
Age	30.25(6.23)	33.16(6.11)	31.70(6.62)	28.22(6.55)
Education	3.10(1.22)	3.75(1.08)	3.55(1.13)	3.14(1.18)
Marriage	0.45	0.53	0.54	0.44
Car	0.24	0.27	0.31	0.24
House	0.25	0.31	0.39	0.27
Failed loans	0.39(1.31)	0.75(1.20)	0.72(2.45)	0.24(0.80)
Credit rating	6.11(2.35)	6.45(1.20)	2.76(1.95)	7.40(0.95)
Log (Quota)	9.07(0.35)	8.35(0.77)	8.54(0.45)	8.01(0.16)
Interest rate	0.11(0.02)	0.21(0.01)	0.16(0.04)	0.17(0.01)
Term	8.85(2.54)	8.13(3.49)	7.86(3.11)	8.67(2.63)
Loan purpose	0.85	0.63	0.53	0.87
Proof of marriage	0.02	0.11	0.16	0.24
Proof of educational	0.03	0.08	0.11	0.02
Proof of buying house	0.04	0.15	0.13	0.05
Proof of buying car	0.06	0.20	0.14	0.02
Proof of borrowing purpose	0.85	0.65	0.78	0.94
Proof of personal	0.41	0.09	0.85	0.23
Proof of social relations	0.07	0.24	0.18	0.06
Proof of consumption	0.12	0.41	0.36	0.07
Proof of repayment ability	0.21	0.57	0.42	0.18
Proof of debt	0.05	0.04	0.03	0.02

Note: The standard deviation of continuous variables is in parentheses.

Table 1 present the descriptive statistical analysis. In the column of successful loan, the mean of credit rating is 2.76, and mean of SBT is 6.45. In the column of lending fails, the mean of credit rating is 7.40, and mean of SBT is 0.13. In the column of default repayment, the mean of is 2.12. This shows that the credit rating has a greater impact on the success of the loan. SBT is higher, probability of successful loan is higher, probability of failed loan is lower, and probability of default repayment is lower.

3.3 Empirical model

3.3.1 Model building

This paper uses theoretical model as: $P_{i} = Pr (y = 1 | {Times}_{i}, {Control}_{i}) = \frac{1}{1 + e^{- (α + β {Times}_{i} + γ {Control}_{i})}}$ . Model (1) shows that the regression model of the successful loans affected by the SBT. Model (2) shows that the regression model of default loans affected by the SBT. The models are as: ${Success}_{i} = α_{01} + β_{11} {Times}_{i} + γ_{11} {Control}_{i} + ɛ_{i}$ (1) ${Default}_{i} = α_{02} + β_{22} {Times}_{i} + γ_{22} {Control}_{i} + ɛ_{i}$ (2)

Where we use Success_i to denote lending status. the lending success donate with a value of 1, otherwise donate with a value of 0. Where we use Default_i to denote default status, the default loan donate with a value of 1, otherwise donate with a value of 0. We use Times_i to denote SBT, use Control_i to denote control variable, use ɛ_i to denote random error term. If the test hypothesis H1, β₁1 must be a positive coefficient in the model of (1), which indicates SBT has a positive impact on the successful loan. That is, SBT is higher, and probability of borrowing failure is lower; SBT is lower, probability of borrowing failure is higher. If the test hypothesis H3, β₂2 must be a negative coefficient in the model of (2), which indicates the SBT has a negative impact on the default loan. That is, SBT is higher, and probability of lending default is lower; SBT is lower, and probability of lending default is higher. If the test hypothesis H4, β₂2 must be a positive coefficient or significantly positive. Simultaneously, keeping the control variables unchanged. We will test hypothesis H5 and H6, combining with the empirical results of model (1) and model (2). β₁1 is significantly negative, and β₂2 is significantly positive, which indicates the borrowers with a higher SBT have a higher probability of default loan and a lower probability of successful loan. Or β₁1 is significantly positive, and β₂2 is significantly negative, which indicates the borrowers with a higher SBT have a lower probability of default loan and a higher probability of successful loan. Or β₁1 and β₂2 are not significant, which indicates SBT has no impact on default loan and successful loan. That is, hypothesis H5 is supported, or hypothesis H6 is supported.

3.3.2 Empirical analysis

For the purpose of testing robustness, except SBT be set as continuous variables, SBT1 and SBT2 be set as dummy variables. Table 2 shows SBT has impact on the successful loan. Columns 1,2,3 are the benchmark regression results of the model, no addition of other control variables, only the independent variable makes a direct regression on the dependent variable. In the benchmark model, SBT has a positive coefficient, which shows that SBT is higher, the probability of successful loan is higher. SBT has a positive impact on the successful loan, Hypothesis H1 is supported. The regression result in column 4 is to add borrower’s information and borrowing list information as control variables in the benchmark model. The result of β₁1 is 0.09, hypothesis H1 is supported. This shows that in the case of other control variables, if the SBT is higher, the probability of successful loan is higher. For the purpose of model robustness. In this paper, SBT1 and SBT2 be set as dummy variables, the control variables remain unchanged, got two new models, the regression results are in columns 5 and 6, found that β₁1 is consistent with the regression results in column 4, still is significantly positive at the 1% level.

Table 2
The impact of SBT on the successful borrowing (Success)

1 2 3 4 5 6 7 8 9

SBT 0.23^* (0.00) 0.09^* (0.01) 0.05^* (0.01)

SBT1 0.81^* (0.00) 0.35^* (0.01) 0.12^* (0.00)

SBT2 0.63^* (0.01) 0.25^* (0.00) 0.11^*** (0.01)

borrower’s information no no no yes yes yes yes yes yes

List information no no no yes yes yes yes yes yes

Proof material information no no no no no no yes yes yes

constant yes yes yes yes yes yes yes yes yes

Number of samples 987754 987754 987754 987754 987754 987754 987754 987754 987754

LogLikelihood –52122.1 –51442.5 –53216.7 –13214.2 –13654.3 –13574.6 –11564.5 –11754.8 –11632.1

	1	2	3	4	5	6	7	8	9
SBT	0.23^*** (0.00)			0.09^*** (0.01)			0.05^*** (0.01)
SBT1		0.81^*** (0.00)			0.35^*** (0.01)				0.12^*** (0.00)
SBT2			0.63^*** (0.01)			0.25^*** (0.00)			0.11^*** (0.01)
borrower’s information	no	no	no	yes	yes	yes	yes	yes	yes
List information	no	no	no	yes	yes	yes	yes	yes	yes
Proof material information	no	no	no	no	no	no	yes	yes	yes
constant	yes	yes	yes	yes	yes	yes	yes	yes	yes
Number of samples	987754	987754	987754	987754	987754	987754	987754	987754	987754
LogLikelihood	–52122.1	–51442.5	–53216.7	–13214.2	–13654.3	–13574.6	–11564.5	–11754.8	–11632.1

Note: Significance.: ^*p < 0.10; ^**p < 0.05; ^***p < 0.01. The upper side is the coefficient, and the parentheses are the standard deviation.

In order further considers the impact of the proof materials uploaded by borrowers on investors’ investment decisions. In subsequent models, the information disclosed in the proof materials uploaded by the borrower is also put into the model as a control variable in this paper. First, consistent with the empirical process of hypotheses H1, and H2, we add the amount of information disclosed in the proof material as a control variable to the regression model, and combined with different independent variables, got three different models, the empirical results are shown in columns 7, 8, and 9 of Table 2. In the column 7, 8, and 9, b₁₁ is significantly positive at 0.1% level, still remains unchanged, Hypothesis H1 is supported again. In all of the regression model, all variables have VIF values below 3, the article believes that there is basically no problem of collinearity. In summary, borrowers with more SBT make a lower probability of failed loan, so hypothesis H1 is supported.

For the purposes of robustness test, in addition to SBT be set as continuous variables, SBT1 and SBT2 be set as dummy variables. In Table 3, columns 1, 2, and 3 are the results of the regression of the above three independent variables directly on the dependent variable. As a benchmark model, they are not added to the control variable. In these regression results, β₂2 is significantly negative at 0.1%, hypothesis H3 is supported. Subsequent regression results obtained by adding control variables, column 4 is the regression result obtained by directly putting the continuous variables of SBT into the model, also add control variables related to the borrower and related to the loan list, this is significantly negative at the 0.1% level, The hypothesis H3 is also supported. Then introduce two dummy variables with SBT indicators instead of continuous variables into the model, in the case other control variables are unchanged, the regression results are shown in columns 5 and 6. In these regression results, β₂2 is significantly negative at 0.1% level once again support the hypothesis H3.

Table 3

The impact of SBT on the default(Default)

	1	2	3	4	5	6	7	8	9
SBT	–0.16^*** (0.02)			–0.18^*** (0.12)			–0.15^*** (0.23)
SBT1		–0.51^*** (0.04)			–0.57^*** (0.03)			–0.55^*** (0.06)
SBT2			–0.33^*** (0.02)			–0.37^*** (0.03)			–0.34^*** (0.02)
borrower’s information	no	no	no	yes	yes	yes	yes	yes	yes
List information	no	no	no	yes	yes	yes	yes	yes	yes
Proof material information	no	no	no	no	no	no	yes	yes	yes
constant	yes	yes	yes	yes	yes	yes	yes	yes	yes
Number of samples	228875	228875	228875	228875	228875	228875	228875	228875	228875
LogLikelihood	–17523.1	–17412.2	–17565.5	–13144.2	–13525.4	–13612.3	–13412.9	–13735.6	–13524.7

Note: Signif.: ^*p < 0.10; ^**p < 0.05; ^***p < 0.01. The upper side is the coefficient, and the parentheses are the standard deviation.

The article further considers the proof materials uploaded by the borrower, first, we will prove that the amount of information disclosed in the material is used as a control variable into the model. at the same time, combined with the three independent variables set in the article, got three different regression models, the results of the empirical analysis are listed in columns 7, 8, and 9 of Table 3, in these regression results, the regression coefficients of β₂2 are significantly negative at the 0.1% level, and the regression results remain unchanged, the hypothesis H3 is supported. In all of the regression models, all variables have VIF values below 3, the article believes that there is basically no problem of collinearity. In summary, borrowers with more SBT make lower probability of default loan, accordingly, the hypothesis H3 is supported. This indicates that SBT is higher, the probability of successful loan is higher, and probability of default loan is lower. Investors prefer highly borrowers with the higher SBT. The empirical results of hypothesis H1, H2, H3 and H4 have shown SBT is higher, the probability of default loan is lower, and probability of successful loan is higher. Empirical results of the composite hypothesis H1, H2, H3 and H4, found that the online lending market can recognize the value of SBT, investors can reasonably use SBT to determine the credit risk, and there is no deviation. Thus, the hypothesis H5 is confirmed.

4 Robustness check

In this paper, the Probit model is selected to perform stability tests on the regression results of independent variables and control variables in Tables 2 and 3. The regression results are shown in Tables 4 and 5. Table 4 checks the impact of SBT on the default loan, Table 5 checks the impact of SBT on the successful loan. In the default loan and loan successful loan both cases, Using Probit model regression, the results are basically consistent with the regression results obtained by the Logistic model. This shows that the regression results of the article are robust.

Table 4
The impact of SBT on the default loan (Default)

1 2 3 4 5 6

SBT –0.76^(0.01) –0.35^(0.00)

SBT1 –0.31^* (0.00) –0.21^* (0.01)

SBT2 –0.18^* (0.02) –0.23^* (0.01)

Borrower information yes yes yes yes yes yes

List information yes yes yes yes yes yes

Proof material information yes yes yes no no no

constant yes yes yes yes yes yes

Number of samples 228875 228875 228875 228875 228875 228875

LogLikelihood –11256.1 –11242.0 –11257.3 –13542.3 –13521.7 –13563.5

	1	2	3	4	5	6
SBT	–0.76^***(0.01)			–0.35^***(0.00)
SBT1		–0.31^*** (0.00)			–0.21^*** (0.01)
SBT2			–0.18^*** (0.02)			–0.23^*** (0.01)
Borrower information	yes	yes	yes	yes	yes	yes
List information	yes	yes	yes	yes	yes	yes
Proof material information	yes	yes	yes	no	no	no
constant	yes	yes	yes	yes	yes	yes
Number of samples	228875	228875	228875	228875	228875	228875
LogLikelihood	–11256.1	–11242.0	–11257.3	–13542.3	–13521.7	–13563.5

Note: Signif.: ^*p < 0.10; ^**p < 0.05; ^***p < 0.01. The upper side is the coefficient, and the parentheses are the standard deviation.

Table 5

The impact of SBT on the successful loan (Success)

	1	2	3	4	5	6
SBT	0.02^*** (0.01)			0.02^*** (0.00)
SBT1		0.06^*** (0.01)			0.05^*** (0.02)
SBT2			0.04^*** (0.01)			0.03^*** (0.00)
Borrower information	yes	yes	yes	yes	yes	yes
List information	yes	yes	yes	yes	yes	yes
Proof material information	yes	yes	yes	no	no	no
constant	yes	yes	yes	yes	yes	yes
Number of samples	987754	987754	987754	987754	987754	987754
LogLikelihood	–116231.4	–116453.7	–116534.5	–125442.1	–125361.3	–125457.8

Note: Signif.: ^*p < 0.10; ^**p < 0.05; ^***p < 0.01. The upper side is the coefficient, and the parentheses are the standard deviation.

5 Conclusion

This article takes the list data of the PaiPaiDai platform as a sample, an empirical analysis of SBT role in online lending was performed. In the online loan market, the study found that the SBT has a significant negative impact on default behavior, which indicates that the higher SBT lower the default loan. Furthermore, investors also favor borrowers with higher SBT lead to have a higher probability of successful loan. This shows that in the PaiPaiDai platform, there is no deviation in the use of the SBT to discriminate credit risk. In the online lending market, we can effectively recognize the value of the SBT, and reduce the risk of investment. However, in the P2P online lending platforms, low credit borrower use multiple successful loans and multiple repayments in order to improve credit rating, deliberately increase SBT, in order to gain investor trust. Therefore, investors can reasonably use the information of the borrower’s lending list and uploaded proof materials to reduce information asymmetry, the borrower’s breach of contract is predicted through information disclosed by the platform.

This research results providing useful guidelines for the actual operation of finance in reality. Moreover, SBT can be used to identify credit risks in financial institutions or P2P lending platforms. This study can be used as a reference for P2P platform investment, specifically use SBT as the basis for credit risk judgment. This guarantees the smooth and healthy development of P2P online loans.

The findings of this study should be interpreted in light of its limitations. The transaction data are collected from PaiPaiDai. Caution should be taken generalizing the conclusions of SBT to other P2P platforms because diverse their diverse settings for financing loans may affect SBT in market. Future studies should examine data from multiple platforms in different countries to build a more comprehensive understanding of the role of SBT in the P2P lending market.

References

Xie

, Zhou

. Research on internet finance model. Journal of Financial Research. 2012;12:11–22.

Boase

, Ling

. Measuring mobile phone use: Self-report versus log data. Journal of Computer-Mediated Communication. 2013;18(4):508–19.

Guo

, Zhou

, Luo

, Liu

and Xiong

. Instance-based credit risk assessment for investment decisions in P2P lending. European Journal of Operational Research. 2016;249(2):417–26.

Wang

, Xu

. China’s Online Lending Industry in 2015, Tsinghua University Press, 2015.

Mild

, Waitz

, Wockl

. How low can you go? Overcoming the inability of lenders to set proper interest rates on unsecured peer-to-peer lending markets. J Bus Res. 2015;68:1291–305.

Duarte

, Siegel

, Young

. Trust and credit: The role of appearance in peer-to-peer lending. Rev Finance Stud. 2012;25:2455–83.

Peng

, Qin

. The impact of historical investment experience on investor’s decision in P2P network leading. Finance Forum. 2018;5:24–41.

Yao

. Empirical study on the market effectiveness of China’ s P2P network lending market. Southwest Finance. 2018;(2):137–45.

Lian

, Jing

. The influence of social capital on the default risk of borrowers in P2P online lending. Finance Forum. 2013;(4):8–17.

10.

Herzenstein

, Dholakia

, Andrews

. Strategic herding behavior in peer-to-peer loan auction[J]. Journal of Interactive Marketing. 2011;25(1):27–36.

11.

Pope

, Sydnor

. What’s in a picture? Evidence of Discrimination from Prosper.com[J]. Journal of Finance. 2011;(1):53–92.

12.

, Guo

, Zhang

. Analysis of factors influencing the success rate of borrowing in China’s P2P small loan market. Journal of Financial Research. 2013(7):126–39.

13.

Duarte

, Siegel

, Young

. Trust and credit: The role of appearance in peer to peer lending. Review of Financial Studies. 2012;25(8):2455–84.

14.

Freedman

, Jin

. Do social networks solve information problems for peer-to-peer lending? Evidence from Prosper.com. NET Institute Working Paper. 2008;(4):08–43.

15.

Lin

, Nagpurnanand

, Viswanathan

. Judging borrowers by the company they keep: Friendship networks and information asymmetry in online peer-to-peer lending. Management Science. 2013;59(1):17–35.

16.

Liao

, Ji

, Zhang

. Can the language be trusted? The role of language in lending markets. Evidence from P2P platforms. Journal of Tsinghua University(Science and Technology). 2015;(4):413–21.

17.

, Gao

, Li

, Cai

, Wang

, Yang

. The influence of borrower’s description on investors’decision: Analyze based on P2P online lending. Economic Research. 2014;(1):143–55.

18.

Puro

, Teich

, Wallenius

. Borrower decision aid for people-to-people lending. Decision Support Systems. 2010;49:52–60.

19.

Zhang

, Liu

. Rational herding in microloan markets. Management Science. 2012;58(5):892–912.

20.

Chen

, Lin

. Rational or Irrational Herding in Online Microloan Markets: Evidence from China, SSRN Working Paper Available at SSRN 2425047, 2014.

21.

Jiang

, Zhou

. Do there exist regional discrimination in P2P lending?: An empirical on renrendai.com. Journal of Central University of Finance & Economics. 2016;(9):29–40.

22.

Guo

. Research on the credit risk in China based on the complex network. Modern Economic Science. 2017;39(3):14–49.

23.

Xiao

, Ye

. Gender discrimination in chinese internet finance. Chinese Review of Financial Studies. 2016;(2):1–16.

24.

Wei

, Ling

. Market mechanisms in online peer-to-peer lending [J]. Management Science. 2016;71(12):4236–57.

25.

Spence

. Job market signaling. The Quarterly Journal of Economics. 1973;87(3):355–74.

26.

Ding

, Lv

, Zhang

. The research on the effects of platform information disclosure and credit rating to the success rate of P2P loan funding. Journal of Industrial Technological Economics. 2016;(10):12–21.

27.

Song

, Jin

. Analysis of incentive effect on credit rating division based on motivation process theory. Journal of Wuhan University of Technology (Information & Management Engineering). 2018;40(1):86–92.

28.

Michels

. Do unverifiable disclosures matter? Evidence from peer -to – peer lending. The Accounting Review. 2012;87(4):1385–413.