A privacy-preserving decentralized credit scoring method based on multi-party information

2023 | Decision Support Systems | Citations: 0

Authors: He, Haoran; Wang, Zhao; Jain, Hemant; Jiang, Cuiqing; Yang, Shanlin

Abstract: With society's wide-scale adoption of information technology, significant inform ... Expand

Abstract: With society's wide-scale adoption of information technology, significant information about borrowers is distributed across various parties, information that can be jointly used to improve credit scoring. However, use of such information faces many challenges, such as the problems of preserving privacy and information redun­ dancy. To address these challenges in leveraging multi-source information for credit scoring, we propose a decentralized multi-party method based on logistic regression. Specifically, we formulate a logistic regression model using the vertical federated learning paradigm. To preserve data privacy during multi-party collaborative model training, we use additively homomorphic encryption based on the second-order Taylor series expansion of the loss function and its gradient. To address information redundancy and to improve the performance of the credit scoring model, we use the proposed HE-DPGD algorithm to estimate the distributed coefficients in a privacy-preserving setting. Empirical evaluation shows that the proposed method can leverage information from multiple parties securely and effectively. Collapse

Topics: privacy homomorphic encryption logistics management online lending missing data

Methods: computational algorithm machine learning federated learning logistic regression experiment

Putting Religious Bias in Context: How Offline and Online Contexts Shape Religious Bias in Online Prosocial Lending

2023 | Management Information Systems Quarterly | Citations: 0

Authors: Sabzehzar, Amin; Burtch, Gordon; Hong, Yili; Raghu, T. S.

Abstract: Biases on online platforms pose a threat to social inclusion. We examine the inf ... Expand

Abstract: Biases on online platforms pose a threat to social inclusion. We examine the influence of a novel source of bias in online philanthropic lending, namely that associated with religious differences. We first propose religion distance as a probabilistic measure of differences between pairs of individuals residing in different countries. We then incorporate this measure into a gravity model of trade to explain variation in country-to-country lending volumes. We further propose a set of contextual moderators that characterize individuals’ offline (local) and online social contexts, which we argue combine to determine the influence of religion distance on lending activity. We empirically estimate our gravity model using data from Kiva.org, reflecting all lending actions that took place between 2006 and 2017. We demonstrate the negative and significant effect of religion distance on lending activity, over and above other established factors in the literature. Further, we demonstrate the moderating role of lenders’ offline social context (diversity, social hostilities, and governmental favoritism of religion) on the aforementioned relationship to online lending behavior. Finally, we offer empirical evidence of the parallel role of online contextual factors, namely those related to community features offered by the Kiva platform (lending teams), which appear to amplify the role of religious bias. In particular, we show that religious team membership is a double-edged sword that has both favorable and unfavorable consequences, increasing lending in general but skewing said lending toward religiously similar borrowers. Our findings speak to the important frictions associated with religious differences in individual philanthropy; they point to the role of governmental policy vis-à-vis religious tolerance as a determinant of citizens’ global philanthropic behavior, and they highlight design implications for online platforms with an eye toward managing religious bias. Collapse

Topics: electronic market online community crowdsourcing crowdfunding peer-to-peer model

Methods: robustness check descriptive statistic qualitative content analysis survey archival research

Theories: theory of economic growth

Can Social Disconnectedness Inhibit Online Trade? Examining the Effects of Digital Distance on Peer-to-peer Lending

2023 | International Conference on Information Systems | Citations: 0

Authors: Sabzehzar, Amin; Zhou, Yingxin; HOU, JINGBO

Abstract: The extant literature has shown that offline group proximity manifests in onlin ... Expand

Abstract: The extant literature has shown that offline group proximity manifests in online peer-topeer lending platforms, inhibiting online transactions in those markets. The findings of this research suggest that digital distance, as measured by the rate of Facebook friendship between country pairs, can also influence lending actions in bi-country lending. Building on a dataset from Kiva.org, we show that digital distance significantly and negatively affects bi-country lending actions, on top of other distance-related barriers discussed in the literature. The results also shed light on the role of government policies regarding local IT infrastructure and Internet freedom, revealing that greater levels of IT infrastructure and Internet freedom can compensate for the negative effect of digital distance on prosocial lending. Collapse

Topics: internet technology information technology infrastructure Facebook crowdsourcing virtual teamwork

Methods: descriptive statistic regression analysis method

Depicting Risk Profile over Time: A Novel Multiperiod Loan Default Prediction Approach

2023 | Management Information Systems Quarterly | Citations: 0

Authors: Wang, Zhao; Jiang, Cuiqing; Zhao, Huimin

Abstract: With the rapid development of fintech, the need for dynamic credit risk evaluat ... Expand

Abstract: With the rapid development of fintech, the need for dynamic credit risk evaluation is becoming increasingly important. While previous studies on credit scoring have mostly focused on single-period loan default prediction, we call for a new avenue-multiperiod default prediction (MPDP)-to depict risk profiles over time. To address the challenges raised by MPDP, such as monotonic default probability prediction and complex relationship accommodation, we propose a novel approach, hybrid and collective scoring (HACS). We design a hybrid modeling strategy to predict whether and when a borrower will default separately through a default discrimination model and a default time estimation model, respectively, and synthesize them through a probabilistic framework. To accommodate various possible patterns of default time and measure the distribution of default probability over successive time intervals, we propose a joint default modeling method to train the default time estimation model. Empirical evaluations at the model (time-to-default prediction performance and discrimination performance) and mechanism (identifiability and discriminability) levels, as well as impact analyses at the application (granting performance and profitability performance) level, show that HACS outperforms the benchmarked survival analysis and multilabel learning methods on all fronts. It can more accurately predict time-to-default and provide financial institutions and investors better decisionsupport in granting loans and selecting loan portfolios. Collapse

Topics: online lending risk management decision making analytical information system

Methods: survival analysis case study machine learning design science design artifact

Theories: portfolio theory

Specialists, Generalists, or Both? Founders’ Multidimensional Breadth of Experience and Entrepreneurial Ventures’ Fundraising at IPO

2023 | Organization Science | Citations: 0

Authors: Souitaris, Vangelis; Peng, Bo; Zerbinati, Stefania; Shepherd, Dean A.

Abstract: Different streams of research have led to contradictory conclusions about the ve ... Expand

Abstract: Different streams of research have led to contradictory conclusions about the venture performance implications of founders’ breadth of experience. Although extant empirical studies have explored the performance implications of founders’ breadth of experience at the start-up stage, we focus on the later stage of the initial public offering (IPO). We theorize that investors categorize venture founders based on two salient dimensions—their industry and functional background—and we relate this categorization to resource acquisition at IPO. To test our model, we use a hand-collected data set of 175 entrepreneurial IPOs in the Alternative Investment Market in London (2002–2013) and two randomized experiments. We theorize and find that compared with entrepreneurial ventures with a lead founder specializing in one industry or one function, investors generally devalue those with a category-spanning lead founder (a generalist). However, devaluation is less severe when a lead founder is a generalist in one dimension (e.g., industry) but a specialist in the other dimension (e.g., function). We also theorize and empirically test trust as a mechanism for the generalist penalty. Specifically, audience members (investors) have low trust in a generalist producer (founder) in contexts where the two parties consider entering into a partnership (equity investment at IPO), and so that generalist producer is devalued. Finally, we show that an external expert endorsement—in our case, from intensive venture capital affiliations—offsets the generalist penalty, especially when category spanning occurs in multiple category dimensions. Funding: This work was supported by the University of St. Gallen and the Natural Science of China [Grant 71832012]. Supplemental Material: The e-companion is available at https://doi.org/10.1287/orsc.2022.1581. Collapse

Topics: marketing management IT career entrepreneurship database system decision making

Methods: experiment theory development robustness check qualitative interview survey

Theories: social identity theory

Risk-Averse Two-Stage Stochastic Minimum Cost Consensus Models with Asymmetric Adjustment Cost

2022 | Group Decision and Negotiation | Citations: 0

Authors: Ji, Ying; Li, Huanhuan; Zhang, Huijie

Abstract: In the process of reaching consensus, it is necessary to coordinate different vi ... Expand

Abstract: In the process of reaching consensus, it is necessary to coordinate different views to form a general group opinion. However, there are many uncertain factors in this process, which has brought different degrees of influence in group decision-making. Besides, these uncertain elements bring the risk of loss to the whole process of consensus building. Currently available models not account for these two aspects. To deal with these issues, three different modeling methods for constructing the two-stage mean-risk stochastic minimum cost consensus models (MCCMs) with asymmetric adjustment cost are investigated. Due to the complexity of the resulting models, the L-shaped algorithm is applied to achieve an optimal solution. In addition, a numerical example of a peer-to-peer online lending platform demonstrated the utility of the proposed modeling approach. To verify the result obtained by the L-shaped algorithm, it is compared with the CPLEX solver. Moreover, the comparison results show the accuracy and efficiency of the given method. Sensitivity analyses are undertaken to assess the impact of risk on results. And in the presence of asymmetric cost, the comparisons between the new proposed risk-averse MCCMs and the two-stage stochastic MCCMs and robust consensus models are also given. Collapse

Topics: decision making peer-to-peer lending crowdsourcing innovation management laptop computer

Methods: computational algorithm optimization model experiment sensitivity analysis conceptual modelling

Know Where to Invest: Platform Risk Evaluation in Online Lending

2022 | Information Systems Research | Citations: 3

Authors: Wang, Zhao; Jiang, Cuiqing; Zhao, Huimin

Abstract: Although enjoying rapid development, online lending also endures some unusual ri ... Expand

Abstract: Although enjoying rapid development, online lending also endures some unusual risk, that is, platform risk. As prior research on default risk evaluation in online lending largely focuses on the micro listing level, we advocate a new research problem at the macro platform level, platform risk evaluation, and explore types of information and methods that are effective in predicting platform risk. We identify four types of information, that is, platform characteristic, risk management, commercial competition, and online word of mouth, by categorizing the available features based on the aspects they reflect and examine their utilities, separately and jointly, in predicting platform risk. We also propose the use of survival analysis, especially the mixture survival model, in predicting whether and when a platform will default. Considering the essential causes and characteristics of different default events, we differentiate two types of default platforms, namely, problematic and failed platforms, and accommodate them using competing risk analysis. We carry out a cross-stage analysis using data crawled from two leading web portals for online lending in China with the two stages separated by the recent dramatic policy intervention. Our results demonstrate the competitive predictive ability of survival analysis as compared with classification-based models. The results also reveal the differences among the four identified factors in terms of predictive utility, the heterogeneity between the two types of default platforms, and differences between the start-up and stable periods of platform development. Additionally, we identify some key features using Shapley values and examine the effects of these key features. Based on the results, we derive some insights and examine the cross-stage changes and commonalities. We provide both lessons learned from the past and practical implications for market managers and lenders in the current online lending market. Collapse

Topics: online lending shapley additive explanation risk management moral hazard crowdsourcing

Methods: survival analysis classification method experiment case study machine learning

A Strategic Analysis of Algorithm Manipulation: a Lending Game perspective

2022 | International Conference on Information Systems | Citations: 0

Authors: ZHOU, Jiali; ZHENG, Jiexin

Abstract: ... Expand

Abstract: Collapse

Topics: online lending productivity

Methods: computational algorithm machine learning deep learning classification method feature engineering

Where You Live Matters: Local Bank Competition, Online Marketplace Lending, and Disparity in Borrower Benefits

2021 | Information Systems Research | Citations: 0

Authors: Alyakoob, Mohammed; Rahman, Mohammad S.; Wei, Zaiyan

Abstract: In the past decade, the proliferation of online marketplace lending has been dis ... Expand

Abstract: In the past decade, the proliferation of online marketplace lending has been disrupting the consumer credit market, especially personal loans used for debt consolidation. These lenders, for example, Lending Club, transcend the geographic boundaries within which local banks operate and offer homogeneous access and terms to borrowers. However, the ultimate benefits borrowers derive from marketplace lending can differ significantly because local alternatives may be used to replace marketplace loans when they are available and favorable. Correspondingly, if local bank competition drives the substitution of an existing marketplace loan with a traditional bank loan, the promise of equal benefits to all borrowers from marketplace lending is unlikely to fully materialize. This competitive dynamic also has implications for policy making, particularly in judging the ramifications of bank mergers and acquisitions. We utilize data from Lending Club, a major peer-to-peer (P2P) lending platform, to study whether local market structure drives heterogeneous incentives of traditional banks to convert borrowers of marketplace lending to conventional bank loans. The results indicate that a borrower who resides in a more competitive market is more likely to pay off a P2P loan early by making a large, one-time payment compared with a borrower from a less competitive market. Thus, borrowers from different markets do not benefit equally from online marketplace lending disrupting the consumer credit market. Our study has implications for online marketplace lending, other FinTech-based markets, and the consumer credit market in general. Collapse

Topics: crowdsourcing peer-to-peer model electronic market Prosper peer-to-peer lending

Methods: robustness check instrumental variables estimation probit regression experiment parametric test

Borrower Learning Effects: Do Prior Experiences Promote Continuous Successes in Peer-to-Peer Lending?

2021 | Information Systems Frontiers | Citations: 0

Authors: Chen, Qun; Li, Ji-Wen; Liu, Jian-Guo; Han, Jing-Ti; Shi, Yun; Guo, Xun-Hua

Abstract: There are a large amount of studies on the impact of individual experience on le ... Expand

Abstract: There are a large amount of studies on the impact of individual experience on learning outcome, which ignore the conditioning and implicitly assume that the learning behaviors are bound to arise. This study develops a two-stage conceptual model for the interpretation of individual learning, and aims to explore how individual’s prior experience affects the initiating of subsequent learning behavior first and then impacts the success of subsequent learning behavior. Using a large sample of loan application data over a five-year period from a Chinese P2P lending platform Renrendai, we found strong evidence that both prior successes and prior failures can motivate borrowers to take subsequent borrowing behaviors again. However, in regard to the impact of individual experience on learning outcome, our finding in online environment is contrary to the prevalent view in organizational learning—“we learn more effectively in failure than in success”. Specifically, we found only successes can lead to desirable success whereas failures can’t, which indicates that borrowers with failure experience are willing to repeat borrowing but unable to learn from their failures. Further, we find a beneficial interaction effect between borrower’s prior successes and prior failures on subsequent borrowing success. These results are robust across a variety of settings. Our study enriches the understanding of individual learning by focusing not only on outcome improvement but also on conditioning. Moreover, we investigate learning process in an online environment which is anonymous and lack of peer communication that offline organizations have, we draw some different findings. Our findings also offer practical insights for borrowers under financial difficulties and platform managers to correctly interpret prior experience and enhance the probability of subsequent funding success in P2P lending. Collapse

Topics: crowdsourcing peer-to-peer lending online lending logistics management peer-to-peer model

Methods: conceptual modelling statistical hypothesis test descriptive statistic regression analysis method robustness check

Theories: learning theory