257
INTERNATIONAL ISLAMIC ECONOMIC SYSTEM CONFERENCE (I-iECONS 2021)
Self-Disclosure of Social Media Accounts And Borrowers Creditworthiness in Islamic Bank
Muhammad Syaeful Fahmi
Faculty of Economics and Business, Universitas Indonesia, Indonesia E-mail: [email protected]
Zuliani Dalimunthe
Faculty of Economics and Business, Universitas Indonesia, Indonesia E-mail: [email protected]
Abstract
There is a standard method to use credit scoring to estimate borrowers' creditworthiness in a bank's consumer loan. Recently, banks consider adding information regarding borrowers' activities on social media to calculate credit scores. This study aims to examine the relationship between social media accounts' self-disclosure with borrowers' creditworthiness in Islamic Bank. We measure self-disclosure of social media accounts variables as to whether the borrower fills their loan application form with their phone number connected to their social media account, specifically Facebook and WhatsApp. The analysis was carried out on 490 existing customer data at Islamic Bank in Indonesia. The creditworthiness probability model is estimated using a binary logistic regression model. We found that self-disclosure of phone numbers connected to borrowers' social media Facebook has a significant and positive effect on creditworthiness, but this is not the case for the phone number connected to WhatsApp. We recommend that bank management consider this finding to improve their credit scoring calculation.
Keywords: Credit Scoring; Social Media; Creditworthiness; Consumer Credit; Islamic Banking.
1. Introduction
Indonesia is a country with the largest Muslim population globally (Pew Research Center, 2019). This condition is a potential market for products and services by Islamic principles, including Islamic banking services. Since first appearing in 1991, Islamic banking in Indonesia has a market share of 6.24% in September 2020 (Indonesia Financial Services Authority, 2020). However, the level of non-performing financing was still higher than non- Islamic banking at 3.20% in the position of January 2021 (Indonesia Financial Services Authority, 2021).
Furthermore, the PwC Indonesia survey results show that in the 2013 to 2018 period, credit risk remains the main threat for banking in Indonesia to grow (PwC Indonesia, 2018).
One of the efforts to minimize credit risk is to maintain credit quality through classic creditworthiness analysis using character, capital, capacity, economic condition, and collateral analysis – also known as the 5 C method (Abrahams & Zhang, 2008). However, the classical analysis method has weaknesses, such as the inconsistency of decisions given by people with different levels of expertise. Therefore, other methods, such as credit scoring methods, are developed using statistical models to produce more efficient credit decision-making.
A credit scoring system based on statistical data processing for consumer financing has been widely used in the financial services industry, such as in banking companies (Abdou & Pointon, 2011). Other studies have also been conducted for credit scoring models for consumer financing with limited data sources (Šušteršic, Mramor, & Zupan, 2009). The development of research on credit scoring models has also utilized Big Data to evaluate potential
258
customer risks (Öztürk & Onay, 2018). The credit scoring model has also been used to project financial risk on consumer financing (Thomas, 2000). Credit scoring model development continues to be carried out by utilizing customer's social media (Tan & Phan, 2018).
The increase in social media users is driven by the increase in internet penetration, which has reached 73.7% of Indonesia's population in the second quarter of 2020 (Irawan, Yusufianto, Agustina, & Dean, 2020). In January 2020, active social media users in Indonesia increased 8.1% from April 2019, and the majority accessed via mobile devices. Based on Indonesia Digital Data 2020, most social media platforms are YouTube, WhatsApp, and Facebook (Hootsuite, 2020).
The increase in community activity in using social media has attracted the attention of researchers to use social media-related activity data in predicting credit default. Recently, social media data can predict credit default on online peer-to-peer Islamic lending in Indonesia (Khilfah & Faturohman, 2020). Previous research has also been conducted in utilizing social media information to predict and prevent credit default in peer-to-peer lending companies in China (Ge, Jeng, Gu, & Zhang, 2017).
Social media data to strengthen the credit scoring model is an alternative to speed up the process, evaluate the credit risk, increase the ability to predict creditworthiness, and even reach customers who do not have a credit history (Thomas, 2000; Berger, Frame, & Miller, 2005; Abdou & Pointon, 2009). However, research using social media data to build a credit scoring model is still limited to peer-to-peer lending companies. Recent research on peer-to-peer online finance companies states that soft information in social network data can improve the screening of potential finance recipients with low guarantees and risks (Liu, Shang, Wu, & Chen, 2020). However, social media data to identify creditworthiness in the Islamic banking industry has not been studied. Therefore, this research can provide valuable insights for the Islamic banking industry in consumer credit assessment with the following hypothesis:
H1. The self-disclosure of Facebook accounts has a significant influence on creditworthiness.
H2. The self-disclosure of WhatsApp accounts has a significant influence on creditworthiness.
2. Methodology & Data 2.1 Literature Review
The institution that formulates standards and guidelines for general banking supervision, The Basel Committee, states that credit risk is the potential failure of bank borrowers to fulfill their obligations by agreed terms (Basel Committee, 2000). Therefore, in carrying out their business activities, Islamic Banks must carry out credit management properly. Credit management is essential to maintain banking stability and performance in the long term, managing credit risk to get lower (Supiyadi, Dodi, & Machmud, 2017; Lassoued, 2018). The importance of credit risk management is inseparable from its impact on reputation and financial performance. According to a study that examines the relationship of credit risk with the profitability and liquidity of Islamic banks in Indonesia, credit risk has a significant and negative impact on the profitability of Islamic banks (Supriyadi & Nugraha, 2018).
Activities of channeling funds by Islamic banks in the form of consumer financing will be recorded as productive assets in the financial statements of Islamic banks (Indonesia Financial Services Authority, 2014;
Bidabad & Allahyarifard, 2019). According to the regulations of the Financial Services Authority of Indonesia, productive assets are defined as "the investment of Bank funds in both rupiah and foreign currency to generate income, in the form of financing, sharia securities, placements with Bank Indonesia and the government, claims for Islamic securities purchased with a promise to be resold (reverse repurchase agreement), acceptance receivables, derivative receivables, investments, placements with other banks, off-balance-sheet transactions, and other forms of provision of funds that are comparable to them" (Indonesia Financial Services Authority, 2014).
In practice, Islamic Banks classify the quality of productive assets in financing into five collectability groups.
The grouping consists of earning assets in the categories Current, Special Mention, Substandard, Doubtful, or Loss (Indonesia Financial Services Authority, 2014). If it is related to the number of days in arrears, then the financing in the default category is financing with the earning assets group of Substandard, Doubtful, or Loss due to having arrears on claims 90 days or more.
Analyzing the feasibility of consumer financing in Islamic banking in Indonesia has used many system assistances in credit scoring. The credit scoring system is used to measure a person's ability and willingness to pay based on relevant risk factors (Abdou & Pointon, 2009; Bolton, 2009; Djeundje, Crook, Calabrese, & Hamid, 2021).
259
Recent research on the modification of predictors used in credit scoring shows that the form of psychometric data and other alternative data related to the characteristics of email use can improve the prediction accuracy of the credit scoring model (Djeundje, Crook, Calabrese, & Hamid, 2021).
Islamic Bank must be based on the principle of prudence in terms of financing the customers and sharia principle (Indonesia Financial Services Authority, 2014). Therefore, methods commonly used in feasibility analysis of financing disbursement must pay attention to at least five factors: character, capital, capacity, economic condition, and Collateral or often referred to as the 5 C method (Abrahams & Zhang, 2008; Wasiuzzaman, Nurdin, Abdullah,
& Vinayan, 2020; Fernando & Siagian, 2021).
The thing that is a challenge for Islamic banking from the five factors that must be analyzed is the character factor. The problem of limited financial information to assess the character of customers in the credit analysis process has been resolved through the use of currently developed technology (Wei, Yildirim, Van den Bulte, &
Dellarocas, 2016; Kshetri, 2016). The development of technology and increasing internet penetration in Indonesia can solve these challenges. One of them is by conducting social assessments through social media information.
Previous research has shown that social assessment can reduce lender doubts by looking at character through social network information and a broader view of consumers' social status (Wei, Yildirim, Van den Bulte, & Dellarocas, 2016).
As a parameter in assessing creditworthiness, social media data has been carried out by Lenddo, a scoring company, and identity verification technology (Groenfeldt, 2015). In addition, several other studies that utilize social media data in the credit scoring model show that credit scoring techniques based on social networks can increase predictive ability by 14.6% compared to traditional non-network methods (Tan & Phan, 2018). Also, some studies reveal that non-credit variables such as internet usage behavior can provide additional information beyond basic demographic information such as income and age in predicting credit default (Wu, Xu, Zhao, & Liu, 2020).
3. Research Method & Data
The type of this research is causal explanatory with a quantitative approach, which studies the relationships between variables tested. In this research, the relationship between the self-disclosure of social media accounts and creditworthiness will be analyzed. Therefore, the regression model used to find the relationship between a binary dependent variable and the independent variable is the logistic regression model (Hosmer & Lemeshow, 2000). The research was conducted using a sample with data sources of consumer financing customers in Islamic banks. The sampling technique is carried out by taking customer data based on the collectability quota of current customers, having arrears, and defaults so that this method is a nonprobability sampling technique through purposive sampling method, quota sampling type (Cooper & Schindler, 2014).
Four hundred and ninety customer data have been successfully obtained with information in the customer name, gender, date of birth, mobile phone number, and collectability. All data prepared are processed based on the variables tested using the binary logistic regression method using SPSS software. The operationalization of each variable used in this study is shown in Table 1.
Table 1. List of operationalization of research variables.
Variables Type Description References
Creditworthiness IV Creditworthiness status from collectability transformation
(Arrears/Non-arrears). Abdou et al., 2016; Khilfah
& Faturohman, 2020 Ln(Age) CV The natural logarithm of the age, calculated from the date
of birth obtained from the data source until 31 December 2020 (in years).
Khilfah & Faturohman, 2020
Gender CV Male/Female Khilfah & Faturohman,
2020 Debt Service
Ratio (DSR) CV The ratio between total instalments (in Rupiah) divided by net income (after deducting installments). This ratio is expressed as a percentage and obtained directly from the Bank system.
Abdou, Tsafack, Ntim, &
Baker, 2016
Ln(Max) CV The natural logarithm of the maximum credit value. Abdou, Tsafack, Ntim, &
260
Variables Type Description References
Baker, 2016 LinkFB DV The connection of the mobile phone number registered to
the Bank's system with the Facebook account. Ge et al., 2017 LinkWA DV The connection of the mobile phone number registered to
the Bank's system with the WhatsApp account. Ge et al., 2017 Notes: IV is independent variable, CV is control variable, and DV are the dependent variable.
4. Result & Discussion
In this section, we first describe descriptive statistics of the data we analyzed. Table 2 shows that the range of each variable with a ratio scale varies with the largest variation in the DSR variable. Furthermore, the frequencies of each of the other variables in the form of dummy variables and having a nominal scale are shown in Table 3.
Table 2. Descriptive statistics.
Variables N Min Max Average St. Dev Var
Ln(Maks) 490 15 22 18.79 1.008 1.016
DSR 490 3 50 28.57 10.040 100.797
Ln(Age) 490 3 4 3.83 0.379 0.144
Table 3. Dummy variable frequency.
Variable 0 1 N
Gender 92 398 490
Gender (%) 18.78% 81.22%
LinkFB 256 234 490
LinkFB (%) 52.24% 47.76%
LinkWA 197 293 490
LinkWA (%) 40.20% 59.80%
Creditworthiness 246 244 490
Creditworthiness (%) 50.20% 49.80%
Five other dummy variables will be tested using logistic regression. These variables are qualitative variables that will be tested for their effect on continuous variables. Therefore, the qualitative variables are quantified in categorical form, namely the variable Gender (0 = female, 1 = male), LinkFB (0 = mobile phone number not connected to FB, 1 = mobile phone number connected to FB), LinkWA (0 = mobile phone number not connected to WA, 1 = mobile phone number connected to WA), Creditworthiness (0 = there are arrears, 1 = no arrears).
The frequency data of each variable is in Table 3. shows that the male gender dominates the gender variable.
According to the best practices in proposing consumer financing in banks, this condition is that men dominate as customers. This condition is because the Bank will assess the repayment capacity of a prospective customer based on their source of income. Therefore, as a source of income recipients, customers with the male gender dominate the application for consumer financing to the Bank. Meanwhile, other variables, namely LinkFB, LinkWA, and Creditworthiness, have almost the same frequency portion of the total sample observed.
The prediction model is obtained from the test results using a binary logistic regression model for all independent variables and their effects on the dependent variable. The resulting model for predicting customers in default (in arrears) based on the variables that significantly affect the dependent variable is shown in Equation (1).
(1) Where:
261
Logit(Creditworthiness) is creditworthiness probability, X1 is LinkFB variable, X2 is LinkWA variable, X3 is Ln(Age) variable, X4is Gender variable, X5 is DSR variable, and X6 is Ln(Maks) variable.
The test results using the binary logistic regression model are shown in Table 4. In addition, statistical testing was carried out to produce a model that has a good predictive ability. Tests conducted include the fit model test, classification accuracy test, determination coefficient test, and partial test (Pampel, 2000).
Table 4. Result of the binary logistic regression model Covariate Logistic
coefficient Standard error value Odds ratio 95% CI
Intercept -2.585 0.352 0.454 0.075 57.68 – 228.88
LinkFB 4.744 0.332 0.000 114.903 0.54 – 2.00
LinkWA 0.043 0.449 0.897 1.044 0.45 – 2.60
Ln(Age) 0.075 0.411 0.867 1.078 0.66 – 3.32
Gender 0.395 0.017 0.337 1.484 0.92 – 0.98
DSR -0.053 0.170 0.003 0.949 0.77 – 1.50
Ln(Max) 0.073 3.449 0.668 1.075
Hosmer-Lemeshow goodness-of-fit test: = 0.159
Additional testing of the model fit was carried out using the overall model fit method. Testing is done by calculating the difference between -2 initial Log-Likelihood (without using test variables) and using test variables.
The difference between -2 Log Likelihood is 401,728. With the Chi-Square value in the table (df = 6, 0.05) of 12.592. So that because the difference in -2 Log Likelihood is greater than the Chi-Square value in the table, the variables can improve the logistic regression model (Pampel, 2000).
The prediction accuracy of the regression model was tested using the classification accuracy test. The results of the classification accuracy test are shown in Table 5.
Table 5. Classification accuracy data.
Observed
Predicted
Creditworthiness Percentage Correct Default Pass
Creditworthiness Default 231 15 93.9
Pass 28 216 88.5
Overall Percentage 91.2
Based on the data in Table 5, it is found that the overall classification accuracy of the Logistic Regression Model is 91.2%. This result shows that the accuracy in predicting the effect of the independent variables on the dependent variable is 91.2%. The Nagelkerke's R Square shows the magnitude of the influence of the independent variables X1, X2, X3, X4, X5, and X6 (simultaneously) affects the dependent variable Y by 74.6%.
The logistic regression model results can complement existing credit scoring models and a model for stand-alone predictions. In addition, this study shows the independent variable that has a significant effect on the probability of default compared to the control variable that has been used as one of the variables in credit scoring in the Bank, namely DSR.
The default probability value shows the creditworthiness of consumer financing distribution. The probability result is close to number one (maximum 1), indicates that the customer/prospective customer has a smaller chance of arrears or good creditworthiness. On the other hand, the probability result is close to 0 (at least 0), indicating that the customer or prospective customer has a greater chance of arrears or has poor creditworthiness.
Based on these results, it is known that the variable connectedness of the customer's mobile phone number with their Facebook account has a significant and positive effect on creditworthiness. Based on these results, it is known that the first hypothesis of this study, namely "there is a significant influence between the self-disclosure of Facebook social media accounts on creditworthiness," can be proven.
The connection between the mobile phone number and the Facebook account is a proxy for the Bank customer's self-disclosure of their social media account. This connection is based on customer's openness in conveying social
262
information described through their social media accounts. This result in line with previous research that borrower's self-disclosure of their social media account and their social media activities can be used to predict borrower's default probability (Ge, Jeng, Gu, & Zhang, 2017). The customer's mobile number and Facebook account connection show that the customer shows an open nature to their social environment information. However, this is not directly related to their ability to pay, as explained in another study which states that unlike traditional predictors such as household income, social network information might not directly capture one's ability to repay but valuable to help infer one's creditworthiness (Tan & Phan, 2018). This assessment can help with the weakness of the behavioral scoring model, which can only be applied to accounts that have sufficient data on customer characteristics (Thomas, 2000).
The Odds Ratio of the LinkFB variable shows the opportunity for customers who have a connected mobile phone number with a Facebook account to have a creditworthiness opportunity (Y = 1) of 57,68 to 228,88 times.
This result in line with previous research, which explains that disclosure of information on social media accounts shows a lower probability of default (Ge, Jeng, Gu, & Zhang, 2017).
The nature of openness is an essential parameter in assessing the character of a customer. In the classical analysis method using the 5 C method, the character factor is critical for analyzing creditworthiness. This result is consistent with previous research, suggesting that character assessment is the second most important in the 5 C creditworthiness analysis after the capacity assessment (Peprah, Agyei, & Oteng, 2017). The customer's character who is open in conveying correct and valid information determines the study of other factors, namely capital, capacity, conditions, and collateral, to be reliable. Identifying the nature of openness has also been proven through research that digital records a person's behavior on social media (Kosinski, Stillwell, & Graepel, 2013).
The use of the variable connectedness of a mobile phone number with a Facebook account is also practically possible in the credit scoring model in banking. Therefore, this finding can improve the measurement of a person's ability and willingness to pay based on risk factors following existing credit scoring predictors (Abdou & Pointon, 2009; Bolton, 2009; Djeundje, Crook, Calabrese, & Hamid, 2021). The applicability of this finding is based on that mobile phone number information has become one of the mandatory information requested by the Bank in the credit application form. In addition, testing the connection of the customer's mobile phone number with their Facebook account does not have the potential to violate the principle of privacy because the Bank does not collect the information listed on their social media accounts but only ensures the connection of the mobile phone number with their Facebook account.
Another independent variable analyzed in this study is the connection between the customer's mobile phone number and the WhatsApp account. Based on the results of statistical tests through partial test analysis, the influence and significance of this variable indicate that there is no significant relationship. Therefore, based on these results, it is known that the second hypothesis of this study has not been proven.
These results explain that the variable connection of a customer's mobile number with a WhatsApp account cannot describe creditworthiness. Furthermore, this result indicates that WhatsApp social media account is considered a general social media account and is not private, so the disclosure of WhatsApp account information cannot describe the character of customer openness. The findings of this study reveal that the use of social media data, significantly Facebook, can help assess creditworthiness in managing credit risk as part of credit management (Supiyadi, Dodi, & Machmud, 2017; Lassoued, 2018).
5. Conclusion
Research on the effect and significance of self-disclosure of social media accounts on consumer financing performance has been carried out. Some of the findings from this study can conclude that statistically, self- disclosure of Facebook social media accounts shows a significant and positive effect on consumer financing performance in Islamic banks. On the other hand, the self-disclosure of WhatsApp social media accounts does not significantly affect the performance of consumer financing in Islamic banks. Thus, the use of social media data such as self-disclosure of customers' Facebook accounts can be considered by Islamic banking management. The utilization of social media data can strengthen their financing assessment system in managing credit risk.
Some limitations in this study include data from one Islamic Bank and the period of the observations are carried out at one time. Therefore, future research can use data samples from several Islamic banks and observations to continue this research. Thus, using social media account data as a credit scoring system in Islamic banking can minimize credit risk.
263 References
Abdou, H. A., & Pointon, J. (2009). Credit scoring and decision making in Egyptian public sector banks. International Journal of Managerial Finance, 5(4), 391-406.
Abdou, H. A., & Pointon, J. (2011). Credit scoring, statistical techniques and evaluation criteria: A review of the literature. Intelligent Systems in Accounting, Finance & Management, 18(2-3), 59-88.
Abdou, H. A., Tsafack, M. D., Ntim, C. G., & Baker, R. D. (2016). Predicting creditworthiness in retail banking with limited scoring data.
Knowledge-Based Systems, 103, 89-103.
Abrahams, C., & Zhang, M. (2008). Fair Lending Compliance. New Jersey: John Wiley & Sons, Inc.
Basel Committee. (2000). Principles for the Management of Credit Risk. Basel: Basel Committee on Banking Supervision.
Berger, A. N., Frame, W. S., & Miller, N. H. (2005). Credit Scoring and the Availability, Price, and Risk of Small Business Credit. Journal of Money, Credit and Banking, 37(2), 191-222.
Bidabad, B., & Allahyarifard, M. (2019). Assets and Liabilities Management in Islamic Banking. International Journal of Islamic Banking and Finance Research, 3(2), 32-43.
Bolton, C. (2009). Logistic regression and its application in credit scoring. Pretoria: University of Pretoria.
Cooper, D. R., & Schindler, P. S. (2014). Business Research Methods. New York: McGraw-Hill.
Djeundje, V. B., Crook, J., Calabrese, R., & Hamid, M. (2021). Enhancing credit scoring with alternative data. Expert Systems with Applications, 163, 113766-113777.
Fernando, E., & Siagian, P. (2021). Proposal to use the Analytic Hierarchy Process Method Evaluate Bank Credit Submissions. Procedia Computer Science, 179, 232-241.
Ge, R., Jeng, J., Gu, B., & Zhang, P. (2017). Predicting and deterring default with social media information in peer-to-peer landing. Journal of Management Information Systems, 34(2), 401-424.
Groenfeldt, T. (2015). Retrieved from Forbes: https://www.forbes.com/sites/tomgroenfeldt/2015/01/29/lenddo-creates-credit-scores-using-social- media/?sh=20c951c92fde
Hootsuite. (2020). Digital 2020 Indonesia. Vancouver: Hootsuite.
Hosmer, D. W., & Lemeshow, S. (2000). Applied Logistic Regression . New Jersey: John Wiley & Sons.
Indonesia Financial Services Authority. (2014). Peraturan Otoritas Jasa Keuangan No.16/POJK/03/2014 Tentang Penilaian Kualitas Aset Bank Umum Syariah dan Unit Usaha Syariah. Jakarta: Lembaran Negara RI.
Indonesia Financial Services Authority. (2020). Snapshot Perbankan Syariah Indonesia 2020. Jakarta: Otoritas Jasa Keuangan.
Indonesia Financial Services Authority. (2021). Statistik Perbankan Syariah Januari 2021. Jakarta: Otoritas Jasa Keuangan.
Irawan, A. W., Yusufianto, A., Agustina, D., & Dean, R. (2020). Laporan Survei Internet APJII. Jakarta: Indonesia Survey Center.
Khilfah, H. N., & Faturohman, T. (2020). Social media data to determine loan default predicting method in an islamic online P2P lending.
Journal of Islamic Monetary Economics and Finance, 6(2), 243-274.
Kosinski, M., Stillwell, D., & Graepel, T. (2013). Private traits and attributes are predictable from digital records of human behavior. Proceedings of the National Academy of Science of the United States of America, 110, pp. 5802-5805. Washington DC.
Kshetri, N. (2016). Big data’s role in expanding access to financial services in China. International Journal of Information Management, 36, 297- 308.
Lassoued, M. (2018). Comparative study on credit risk in Islamic banking institutions: The case of Malaysia. The Quarterly Review of Economics and Finance, 70, 267-278.
Liu, Z., Shang, J., Wu, S.-y., & Chen, P.-y. (2020). Social collateral, soft information and online peer-to-peer lending: A theoretical model.
European Journal of Operational Research, 428-438.
Öztürk, E., & Onay, C. (2018). A review of credit scoring research in the age of Big Data. Journal of Financial Regulation and Compliance, 26(3), 382-405.
Pampel, F. C. (2000). Logistic Regression: A Primer. California: Sage Publications, Inc.
Peprah, W. K., Agyei, A., & Oteng, E. (2017). Ranking The 5C’s Of Credit Analysis: Evidence From Ghana Banking Industry. International Journal of Innovative Research and Advanced Studies, 4(9), 78-80.
Pew Research Center. (2019). Retrieved from pewresearch.org: https://www.pewresearch.org/wp- content/uploads/2019/03/FT_19.03.29_muslimChristianPopulations_muslim.png
PwC Indonesia. (2018). 2018 Indonesia Banking Survey . Jakarta: PwC Indonesia.
Supiyadi, D., Dodi, & Machmud, A. (2017). Factors that Influences credit risk of financing institution in Indonesia. In A. G. Abdullah, A. B.
Nandiyanto, N. Budiwarti, I. Waspada, A. Machmud, L. Permana, & Y. Rohmana (Ed.), Proceedings of the 2nd International Conference on Economic Education and Entrepreneurship (pp. 159-163). Bandung: Scitepress.
Supriyadi, D., & Nugraha, M. A. (2018). The determinants of bank profitability: Empirical evidence from Indonesian sharia banking sector. In A.
G. Abdullah (Ed.), Advances in Economics, Business and Management Research (pp. 21-26). Bandung: Atlantis Press.
Šušteršic, M., Mramor, D., & Zupan, J. (2009). Consumer credit scoring models with limited data. Expert Systems with Applications, 36, 4736- 4744.
Tan, T., & Phan, T. Q. (2018). Social media-driven credit scoring: the predictive value of social structures. Retrieved from https://ssrn.com/abstract=3217885 or http://dx.doi.org/10.2139/ssrn.3217885
Thomas, L. C. (2000). A Survey of credit and behavioural scoring: forecasting financial risk of lending to consumers. International Journal of Forecasting, 16, 149-172.
Wasiuzzaman, S., Nurdin, N., Abdullah, A. H., & Vinayan, G. (2020). Creditworthiness and access to finance of SMEs in Malaysia: do linkages with large firms matter? Journal of Small Business and Enterprise Development, 27(2), 197-217.
Wei, Y., Yildirim, P., Van den Bulte, C., & Dellarocas, C. (2016). Credit scoring with social network data. Marketing Science, 35(2), 234-258.
Wu, W., Xu, D., Zhao, Y., & Liu, X. (2020). Do consumer internet behaviours provide incremental information to predict credit default risk?
Economic and Political Studies, 8(4), 482-499.