alliance
DownloadTélécharger
Actions
Vote :
ScreenshotAperçu
Informations
Catégorie :Category: nCreator TI-Nspire
Auteur Author: msds3233
Type : Classeur 3.0.1
Page(s) : 1
Taille Size: 9.79 Ko KB
Mis en ligne Uploaded: 23/06/2025 - 20:30:09
Uploadeur Uploader: msds3233 (Profil)
Téléchargements Downloads: 1
Visibilité Visibility: Archive publique
Shortlink : https://tipla.net/a4745090
Type : Classeur 3.0.1
Page(s) : 1
Taille Size: 9.79 Ko KB
Mis en ligne Uploaded: 23/06/2025 - 20:30:09
Uploadeur Uploader: msds3233 (Profil)
Téléchargements Downloads: 1
Visibilité Visibility: Archive publique
Shortlink : https://tipla.net/a4745090
Description
Fichier Nspire généré sur TI-Planet.org.
Compatible OS 3.0 et ultérieurs.
<<
Alliant Credit Union (Alliant) had to decide whether to partner with Upstart Holdings, Inc. (Upstart), a financial technology (fintech) company that offered a platform to connect borrowers and lenders. Upstarts underwriting models used artificial intelligence (AI)/machine learning (ML) algorithms to analyze both standard financial variables and alternative data on borrowers (e.g., their education history). Studies found that Upstarts approach to underwriting resulted in fewer defaults and more approvals relative to conventional models based on credit scores. However, the use of alternative data in the underwriting process raised fair-lending concerns. In 2020, a nonprofit claimed that Upstarts use of educational variables led to discriminatory outcomes. While Upstart disputed this claim, it agreed to reform aspects of its models to ensure fairness. The case requires you to evaluate tradeoffs associated with the use of new data and technology in the underwriting process. On the one hand, Upstarts underwriting models expanded access to credit, particularly benefiting borrowers with limited credit histories. On the other hand, the use of alternative data raised fairness concerns because the variables were often correlated with factors such as race, ethnicity, and age. The case also provides an opportunity to discuss fair-lending laws in the United States, the economics of underwriting, and funding models used by fintech firms. 1. What problem is Upstart trying to solve? Upstart seeks to promote financial inclusion via enhanced access to credit. How big of a problem do you think unequal access to credit is? Only 45% of Americans have access to bank quality credit. As an alternative, many borrowers resort to relatively expensive forms of credit (e.g., payday loans, checking account overdrafts).Case Exhibit 3 shows that access to credit varies significantly by race and ethnicity. Not surprisingly, income is positively correlated with the likelihood of using credit and negatively correlated with the likelihood of being unbanked for all groups (Black, Hispanic, and White). However, within income brackets, there are striking differences across groups. For example, the use of credit by White households is approximately 25 percentage points higher than that of Black households (74.5% vs. 49.4%) for incomes between $30,000 and $50,000. Infact, the use of credit by White households in this income range is nearly the same as that of Black households making $75,000 or more. The relation between race and ethnicity and the likelihood of being unbanked displays similar patterns, with White households less likely to be unbanked than Black and Hispanic households with similar incomes. T o understand how Upstart promotes financial inclusion, it is helpful to first think about the consumer lending value chain. At a high level, the value chain consists of three primary activities: customer acquisition,underwriting, and servicing. Customer acquisition includes marketing and sales. Underwriting includes borrower risk assessment and loan pricing. Servicing includes payment processing and handling delinquencies.Upstart primarily targets underwriting, though its AI/ML models are used in other aspects as well (e.g.,customer acquisition, as noted in case T able 2). What Is the main friction that underwriters face? In a frictionless world, the riskiness ( type ) of each borrower would be perfectly observable and underwriters could accordingly price credit. Of course, many frictions existfor example, some borrowers lack credit scores or have higher uncertainty regarding future income. The main issue here is information. Specifically, underwriting is subject to information asymmetries between the borrower and the lender that can give rise to adverse selection. 2. Why does alternative data help to mitigate information frictions? A useful starting point is to compare Upstart risk grades to FICO scores. What information is used to calculate FICO scores? How does this compare to that used by Upstart? . FICO scores are calculated using a number of factors, including a borrowers payment history, available credit, negative public records (e.g., bankruptcies, liens), and length of credit history (case T able 1). Like FICO scores, Upstarts models use borrower financial information from credit report agencies, but they augment this with alternative data provided by borrowers, including education history (e.g., college major), cost of living, and loan application interactions (case page 5). Do standard financial data and alternative data measure the same thing? An important point for students to understand is that while both types of information are used to assess credit worthiness, they go about this in different ways. Many of the factors used to calculate FICO scores are backward looking (e.g., payment history, previous bankruptcies, length of credit history). A key assumption underlying the use of FICO scores is that past credit
[...]
>>
Compatible OS 3.0 et ultérieurs.
<<
Alliant Credit Union (Alliant) had to decide whether to partner with Upstart Holdings, Inc. (Upstart), a financial technology (fintech) company that offered a platform to connect borrowers and lenders. Upstarts underwriting models used artificial intelligence (AI)/machine learning (ML) algorithms to analyze both standard financial variables and alternative data on borrowers (e.g., their education history). Studies found that Upstarts approach to underwriting resulted in fewer defaults and more approvals relative to conventional models based on credit scores. However, the use of alternative data in the underwriting process raised fair-lending concerns. In 2020, a nonprofit claimed that Upstarts use of educational variables led to discriminatory outcomes. While Upstart disputed this claim, it agreed to reform aspects of its models to ensure fairness. The case requires you to evaluate tradeoffs associated with the use of new data and technology in the underwriting process. On the one hand, Upstarts underwriting models expanded access to credit, particularly benefiting borrowers with limited credit histories. On the other hand, the use of alternative data raised fairness concerns because the variables were often correlated with factors such as race, ethnicity, and age. The case also provides an opportunity to discuss fair-lending laws in the United States, the economics of underwriting, and funding models used by fintech firms. 1. What problem is Upstart trying to solve? Upstart seeks to promote financial inclusion via enhanced access to credit. How big of a problem do you think unequal access to credit is? Only 45% of Americans have access to bank quality credit. As an alternative, many borrowers resort to relatively expensive forms of credit (e.g., payday loans, checking account overdrafts).Case Exhibit 3 shows that access to credit varies significantly by race and ethnicity. Not surprisingly, income is positively correlated with the likelihood of using credit and negatively correlated with the likelihood of being unbanked for all groups (Black, Hispanic, and White). However, within income brackets, there are striking differences across groups. For example, the use of credit by White households is approximately 25 percentage points higher than that of Black households (74.5% vs. 49.4%) for incomes between $30,000 and $50,000. Infact, the use of credit by White households in this income range is nearly the same as that of Black households making $75,000 or more. The relation between race and ethnicity and the likelihood of being unbanked displays similar patterns, with White households less likely to be unbanked than Black and Hispanic households with similar incomes. T o understand how Upstart promotes financial inclusion, it is helpful to first think about the consumer lending value chain. At a high level, the value chain consists of three primary activities: customer acquisition,underwriting, and servicing. Customer acquisition includes marketing and sales. Underwriting includes borrower risk assessment and loan pricing. Servicing includes payment processing and handling delinquencies.Upstart primarily targets underwriting, though its AI/ML models are used in other aspects as well (e.g.,customer acquisition, as noted in case T able 2). What Is the main friction that underwriters face? In a frictionless world, the riskiness ( type ) of each borrower would be perfectly observable and underwriters could accordingly price credit. Of course, many frictions existfor example, some borrowers lack credit scores or have higher uncertainty regarding future income. The main issue here is information. Specifically, underwriting is subject to information asymmetries between the borrower and the lender that can give rise to adverse selection. 2. Why does alternative data help to mitigate information frictions? A useful starting point is to compare Upstart risk grades to FICO scores. What information is used to calculate FICO scores? How does this compare to that used by Upstart? . FICO scores are calculated using a number of factors, including a borrowers payment history, available credit, negative public records (e.g., bankruptcies, liens), and length of credit history (case T able 1). Like FICO scores, Upstarts models use borrower financial information from credit report agencies, but they augment this with alternative data provided by borrowers, including education history (e.g., college major), cost of living, and loan application interactions (case page 5). Do standard financial data and alternative data measure the same thing? An important point for students to understand is that while both types of information are used to assess credit worthiness, they go about this in different ways. Many of the factors used to calculate FICO scores are backward looking (e.g., payment history, previous bankruptcies, length of credit history). A key assumption underlying the use of FICO scores is that past credit
[...]
>>
