FinTech Methods and Practice
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
FinTech Methods and Practice
Final Exam: Payday and Installment Loans, FL Regulation
Due: May 10, 2023 (midnight CST)
• Please submit a write-up of results, along with your R code (R Markdown optional). You can use a different statistical programming software, e.g., Stata, if you like.
• Answer the questions in the order given here.
• This final exam has to be done individually. No help can be solicited or provided.
• 150 points (10 bonus points).
Introduction
In this assignment, you will use a mock dataset resembling credit report data from Clarity Services, the largest Fair Credit Reporting Act (FCRA)-regulated credit bureau to focus on the “thin credit” segment. Clarity covers consumers that have only very “thin”credit reports from mainstream credit bureaus, these are typically low-income consumers that rely on payday loan providers for example.
You want to understand whether switching from payday loans (usually short-term, high-cost loans, generally for $500 or less, that is typically due on your next payday) to more traditional installment loans with larger loan amounts, longer duration, and regular repayments, helps consumers (using delinquencies as your main outcome variable). To do so, you use a regulation in FL that was passed in March 2018 and put into place on 1st of July 2019. The regulation wllowed lenders to make “installment”loans up to $1000 (previously was $500), apart from the payday loans they originally offer. The regulation mandates at least 2 weeks between payments, not more than one year. The repayment is over 60 to 90 days (previously it was 7 to 31 days). Maximally 8% of the loan balance can be charged in interest over two weeks (previously, it was 10% max per $100 borrowed, which corresponds to a max APR of 304%).
On Canvas, under Files, you can find the mock dataset. Credit report data always consists of a snapshot of all tradelines, i.e., all information on the approved loans. Here is a data description:
• portfolio: type of loan portfolio the tradeline is for
• loan: account type, type of loan classification
• account: account status, current condition of the account
• duration: terms duration, the expected number of periods for the duration of the loan,“period”is how often the payment becomes due
• due: terms frequency, the frequency period of payments due: e.g. the payment is due biweekly (7) or quarterly (90)
• open_date: the date the account was originally opened
• close_date: the date the account was closed to further purchases, paid in full, sold, or Charged-off
• first_due_date: the date the first payment is expected
• last_payment_date: the date of the most recent consumer payment received
• delinquency_date: the first reported date of the delinquency of the loan, should only be filled in while the payment is still past due, once the payment is paid this should be null, the next delinquency overwrites the last
• highest_credit: the original amount of the loan, or highest balance attained, this excludes fees and interest payments
• current_balance: current balance owed on the account, should include principal, current interest due, late charges, and fees
• expected_payment: scheduled payment, the whole positive dollar amount expected for the next scheduled payment
• amount_past_due: the total amount of payments due based on delinquency, this should include late charges and fees (if applicable) that are past due
• charge_off: original charge-off, the original amount charged off due to loss by the lender
• rating: payment rating, display a single character value that properly identifies whether the account is current, past due, in collections, or charged-off within the activity period reported
Tasks:
1) [15 points] Load the mock credit report data, order the variables to get familiar with the dataset. Turn all the date variables into a format that R can understand. Make new variables that document the month, week(the Nth week in the year), year-week, year-month of the opening, closing, first due, last payment, and delinquency dates of each tradeline. Hint: we recommend using the following packages
- data.table, dplyr, lubridate, tsibble, and tidyverse - to make your work easier. You can use either read.csv() or read.delim2() to load the dataset in R. as.Date can be used to set the dates, and you can use the week, year, month, yearmonth, and yearweek functions to get the other date variables.
2) [15 points] Make two new variables: the log of highest_credit and the percent deviation from individual-level means of the highest_credit variable. Hint: to generate the mean variable, you’ll need to use a combination of group_by(id) and arrange() in a tibble structure with %>%. Use an ifelse() statement to generate the deviation variable. To generate the log of the highest_credit, use log(abs(CreditReportData$highest_ credit)+1). The percent deviation from individual-level means of the highest_credit variable is: for each id group, the highest_credit divided by the id’s mean of highest_credit. Assign 0 if the denominator is zero.
3) [10 points] Create a dummy variable for when a given loan/tradeline is opened after the treatment, i.e., after the FL regulation has taken place (July 1st, 2019). Hint: you can use an ifelse() statement to generate the dummy variable. Conditions can be combined with &.
4) [10 points] Create dummy variables for when a given loan/tradeline is either an installment loan, a real-time installment loan, or a single-payment “payday”loan.
5) [15 points] Plot the number of installment and payday loans that were opened in any given week over the sample period. Then plot the share of each of the loan types in all loans for any given week over the sample period. Hint: use the collapse library to generate a new dataset with a total of loans in each category. Use ggplot with geom_line() option for the plots.
6) [15 points] Generate a dummy variable as the treatment intensity, i.e., 1 if the individual resides in Florida and 0 otherwise. Also generate dummy variables for when a loan is delinquent and charged off.
7) [15 points] Regress whether a new loan is an installment loan or a payday loan on the treatment * treatment intensity, i.e., the interaction of the FL state and post regulation dummies controlling for individual and week-of-opening-of-the-loan fixed effects. Hint: Use the lfe library and the following regression structure: felm(regression formula | fixed effects | IV formula | variables for clustering).
8) [15 points] Regress the duration of the loans and the highest credit (all three variables measuring highest credit) on the triple interaction of treatment * treatment intensity * installment loan controlling for individual and week-of-opening-of-the-loan fixed effects. Also regress whether a loan is delinquent on the triple interaction.
9) [15 points] Make density plots for loan amounts and repayment amounts pre/post regulation, Florida versus other states, and installment versus payday loans. Hint: create a subset of the dataset with subset() and required conditions. In ggplot, use geom_density(aes(color = as.factor(dummy for pre/post FL regulation))). Expected payment is on the x-axis.
10) [15 points] Explain why you cannot compare, e.g., delinquencies, of installment versus payday loan borrowers directly, but instead need the FL regulation to answer the research question of whether installment loans reduce delinquencies.
11) [10 bonus points] Is there a remaining endogeneity concern with using the triple interaction of treatment * treatment intensity * installment loan as your regressor?
2023-05-08
Payday and Installment Loans, FL Regulation