Assignment to be entirely done in Jupyter Notebook. It is to be done using data in the attached zip file.
25752 Bank Lending and Analytics Individual Assignment Instructions • This assignment is to be completed individually • This assignment is to be submitted in the lecture of Week 12 • All assumptions should be stated • Use the data set mortgage for all questions • The assignment has to be done in Python. Please ask for prior approval if you consider using other packages • This assignment accounts for 70% of the total marks in the subject. Total marks is 70 • Page limit is one page per 5 marks for a total of approximately 14 pages • The submissions will be checked for plagiarism using Turnitin. Make sure every code and word is written by yourself! Problems 1. PD modelling [20 marks] a. Estimate a basic credit risk model for mortgage default probabilities (PD) (you may choose a logit or a probit model). Include two standard explanatory variables which are FICO and LTV at origination. Compute the estimated PD for all mortgage loans and periods. Plot the average probability of default by time. Provide your code, output for the model and interpret the output. [5 marks] b. Extend the model to account for loan age. Compute the estimated PD for all mortgage loans and periods. Plot the average probability of default by loan age. Provide your code, output for the model, the plot and interpret the output. [5 marks] c. Suggest two additional variables from mortgage data that you think can explain for mortgage default. Explain the rationale (relation to default probabilities). Estimate the PD model again by including explanatory variables in part (a), part (b) and your suggested variables in one regression. Compute the estimated PD for all mortgage loans and periods. Plot the average probability of default by loan age and period in two separate charts. Provide your code, a output for the model, the plots and interpret the output. [5 marks] d. Compare the accuracy of three models from steps 1a to 1c. Present and explain your findings with regard to model accuracy? [ 5 marks] 2. LGD modeling [15 marks] a. Compute the loss rate given default (LGD) for all mortgage loans and periods from the current loan to value ration LTV assuming a 40% house price decline for all loans in all future periods and no repayment of the outstanding loan amount. Compute the hypothetical LGD for all mortgage loans and periods. Plot a histogram for the distribution of the LGD. Provide your code, plots, and analyse the output. [5 marks] b. Run a linear regression model to predict LGD. Include the same set of explanatory variables used in the question 1c. Compute the estimated LGD for all mortgage loans and periods. Plot the average LGD by period. Provide your code, output, plot and interpret the output. [5 marks] c. Name two factors that may impact LGD next to house prices [5 marks]. 3. Expected loss calculation [5 marks] Compute the level of expected one-period loss for all mortgage loans and periods. PD should be inferred from question 1c and LGD should be inferred from question 2b. You may set the exposure to default to one unit or use actual loan amounts. Plot the average expected loss by loan age and time period in two separate charts. Provide your code, plots and analyse the output. 4. Bank capital allocation [5 marks] Compute the Basel capital ratio for all mortgage loans and periods using the internal ratings based approach. Assume correlation at 15%. PD should be inferred from question 1c and LGD should be inferred from question 2b. You may set the exposure to default to one unit or use actual loan amounts. Plot the average capital ratio by loan age and time period in two separate charts. Provide your code, plots and analyse the output. [10 marks] 5. IFRS 9 loan loss provisioning [25 marks] Please read the attached paper on IFRS 9 modelling: Krüger, S., Rösch, D. and Scheule, H., 2018. The impact of loan loss provisioning on bank capital requirements. Journal of Financial Stability, 36, pp.114-129. a. What are the main differences for the loan loss provisioning under IFRS 9 and US GAAP?[5 marks] b. Compute the level of expected lifetime loss (current expected credit loss, CECL) for all loans making simplifying assumptions (e.g., assume that the remaining time to maturity is 10 for all loans, time constant PDs and LGDs, the discount rate to be equal to the mortgage loan rate). You may set the exposure to default to one unit or use actual loan amounts. Plot the average current expected credit loss by loan age and time period in two separate charts. Provide your code, plots and analyse the output. [10 marks] c. Explain two effects of COVID-19 on IFRS 9 or US GAAP loan loss provisions? Define loan deferment and detail the response of the Australian Prudential Regulation Authority on loan loss provisioning and bank capital calculations in relation to COVID-19 loan deferments [10 marks]