Simulation vs. Statistics
Statistics based approach
- Analyze the system on surface level
- Does not consider the underlying processes
- The estimation error is increasing as we getting further from the sample Simulation
|
|
Simulation
- Understand the problem at bottom level
- Consider every important step in a process
- Modeling error is constant
|
Credit Risk measurement with Microsimulation
The basic idea is to:
- determine the process of a credit repayment considering the possible reasons of a default
- analyze the system on individual, or close to individual level
- calibrate the simulation to fit the observations
The final simulation is used to:
evaluate the macroeconomic variables’ effects on PD and LGD rate
Decision model - Structure
The decision model defines the behavior of the customers (based on experts):
- Households decide the amount of consumption, saving and credit repayment
- Planning to multiple time period
- Firms decision making methods are similar: they consider the profit rate and the liquidity as well
- Customers are connected through endogen variables
- The outcome of the decision model: financial decision
The figure shows regression curve, that is fitted on the observations, may fail to provide appropriate response on the
out of sample section
Decision Model - Segmentation
The customers are grouped based on the following points:
- Regional variables
- Income / Profit rates
- Economic sectors
|
|
Internal segmentation is also used
- Risk aversion
- Forecast ability
- Liquidity
|
Repayment model - Structure
For a given saving / repayment decision
- The income shocks may result postponed credit repayment
- The postponed repayments sums up
- High level of accumulated repayments causes default
The repayment model simulates the queue of repayments in each and every segment, while it provides aggregate information of the PD and LGD
|
