Financial Econometrics and Empirical Market Microstructure
General Scheme of the Model Workflow
The model workflow includes the following steps:
1. The AR(1)-GARCH(1,1) model estimation of each risk-factor (i. e. returns on each stock).
2. Distribution construction for historical St (the stochastic component of the AR - GARCH model error) with a Gaussian smoothing kernel and Pareto distribution for the tails.
3. A constructed distribution for St used for t-copula identification.
4. t-copula used for generating values of St during the forecast period.
5. Build return forecast for each stock based on the identified AR-GARCH model (step 1) and generated St (step 4) N times via the Monte-Carlo approach.
6. Calculation of the profit-loss profile and risk metrics values based on results from step 5.
We analyzed two different use case scenarios:
• Basic scenario: no changes in conditions; we just built a forecast for the next 30 days and calculated a profit-loss profile for the portfolio on the 30th day.
• Stress-scenario: we simulated a 40 % idiosyncratic drop in GMKN stock, built a forecast for the next 10 days after this drop, and after that calculated a profit-loss profile for the portfolio on the 10th day.
Results of the stress tests are shown in Fig. 6 and Table 3.
