This is the GitHub repository for a research project on establishing a robust methodology to forecast the correlations between the risk factors most likely to impact Aviva UK’s balance sheet over a 1-year period. This project was hosted by the London School of Economics and Political Science's Department of Statistics, in collaboration with Aviva UK.
We identified a total of five risk factors which we sought to model, namely interest rates, the yield on high-quality corporate bonds, the risk-free rate, inflation, and Gross Domestic Product (GDP). In addition, two statistical models were used to forecast rolling correlations among pairs of risk factors: 1) an ARMA model based on a rolling window, and 2) the DCC-GARCH model. We found the DCC-GARCH model to be the superior method to forecast these correlations.
The final results are in 'Final_Report.pdf'
Team members: Hudson Yeo, Jeria Kua, Nicholas Hutter, Gábor Ratkovics, Zain Hussein
Awarded Overall Champion of the Challenge.
LSE Practitioners Challenge
├── Code
│ ├── DCC_Forecasts.R
│ ├── corrfx.py
│ ├── notebooks
│ │ ├── arma_testing.ipynb
│ │ ├── data_environment_and_testing.ipynb
│ │ └── forecast_errors.ipynb
│ └── stationarity_test.R
├── Data
│ ├── FTSE100_closing_data.csv
│ ├── UK_GDP_data.csv
│ ├── UK_gilt_yield_data.csv
│ ├── UK_inflation_data.csv
│ ├── UK_risk_free_rate_data.csv
│ ├── forecasts.csv
│ ├── forecasts_t.csv
│ ├── processed_data.csv
│ └── processed_linear.csv
├── Papers
│ ├── DCC Paper 4.pdf
│ ├── DCC Paper 5.pdf
│ ├── Dynamic Conditional Correlation.pdf
│ ├── GARCC paper.pdf
│ ├── Giovannini2006_Article_ConditionalCorrelationsInTheRe.pdf
│ ├── PwC Solvency II Capital Model Survey.pdf
│ ├── Ten Things You Should Know About the Dynamic Conditional Correlation Representation.pdf
│ ├── Time Series Analysis.djvu
│ └── _LCP Solvency II report 2019.pdf
├── README.md
└── Writeups
├── Final_Report.pdf
└── Working Reports
├── Practitioners challenge (PC) : risk factors.pdf
├── Risk Factors final choice (AutoRecovered).docx
└── Risk Factors final.pdf