Risk and insurance: risk identification, measurement and management
This domain tests risk taxonomy, qualitative risk process and quantitative risk measurement, including how insurers aggregate frequency-severity risk and tail risk.
Identification and process
Pure risk has downside only; speculative risk has upside and downside. A risk register records cause, event, consequence, owner, controls and residual rating. ERM frameworks such as ISO 31000 and COSO emphasise governance, appetite, tolerance and monitoring. Controls are avoid, reduce, transfer and retain. Underwriting, claims management and reinsurance are risk-transfer and risk-reduction tools.
Measurement
Frequency-severity models separate how often and how much. Expected loss = expected frequency x expected severity. VaR is a quantile of the loss distribution; TVaR is the average loss beyond VaR. VaR is not subadditive, so standalone VaRs cannot simply be added. Dependence, fat tails and parameter uncertainty matter. Stress and scenario tests complement statistical models. Monte Carlo simulation is used where closed forms are unavailable.
Worked example
A portfolio has three outcomes: 90% no loss, 8% loss INR 10 lakh, 2% loss INR 50 lakh. Expected loss = 0.08 x 10 + 0.02 x 50 = INR 1.8 lakh. The 95% VaR is INR 10 lakh because cumulative probability reaches 98% there. TVaR at 95% averages the worst 5%: (0.03 x 10 + 0.02 x 50) / 0.05 = INR 26 lakh. TVaR captures the severe tail that VaR ignores.
Traps
Adding standalone VaRs ignores diversification but can also understate dependence. Confusing volatility with ruin is common. Do not treat a risk register as a measurement model. Ignoring parameter and model risk overstates precision.