GARCH Volatility Models Quiz

Answers and explanations appear as you take the quiz.

1. Which terms drive the next variance in a GARCH(1,1) model?
   • A constant, past squared innovation and past conditional variance
   • Only the current return sign
   • The next observed return and its future error
   • The number of features and clusters
2. What does a large recent squared shock tend to do to conditional variance?
   • Set the expected return direction to positive
   • Raise the next variance forecast through the ARCH term
   • Force beta to zero
   • Make realized variance known exactly
3. Can a GARCH variance forecast determine whether the next return is positive or negative?
   • Yes, beta is the return sign
   • Yes, higher variance means a positive return
   • No, it models conditional variance rather than return direction
   • Only if the variance forecast is zero
4. What do standardized residuals help assess?
   • The exact future loss amount
   • Whether feature names are ordered
   • The number of seasonal periods
   • Whether innovations scaled by fitted volatility retain structure
5. Why inspect squared standardized residuals?
   • To estimate the mean return direction
   • To look for remaining volatility clustering
   • To calculate a classification threshold
   • To test whether all shocks are positive
6. Under standard GARCH(1,1) assumptions, what does alpha+beta near one suggest?
   • A known positive return
   • No effect of past variance
   • Highly persistent volatility shocks
   • A zero conditional variance
7. Why might a heavy-tailed innovation distribution be considered?
   • It makes the mean equation unnecessary
   • It guarantees every tail event is predicted
   • It sets conditional variance to zero
   • Financial returns may produce more extreme observations than a normal model captures
8. Which measurement caveat applies when validating a variance forecast against realized variance?
   • The proxy itself is noisy and imperfectly measured
   • A proxy is always the true conditional variance
   • Only the return mean can be evaluated
   • The forecast horizon no longer matters