Answer: A nonconstant variance

A covariance stationary time series must have: 1. Constant mean (stationary mean) 2. Constant variance (homoscedasticity) 3. Constant autocovariance structure (autocovariance depends only on lag, not time) Option D (a nonconstant variance) is the exception because it violates the requirement for constant variance. A nonconstant variance indicates heteroscedasticity, which means the time series is not covariance stationary. The other options are all characteristics of covariance stationarity: - A: Stability of autocorrelation (autocorrelation depends only on lag) - B: Stability of the mean (constant mean over time) - C: Stability of the covariance structure (autocovariance depends only on lag)

Author: Nikitesh Somanthe