Explanation

For a correctly specified autoregressive model:

• Residuals should be constant and finite: This is consistent with the assumption of homoskedasticity and stationarity
• Residual autocorrelations should NOT differ significantly from zero: If residual autocorrelations are significantly different from zero, it indicates that the model has not captured all the systematic patterns in the data, suggesting the model is misspecified

Therefore, option A (residual values are constant and finite) is most consistent with a correctly specified autoregressive model. Option B would indicate model misspecification.