Suppose we observe $X \sim \text{Poisson}(\lambda)$. If we use a hierarchical Bayesian model with the prior $\lambda \sim \text{Gamma}(\alpha, \beta)$, how does the posterior variance of $\lambda$ compare to the prior variance?