In a clinical trial, the treatment effect $\Delta$ is modeled using a maximum likelihood estimator $\hat{\Delta}$. A researcher notes that the Wald statistic $W = \frac{\hat{\Delta}^2}{\widehat{Var}(\hat{\Delta})}$ behaves unexpectedly when $\hat{\Delta}$ is near the boundary of the parameter space. Which phenomenon describes this behavior where the Wald statistic decreases as the estimate moves further from the null, potentially leading to a failure to reject $H_0$?