In a survey using a $p$-th order inclusion probability design, let $\pi_i$ denote the first-order inclusion probability and $\pi_{ij}$ the joint inclusion probability. For the Sen-Yates-Grundy (SYG) variance estimator of the total $\hat{Y}_{HT} = \sum_{i \in s} \frac{y_i}{\pi_i}$, which of the following conditions is required for the estimator to be non-negative and unbiased for any fixed sample size $n$?