In a survey using a stratified design, let $N_h$ be the stratum size and $S_h^2$ the stratum variance. If we use power allocation, $n_h = n \cdot \frac{N_h S_h^p}{\sum N_i S_i^p}$, where $p=1$ corresponds to Neyman allocation. If $p=0$, the allocation becomes proportional. For a specific population where $N_1=1000, S_1=2$ and $N_2=1000, S_2=8$, what happens to the sample size $n_1$ as the exponent $p$ increases from $0$ to $2$?