Glacier flow is typically modeled using a power law rheology known as the Glen–Nye flow law, with the power n assumed to be 3. However, recent research and past observations suggest that n=4 may better represent ice flow in some locations. We lack a quantitative understanding of how much this exponent affects ice loss projections, and its significance relative to other sources of uncertainty. Here, we test the effect of n=3 versus n=4 in a series of 300-year forward simulations of the Amundsen Sea Embayment, West Antarctica. We find that in periods of rapid grounding line retreat, uncertainty in n leads to a larger spread in ice loss projections than the spread due to uncertainty in climate forcing. The spatial sensitivity of n is generally proportional to the change in strain rates, so we expect regions undergoing more moderate dynamic change to be less sensitive to n.