Improving community resiliency to wildfire is a challenging problem in the face of ongoing development in fire-prone regions. Evacuation and shelter-in-place are the primary options for reducing wildfire casualties, but it can be difficult to determine which option offers the most protection in urgent scenarios. Although guidelines and policies have been proposed to inform this decision, a formal approach to evaluating protective options would help advance protective-action theory. We present an optimization model based on the premise that protecting a community can be viewed as assigning threatened households to one of three actions: evacuation, shelter-in-refuge, or shelter-in-home. While evacuation generally offers the highest level of life protection, it can place residents at greater risk when little time is available. This leads to complex trade-offs involving expected fire intensity, available time, and the quality and accessibility of in-place shelter. An application of the model is presented to illustrate a range of issues that can arise across scenarios.