Recent events have highlighted the importance of good models of crowds, however many existing crowd models are either computationally inefficient, or are missing a crucial human behaviour in crowds: local pushing. After discussing some essential aspects of force in crowds, and considering some existing models, we propose an efficient agent-based model of crowd evacuation that incorporates pushing forces and injuries. Basing our model on existing work, we extend this model to investigate force effects at different crowd densities. Analysis of our model shows significant effects of force on the crowd, as well as significant effects of crowd density when measuring the number of agents still trapped inside the space after a fixed time.