The effect of shade from one PV module on another is incorporated into a model for the power generated by PV systems. The model is calibrated with data from the Tucson Electric Power solar test yard. Shade de-rating factors from the model are compared with data every minute of the day and every day of the year. The model is then used to predict final yields (kWh/kWDC) for different PV system deployments with various (non-tracking) module orientations and ground-cover ratios. Several heuristics are put forth to help understand how the observed non-linear response to shade can impact the yield from PV systems. In one example, we find that a PV system deployed in the Tucson Electric Power solar test yard could produce 22% more kWh for the month of December (and 3.8% more annually) if the modules were separated by twice as much distance. In another example, we predict that a system in Tucson with south-facing modules at 12-degrees from horizontal can generate 1.5 times as many kWh/yr per square-meter of land compared to a system with modules at 32-degrees (the latitude angle). These examples emphasize the non-linear impact of partial shade on PV system performance.