In this paper we analyze the effects of neighbor shadowing of tracking solar photovoltaic arrays when they are set out in solar farms for large scale generation. Closer tracker spacing yields more power per unit area of land, but less power per tracking unit because of shadowing. A model has been developed to quantify and compare efficiencies for different tracker aspect ratios and field layouts, on an hourly, daily and annual basis. The model accounts for atmospheric absorption as well as neighbor shadowing at low solar altitude angles. We have focused on the case of CPV arrays which are oriented normal to the rays from the sun. The field layout is best characterized by the ratio of total array area to land area (the ground cover ratio or GCR). We explore as a function of GCR both the fraction of all the direct sunlight energy that is intercepted by the arrays (the irradiance collection efficiency) and the energy lost by each array because of shadowing. Examples are worked out for rectangular arrays on dual axis trackers at 33° latitude. We find that for a ground cover ratio of 30% the annual irradiance collection efficiency is 50%, almost independent of the layout pattern or the array aspect ratio. For a ground cover ratio of 40%, the irradiance collection efficiency rises to 65%. The corresponding shadowing losses do depend on aspect ratio, thus for 30% GCR the annual average of shadowing loss is 7.2% for 3:1 aspect ratio, rising to 7.8% for 2:1 aspect ratio. High GCR is not realizable for higher aspect ratios, which lead to large swing radius, but for 2:1 aspect ratio 40% GCR results in shadowing loss of 11.5%. One conclusion is that a solar farm with arrays of 2:1 aspect ratio set out with 40% GCR is good compromise when land is scarce: 64% of all the direct sunlight energy incident on the land is harvested by the arrays, with only 11.5% shadowing loss. We have compared these efficiencies with those for trough CSP systems, which also harvest direct sunlight but with reflectors turning about a single, horizontal N-S axis. For given GCR, the shadowing loss is slightly less (0.5%) than for the above dual-axis arrays, however the irradiance collection efficiency is worse in winter, leading to a lower annual average for a given GCR. For example, at 40% GCR, a single-axis system realizes a 56% irradiance collection efficiency compared to 64% for the dual axis systems.