Land sharing and land sparing are contrasting proposals for minimising the impacts of agriculture on wild
species. Edge effects (biophysical gradients near habitat boundaries) might reduce population sizes on spared
land, particularly in highly-fragmented landscapes, so might change conclusions about whether land sparing
or land sharing is better for species' persistence. We assessed this possibility by modelling the population sizes
of 120 Ghanaian bird species in the presence of a range of hypothetical edge effects under land-sparing and
land-sharing strategies, and at different levels of habitat fragmentation and agricultural production. We found
that edge effects can reduce population densities on spared land, and in highly-fragmented landscapes can – at
modest levels of agricultural production combined with high edge penetration distances – cause the optimal
strategy to switch from land sparing to land sharing. Nevertheless, land sparing maximised population sizes
for more species in most cases tested. This conclusion was best supported for sensitive species with small global
geographical ranges, which are likely to include those of greatest future conservation concern. The size of patches
of spared land affected conservation outcomes: population sizes were maximised under a land-sparing strategy
that spared large blocks of natural habitat of ~1000 or, better, ~10,000 ha. To effect land sparing in practice would
require policies that promoted both increases in agricultural yield and the establishment or protection of natural
habitats on spared land. Because the optimum scale of patches of spared land for edge-sensitive species is generally
larger than the size of individual farms, policies that facilitate coordinated action by farmers or other land managers
might be required.