The traditional concerns about farm animal welfare have centred around the impact of intensive environments and management practices on the animal. This emphasis on the physical environment is changing, however, with greater consideration being given to animal factors and in particular to the selective breeding of farm animals. In this paper we use examples from our own research on dairy cattle and sheep breeding that have made positive and practical contributions towards reducing welfare problems by creating more balanced breeding programmes. In both examples, inclusion of health and fitness traits into breeding indexes can be shown to be more profitable than selecting on production traits alone. In addition, we propose that in principle animal breeding combined with economics research can make a more general contribution towards resolving animal welfare issues, by providing a framework for the quantitative evaluation of the costs and benefits of an animal production system. The advantage of the approaches currently used in multi-trait selection is that they transform all traits (production-based or welfare-based) to a common currency allowing direct comparisons of costs and benefits. Currently the weights applied to traits reflect their economic value to the producer. This approach is likely to underestimate non-economic welfare aspects such as the pain or discomfort associated with lameness, and new approaches are needed to more fully account for these non-economic welfare costs. We therefore propose that consideration is given to the use of approaches such as contingent valuation, which has been widely used in economics to derive values for non-economic activities. The question of who would pay for the addition of these welfare costs to a breeding index remains open, but it would seem most reasonable to treat these as a public good and pay for them as such through appropriate mechanisms.