Inclusive fitness models predict many commonly observed behaviours: among humans, studies of within-household violence3 , the allocation of food and child care find that people favour those to whom they are more closely related. In some cases however, kin-altruism effects appear to be modest. Do individuals favour kin to the extent that kin-altruism models predict? Data on remittances sent by South African migrant workers to their rural households of origin allow an explicit test, to our knowledge the first of its kind for humans. Using estimates of the fitness benefits and costs associated with the remittance, the genetic relatedness of the migrant to the beneficiaries of the transfer, and their age- and sex-specific reproductive values, we estimate the level of remittance that maximizes the migrant worker’s inclusive fitness. This is a much better predictor of observed remittances than is average relatedness, even when we take account (by means of a multiple regression) of covarying influences on the level of remittance. But the effect is modest: less than a third of the observed level of remittances can be explained by our kin-altruism model.