We analyse migration from farm households in Kenya, and argue that many migration decisions should be regarded, not as individual decisions but rather as household decisions, in which, within the constraints given, the household decides to allocate its labour force among activities in such a way as to maximize household utility. The resulting migration tends to be circular. We use a probit model which incorporates this feature, and also takes into account contacts and information, as well as indivisibilities. Our empirical analysis suggests that the pull of high urban wages is much more important for migration decisions than the push of land scarcity. A buoyant local non-agricultural economy does not seem to restrict migration. We have also found that networks of personal contacts are highly significant determinants of migration. One important policy implication is that rural development, desirable though it is on other grounds, is unlikely to have the effect of stemming the flow of migration.