Many countries have undergone a broad retreat from the use of indirect (supply) subsidies to meet low-income housing-affordability problems, shifting to direct subsidies often linked to means-tested income-maintenance systems. Although the reasons for this change of direction are well documented, the efficacy of direct housing subsidies in terms of tackling affordability remains in question. The authors examine in detail one such system, Australia's Rent Assistance (RA) programme, making use of a microsimulation model of the Australian housing market linked to a model of the social security system. It is found that there is considerable targeting error because many low-income renters are ineligible for direct subsidies. It is also found that RA is relatively ineffective in overcoming affordability problems in high housing cost areas of Australia. Although RA does not of itself contribute much to poverty-trap problems, it may deter unemployed households from moving to areas where job vacancies exist. The authors conclude that future comparative work could usefully analyse the distributional and behavioural impacts of different forms of housing allowances. Moreover, alternative policies could be recalibrated within the microsimulation model in order to examine the first-round impacts of policy design change.