The welfare impacts of targeted transfers to poor households

South African researchers and policymakers have for some time now been exploring the idea of basic income grants and extended welfare transfer schemes as a means to reduce poverty and inequality in the country. Despite the fact that poverty in South Africa is widespread there is a general consensus among policymakers about the preference of targeted welfare grants over non-targeted grants due to the huge budgetary implications of the latter. Targeting, however, brings with it increased administrative complexities both with respect to means testing that is required and the challenge of successfully identifying appropriate recipients. These administrative complexities introduce a cost dimension that is as yet largely unexplored. This paper reports the results from a series of simulations that evaluate the general equilibrium and welfare implications of different schemes for targeted transfers. The poverty and inequality implications are evaluated using microsimulation techniques that link the representative households in the CGE model to the household surveys. The general equilibrium model is calibrated with an aggregation of the PROVIDE Social Accounting Matrix (SAM) for South Africa for the year 2000. Distinctive features of the SAM include detailed factor and household accounts: overall the SAM has 32 commodities, 39 activities, 56 factors, including GOS (capital), nine land factors and 46 labour factors, and the 162 household accounts of the PROVIDE SAM. The labour accounts distinguish between types of labour on the basis of occupational category, race and province of residence, while the household accounts distinguish between households on the basis of race, gender and educational achievements of the ‘head’ of household; province of residence further disaggregated according to whether the district of residence was in a former homeland; and, for larger groups, the level of income. A distinctive feature of the SAM is the fact that there is a complete mapping of the individual households in the Income and Expenditure and Labour Force Surveys to the representative households in the SAM. It is this mapping that allows the development of detailed associated microsimulation techniques. The simulations assume that the total value of transfers, excluding administrative costs, is constrained at R15 billion (approximately US$2.5 billion or 2.4% of total government expenditure). The simulations explore the effectiveness of different targeting regimes, e.g., broad targeting (associated with a low per capita transfer value) versus narrow targeting (associated with a high transfer value), under the assumption that the targeting regimes are determined by reference to the characteristics of the representative household groups. For all simulations, the impact of adding an administrative cost component to the expenditure side is also explored. Various tax replacement policies are explored under alternative financing options; from a welfare point of view these are important, especially when tax rates paid by households that are close to the poverty line change. The effectiveness of different targeting schemes in reducing poverty and inequality is explored in depth. This is achieved by extracting results on per capita income changes from the general equilibrium model and feeding these into a micro-level module that calculates poverty and inequality at the individual level. The results indicate that the poverty impact of the R15 billion transfer is typically small: the poverty headcount falls from about 49% in the base to approximately 46% in the simulations. However for some household groups, which are in close proximity to the selected poverty line, poverty actually increases due to the increased tax burden. This highlights the importance of ensuring an equitable distribution of the increased tax burden. Inequality also declines marginally in all the simulations considered, mainly because poor households are targeted while non-poor households typically carry a larger share of the increased tax burden. In as far as the effectiveness of broad versus narrow targeting is concerned the results suggest that narrower targeting generally implies greater reductions in poverty and inequality, although it depends crucially on how far the recipients are located from the poverty line and the funding regime. The paper concludes with a discussion of alternative targeting regimes.