We re-examine the concept of beta-convergence in living standards across countries during the period 1980–2012. In this study, well-being is assessed using the Human Development Index (HDI) which considers income aspects as well as social indicators, thus reflecting the multidimensional nature of this process. The existence of sigma convergence is evidenced in this study and hence beta convergence, as a necessary condition, is also pointed out. However, the linearity of this process has been questioned. Therefore, we apply a semiparametric specification of this process to the HDI and each of its intermediate indices. These models allows for nonlinearities in the estimation of the convergence speed. Our results reveal that absolute convergence in human well-being is satisfactorily represented by the conventional linear specification. However the income and education indices show nonlinear patterns. We also include structural variables to capture differences in the steady-state (conditional convergence). Under this model, convergence speed of all indicators is higher and the convergence process seems to be linear only for the health index.