In the 1990s, Brazil and Mexico were pioneers in the implementation of conditional cash transfer programs (CCTs), which since have benefitted an estimated one billion poor families around the world. However, the initial evaluation strategies pursued by each state were different: Mexican officials partnered with US economists to implement an RCT evaluation, while Brazilians used a combination of statistical simulations and qualitative studies and aimed to secure the generation of policy knowledge to domestic experts. Based on eighteen months of participant observation in Mexico City and Brasília, 100 interviews with political and academic elites, content analysis of 400 policy documents, and historical-process tracing methods, this dissertation explains why these two similar countries, implementing the same policy, took different routes to assess the merits of CCTs, and what unintended consequences followed from these choices. I demonstrate that a key factor to achieve the legitimacy and political viability of CCTs is the knowledge regimes that states create to implement and evaluate these programs. The dissertation shows that while knowledge regimes tend to be understood as technical or apolitical machineries, they are inherently shaped by the politics of legitimation of CCTs and they produce unanticipated consequences for the ways that states combat poverty in the long-run. Only by taking into consideration the role that knowledge production plays in securing the political viability of CCTs, I argue, we can assess the politics and consequences of these programs, and how they relate to poor families on the ground.