A high quality basic education for every child has been enshrined as a development goal since even before there were development goals. Promoting the achievement of this goal around the world, particularly the “high quality” component of the goal, requires a serviceable general positive model of education policies—a coherent causal explanation of why governments actually do what they do. The difference between thinking small and thinking big in research the economics of education and in “policy advice” depends on whether researchers and policy advisers have a correct positive model of the process of the diffusion to scale of new knowledge, of potential innovations and of what factors influence policy adoption. “Normative as Positive” (NAP) is one possible positive model, which explains that the policies actually chosen were chosen because they maximize some aggregate social welfare function. But since NAP is false, “policy recommendations” based on NAP will be relevant to the actual process of policy making only by coincidence. This is equally true of policy recommendations based on more “rigorous” empirical methods like field and randomized experiments—these techniques are intrinsically no more nor less policy relevant than other research methods—they still must be linked with a plausible positive model of policy and policy change if they hope to have influence at scale. A false positive model such as NAP is not only useless as a guide to policy relevance, it is potentially worse than useless as it may point the research and “policy” work of economists and educationists in precisely the wrong direction—towards nation-states, technocrats and bureaucrats (or “policy makers”) as the locus for educational reform rather than students, parents, communities, and teachers.