%0 Manuscript %D Working Paper %T Converging to Convergence %A Michael Kremer %A Jack Willis %A Yang You %X Neoclassical theory predicts convergence towards steady-state income, determined by policies, institutions, and culture. Empirical tests of convergence in the 1990s found that conditioning on institutions mattered: \textit{unconditionally}, the norm was divergence, if anything. We revisit these tests with 20 years of additional data. While the literature on institutions emphasizes their historical origins and persistence, we find substantial changes. First, there has been a trend towards unconditional convergence since the 1960s, leading to convergence since the early 2000s. Second, policies and institutions have converged substantially, towards development-favored institutions - those associated, across countries, with higher levels of income. Third, the institutional changes are larger, on average, than those predicted by the cross-sectional income-institution slope; while the slope itself has remained stable. Fourth, the growth-institution slope - the coefficients of growth regressions - has decreased substantially, resulting in a shrinking in the gap between conditional and unconditional convergence. We discuss the implications of these new patterns for models of growth.    %B Accepted NBER Macro Annual 2021 %G eng