In this paper, I digitize economic census data to study unconditional convergence in manufacturing labor productivity across Mexican states from 1988 to 2018. I document its existence in three-digit industries at a rate of convergence of 1.22% per year. However, this result does not hold at the aggregate level: I find no unconditional convergence in manufacturing wide labor productivity across states. Shift-sharing analysis reveals that the primary reason is the lack of labor reallocation towards more productive industries and the underperformance of some of the largest ones. Unconditional convergence at all levels only occurred during 1988?1998. Afterward, the convergence process broke down and was only observed at disaggregated levels. I provide evidence that one possible cause of this breakdown is the so-called "China shock". Additionally, I show that the convergence process, when it happened, tended to exhibit a catching-down feature, where past leaders have seen their labor productivity decline.