In this paper, I conduct a structural change test that casts doubt on the validity of exogenous growth assumptions. Cross-sectional empirical support for non-stochastic convergence in the neoclassical growth model is the reason that the literature rejects endogenous growth. But, in a stochastic world, both neoclassical and endogenous growth models exhibit disequilibrium adjustment dynamics, thus convergence is not sufficient to reject endogenous growth. After testing for cointegration in regional per-capita incomes, I extract a single common trend to control for non-stationarity in regressions including both linear and stochastic trends. Structural change tests demonstrate that the data contain segmented linear trends, which is inconsistent with an exogenous growth assumption, but is consistent with endogenous growth.