ABSTRACTThis paper considers the historical record and time series properties of United States' industrial production for the period 1860 to 1991, utilizing unit root tests and measures of persistence. The results identify a segmented trend model which is used to assess the time‐series simulation performance of four well‐known models of economic growth: Solow (1957); Mankiw, Romer and Weil (MRW, 1992); Barro and Sala i Martin (BSM, 1992); and Rebelo (1991). Both the MRW and BSM models dominate the Solow model in accounting for twentieth century industrial growth, highlighting the importance of human capital, and the paper suggests a new measure related to higher education. However, the Rebelo model explains the post‐1973 slowdown more successfully than either of the ‘augmented‐Solow’ approaches. The paper concludes with a discussion of the impact of shocks on US indust