We develop the empirical implications of Page's (2006) definition of path dependence as a process where the sequence of historical events affects the final outcome. A critical necessary condition for path dependence in common dynamic models is a time-varying autoregressive parameter whose value becomes 1 at some point. Failure to meet this condition results in a path independent process whose equilibrium outcome is only a function of the current exogenous conditions. This condition is illustrated with a discrete Markov Chain model and with a computational model with continuous variables, which are illustrated with models of partisan change. A Monte Carlo simulation shows how non-linear least-squares estimation can recover the parameters that distinguish path dependence from path independence. This integration of modeling and an estimation strategy is illustrated with data on civil rights attitudes and macropartisanship. The results have implications for discussions of path dependence in a wide range of social science fields.