This paper explores the two common concepts of the natural rate of unemployment: (i) the stable, long-run equilibrium rate of unemployment, and (ii) the equilibrium unemployment rate at which there is no tendency for this rate to change, given the exogenous variables. The first concept (common in the theoretical literature) is impractical for empirical assessment, since it requires reliable estimates of the long-run values of the exogenous variables (which are not available) and since it is inherently unable to provide an analysis of how the NRU changes through time. Consequently the second concept is used in empirical studies. The paper shows that this latter natural rate is not necessarily a reference point (a value toward which the equilibrium unemployment rate tends with the passage of time). Specifically, it is not a reference point in multi-equation labor market models containing lagged endogenous variables and exogenous variables with nonzero long-run growth rates. Since these features are exceedingly common, our analysis casts serious doubts on the usefulness of the natural rate hypothesis as a predictive tool.