Linkages between sector-specific policy and sector employment are explored using a nonstationary Markov chain analysis. When parameterization of transition probabilities between employment sectors includes policy variables, hypothesis tests can determine policy impact relative to other variables. This astructural approach eliminates bias inherent in structural models. Application of this technique to Oregon's forestry sector and national forest policy demonstrates that macroeconomic forces have statistically important effects on employment while national forest policy, measured as timber sold or timber cut, does not. This result raises questions about forest policy impact analysis and assumptions inherent in national forest policy implementation. Copyright 1997, Oxford University Press.