This research extends earlier work conducted by AFSC economists to assess the effects that accounting for two types of variations in inputs (shocks) of regional economic impact models have on the statistical significance of the model estimates. In this research, confidence intervals for regional economic impacts resulting from six changes in saltwater sportfishing harvest limits are calculated using a stated preference model of sportfishing participation and a social accounting matrix (SAM) for southern Alaska. Two important types of input variation are considered in calculating the economic impacts: sample variation in sportfishing-related expenditures and stochastic variation from parameters in the recreation participation model. Results indicate that the estimated confidence interval for the economic impacts (in terms of the change in total regional output) that accounts for both types of variations is significantly wider than the one estimated when considering only the sample variation. In fact, for the six policy scenarios considered, only the change in total regional output associated with a two-fish reduction in the Pacific halibut bag limit is statistically greater than zero when both types of variation are considered in the calculation of economic impacts. For the other five policy scenarios, the 95% confidence intervals contain zero, suggesting reductions in the bag limits are not statistically different from zero. Moreover, using the methods of convolutions to assess differences in estimated impacts between scenarios, we show there are only statistical differences between estimated economic impacts when sampling variation alone is accounted for, but none when stochastic variation is considered. This suggests that in some cases, decision makers may need to look beyond a simple comparison of the point estimates of economic impacts of alternatives as a basis for choosing a preferred alternative due to a lack of statistical differences in the results from regional economic impact models.