This analysis demonstrates that changing the assumptions used to develop emission inventories can confound attempts to show that a region is making progress toward attainment. Ambient carbon monoxide data and four different estimates of the regional emissions in the California South Coast Air Basin from 1990 through 1995 are used to estimate rollback relationships. Mixed models are specified to the ambient concentration data and estimated emissions, allowing the calculation of the expected increase in ambient concentration per ton of carbon monoxide emitted under each emissions inventory scenario. Statistically significant differences are found between rates of expected change in carbon monoxide concentration among the four scenarios tested. Existing guidance for conformity determinations is found to be insufficient.