Emissions of volatile organic compounds (VOCs) and of CO are often underestimated, and correcting this underestimate is important for modeling ozone formation and sensitivity. The California Institute of Technology three-dimensional photochemical model is used to test the official emissions inventory for the Mexico City metropolitan area through a direct comparison of measured and modeled total nonmethane hydrocarbons (NMHCs) and CO. The model is applied to six 2-day periods during the IMADA measurement campaign of March 1997. When using emissions based on the official emissions inventory, the model significantly underestimates measurements of total NMHCs and of CO. A best fit to the measurements (with lowest bias) is found when increasing emissions of CO and VOCs by factors of 2 and 3, respectively. Adjusting total emissions accordingly, the model produces good estimates of ozone and NOx, with average normalized biases over 6 days of 3% and 32%, respectively, and with better agreement during daytime hours. This agreement for ozone and NOx supports the suggestion that VOC and CO emissions are underestimated. Other uncertainties are analyzed, finding that while some uncertainties are important, none is individually significant enough to account for the discrepancy. Correcting the total emissions, the overall model performance is found to be adequate, particularly on 3 days, for the model to be used for analysis of control strategies. Using the results of a chemical mass balance identification of NMHC sources highlights sources that are more likely underestimated, but confidence in the appropriate correction to emissions from different sources is low and should be investigated further.