Existing models for predicting long-term corrosion rates are generally power models with exposure time as the independent variable, with separate models for steel type and environment (i.e., urban, rural, marine). An international data base, with data from eight countries (Belgium, France, Germany, Great Britain, Japan, South Africa, Sweden, and the U.S.), was analyzed to examine the effects of exposure time, sample replication, sample size, steel type, the angle and direction of exposure, and the environment. Four steel types were included in the data: A588, A242, copper, and carbon. The results indicate that a global model would not provide accurate predictions, but it was possible to regionalize data for central Europe. Replication of samples was shown to be important for some types of steels in urban and marine environments, but not important for rural environments. The duration of exposure is very important, with some exposure durations being much more important than others. North-facing specimens were shown to corrode more than those facing south, and vertical specimens corroded more than those exposed at an angle of 30°. The amount of corrosion varied with steel type, but less with the country in which the data were collected. These results indicate that corrosion model accuracy can be improved by accounting for these other factors.