The Arizona I/M program provides one of the first opportunities to examine the costs and effectiveness of vehicle emission repair. This paper examines various aspects of emission reductions, fuel economy improvements, and costs of repair, drawing data from over 80,000 vehicles failing the I/M test in Arizona between 1995 and the first half of 1996. We summarize the wealth of repair data from the Arizona program and highlight its limitations. Because missing or incomplete cost information has been a serious shortcoming for evaluation of I/M programs, we develop a method for estimating the costs of repair when those costs are not reported. We find surprising evidence that almost one quarter of all vehicles that take the I/M test are never observed to pass the test. Using a statistical analysis, we provide some information about the differences between the vehicles that pass and those that do not. Older, more polluting vehicles are much more likely to never pass the I/M test, and their expected costs of repair are much higher than those of new cars. The paper summarizes the evidence on cost and emission reduction in the Arizona program, comparing costs and emission reductions for both cars and trucks. Finally, we examine the potential for more cost-effective repair, first through an analysis of tightening I/M cutpoints, and then by calculating the cost savings of achieving different emission reduction goals when the most cost effective repairs are made first.