Statistical modeling of vehicle emissions from inspection/maintenance testing data: An exploratory analysis

Many metropolitan areas in the United States use vehicle inspection and maintenance (I/M) programs as a means of identifying high-polluting vehicles. While the effectiveness of such programs is debatable, the cost is undeniable, with millions of dollars spent in testing and millions more lost in the time motorists expend to participate in such programs. At the core of these costs is the blanket approach of requiring all vehicles to be tested. This paper sets the groundwork for a procedure that can be used to selectively target those vehicles most likely to be pollution violators. Using I/M data collected in the Seattle area in 1994, carbon monoxide, carbon dioxide and hydrocarbon emissions were modeled simultaneously using three-stage least squares. Our results show that vehicle age, vehicle manufacturer, number of engine cylinders, odometer reading, and whether or not oxygenated fuels were in use all play a significant role in determining I/M emission test results and these statistical findings can be used to form the basis for the selective sampling of vehicles for I/M testing.