On modeling pollution-generating technologies

Distinguishing between intended ("good") production and unintended or residual ("bad") generation, we introduce the concept of by-production. In by-production technologies, pollution is an output that satises a "costly disposability" assumption and violates standard free disposability with respect to pollution-causing inputs. Our approach therefore differs substantially from standard approaches to modeling pollution-generating technologies. We show how by-production can be modeled using data envelopment analysis (DEA) methods. With an electric power plant database, we illustrate shortcomings under by-production of two popular eciency indexes: the hyperbolic index and the directional distance function. We propose and implement an alternative eciency index with superior properties.