This paper examines the two approaches that are presently available in the data envelopment analysis (DEA) literature for use in identifying and analyzing congestion. The first approach, due to Färe et al. (FGL) has been extensively employed to identify congestion in areas such as the provision of health services. The latter approach, due to Cooper et al. was subsequently extended by Brockett et al. to treat tradeoff possibilities between employment and output in Chinese production when congestion is present. In contrast with Cooper et al., examples are provided in this paper which show that the Färe et al. approach can fail in both modes: It can show congestion to be present when this is not consistent with the observed behavior. It can also fail to exhibit congestion when the data show it to be present. Suggestions are offered both for (i) taking precautions before using the FGL approach and for (ii) overcoming these shortcomings.
