Estimating lock congestion

We statistically estimate the relationship between lock usage (the number of barges locked) and lock congestion (the average delay and process time incurred) using a fixed‐effects regression model that accounts for both lock characteristics and lock heterogeneity. In addition, we exploit the geographical variation in demand patterns for individual locks along the Upper Mississippi River system of locks and dams to avoid simultaneity bias that arises from the use of observed data. Using a panel dataset spanning the years 1993‐2010, we find statistically significant evidence of the existence of a quadratic lock congestion function for 600 ft lock technology. While 1,200 ft lock technology on the other hand seemingly operates under free‐flow conditions, the presence of an auxiliary chamber significantly mitigates congestion, with the effect being more pronounced for higher levels of traffic. Unscheduled lock outages attenuate lock congestion for both types of lock technology, while lock age (following a U‐shaped relationship) affects congestion costs for 600 ft locks only. In addition, we find a tempering effect of scheduled outages in the case of 600 ft locks, and non‐commercial vessels in the case of 1,200 ft locks.