This paper gauges efficiency in container ports. Using non-parametric methods, we estimate efficiency frontiers based on information from 86 ports across the world. Three attractive features of the method are: 1) it is based on an aggregated measure of efficiency despite the existence of multiple inputs; 2) it does not assume particular input-output functional relationships; and 3) it does not rely on a priori peer selection to construct the benchmark. Results show that the most inefficient ports use inputs in excess of 20 to 40 percent. Since infrastructure costs represent about 40 percent of total maritime transport costs, these could be reduced by 12 percent by moving from the inefficient extreme of the distribution to the efficient one.

1