Regulators of the railroad industry are tasked with protecting shippers from excessive rates for shipments in which the railroad is market dominant, defined as an absence of effective competition from intramodal and intermodal competition. This task requires accurate measures of shipment costs, markups, and how these markups relate to competing modes of transport. However, the current regulatory accounting approach of allocating costs and markups is heavily criticized. In contrast to the academic literature, which is aggregate in nature and estimates the average cost and markup over the network, I develop a method to measure costs and markups that retains their disaggregate properties. I adapt and apply a quadratic cost function that provides shipment costs and markups and use these results to explore market dominance, wherein the markup and the presence of competing modes of transport determine whether shippers may be eligible to contest the reasonableness of the rate. I find that a movement from monopoly to duopoly leads to an average 6.8% decline in rail markups. The results suggest rail markups are most constrained by rail competition within 10 miles of the origin-destination and that nearby ports decrease the impact of rail competition on rail markups. This approach can be operationalized by regulators and market participants to assess the reasonableness of a rate and to streamline and expedite market dominance inquiry.
In this paper, we use Bayesian techniques to develop nowcasts for the quantity of waterborne traffic in the United States in total and for the four primary commodities. These waterborne traffic levels are released with a considerable time lag, but yet are of current interest. Nowcasts (i.e. predictions of the waterborne traffic levels to be released based on other variables that are available) have been constructed using an array of different variables and techniques. However, the large number of potential predictor variables and changes in the distribution of traffic levels leads to both model and estimation uncertainty, which has likely hampered the accuracy of these existing nowcasts. We use Bayesian Model Averaging (BMA) to create nowcasts, which confronts model and estimation uncertainty directly via the averaging of models with different sets of predictors. We also use rolling window techniques to account for possible changes in the nowcasting relationship over time. Based on a variety of evaluation metrics, we find that BMA substantially improves nowcast accuracy.
In this paper, I adapt and apply a technique for estimating multiproduct cost functions in the railroad industry. Historically, regulators have relied on an accounting cost allocation procedure to determine whether railroads are exploiting their market power and charging excessive rates. But, the current regulatory approach has been heavily criticized. In this application, I develop and estimate a linear quadratic cost function using first and second moments of shipment and railway characteristics. This approach provides a solution to handle the large number of product-origin-destination combinations. Implementing the model in this way allows shipment specific costs to be estimated while also incorporating the shared network technology inherent in railway networks. The result can be used in conjunction with rates to identify excessively high rail rates, it can also be used to estimate the costs attached to a specific rail movement which can be important for shippers in negotiating rate under contracts; shippers can use it to evaluate eligibility for rate relief. Railroads can operationalize this method to set more competitive rates and avoid the dispute resolution process.