Publications:
Working Papers:
Productivity and its growth are key values in determining economies of scale, long-term growth, and long-term viability of firms and industries. Partial deregulation of railroads was led by concerns that existing regulation and changes to the industry led to stagnation in productivity. Policy changes made it easy for firms to increase productivity through mergers and abandoning unprofitable routes as well as technological innovation through the 1980s and early 1990s. However, as the industry has become increasingly consolidated and as more lines have been abandoned, firms may need to rely on technological change to increase productivity. I develop a model that separates changes in productivity due to innovation and those caused by non-innovative factors. By allowing productivity and technology to evolve flexibly over time, I examine changes in railroad productivity and identify its driving component. I find that BNSF, CN, and KCS have all experienced growth in productivity since 1999. Improvements in technology were the driving factor in BNSF's growth, while CN and KCS saw significant growth due to factors other than innovation. Finally, I develop a metric that determines whether firms substitute inputs towards factors that innovation makes more productive. I estimate the probability that each firm takes that action to be around 50% with no discernible pattern over time, indicating firms don't anticipate technological change or don't adjust input allocation to take advantage of innovations.
In this paper, we develop and estimate a model that provides both markups and scale elasticities that vary across railroads and through time for the traffic on their networks. Our model is based on a framework provided by Hall (1988) and Klette (1999) wherein markups and scale elasticities are estimated from production relations.  In our model, we aggregate the shipments over each firm's network, which provides a mapping from inputs and network and shipment characteristics to aggregate outputs over the network.  Markups and scale elasticities are taken to follow a multivariate distribution.  This allows for differences in markups and scale across firms and through time, but also for covariances across firms in markups and scale. We estimate the model with Bayesian methods to find markups that are generally well in excess of marginal costs and scale elasticities that generally point to increasing or constant returns in the industry.
There is a wealth of literature that points to inefficiencies in production. Inefficiencies can arise in the production of outputs from overutilization of inputs in the production process (technical inefficiency) or from errors in optimization that misalign factor prices and optimal input decisions (allocative inefficiency). In examinations of inefficiency, many studies use an inflexible production technology, typically the Cobb-Douglas form, which fails to account for differences in the technology of firms, or fails to control for the effects of competition in limiting inefficiency. In this study, I develop a model that allows for substitutes and complements in production and flexibly accounts for patterns in productivity. I use the model to derive firm cost functions and estimate technical and allocative inefficiencies. Finally, I allow allocative errors to be correlated with the level of competition to examine how the incentive to precisely allocate inputs and minimize costs are affected by competitive pressures.