Skip to content
Licensed Unlicensed Requires Authentication Published by De Gruyter October 7, 2019

The Static-Dynamic Efficiency Trade-off in the US Rail Freight Industry: Assessment of an Open Access Policy

  • Daniel Coublucq , Marc Ivaldi EMAIL logo and Gerard McCullough

Abstract

Considering the US railroad industry, which is characterized by seven integrated firms that provide freight services on tracks they own and maintain, this paper provides a structural model that allows to evaluate the potential effects of opening the rail network to new firms on prices and investment incentives. In particular, we propose a framework for analyzing the tension between static efficiency (pricing behavior) and dynamic efficiency (investment behavior). The investment behavior is rendered endogenous by means of a dynamic model where the current investment depends on the expected future profits. We then use a forward simulation procedure to analyze the effect of an open-access market structure where a new firm uses the network of one of the biggest railroad firm. Under a simple access charge equaled to the marginal cost of access, investment in network infrastructure decreases by 10% per year, leading to a significant decrease in network quality over time. Under this setting, despite the increase of price competition, the decrease in network quality leads to a fall in consumer welfare. Other types of (more evolved) access charges might even allow to relax the tension between static efficiency and dynamic efficiency, allowing more price competition while preserving investment incentives. This topic deserves further research and is beyond the scope of this paper.

JEL Classification: C54; L10; L51; L92

Acknowledgement

The authors are thankful to the participants of the Applied Microeconomics Workshop at Toulouse (February 2012), the Second Workshop on Transport Economics-Competition and Regulation in Railways (Fedea-IEB, Madrid, Spain, March 2012), the Advanced Workshop in Regulation and Competition (Rutgers University, May 2012), and the Kuhmo Nectar Conference on Transportation Economics (Berlin, Germany, June 2012). Daniel Coublucq is in partnership with Compass Lexecon. The views expressed in this article are personal to the author, and do not necessarily represent those of Compass Lexecon.

Appendix 1: Concentration in the US Rail Freight Industry

This appendix presents the concentration over time in the US rail freight industry. Figures A1 and A2 list all the takeovers that happened in the railroad industry. In this paper, we do not consider a takeover as an investment. This is beyond the scope of this paper and can be related to the literature on endogenous mergers. (See Gowrisankaran 1999.) We focus on the issue of panel attrition due to concentration in the US rail freight industry.

Figure A1: Railroad Firms in the Eastern Area.
Figure A1:

Railroad Firms in the Eastern Area.

Figure A2: Railroad Firms in the Western Area.
Figure A2:

Railroad Firms in the Western Area.

Table A1:

Railroad Firms in the Eastern Area.

RailroadYears in dataAbbrevation (used in Figure A1)
Baltimore & Ohio (BO)1978–1985BO (into CSX in 1985)
Chesapeake & Ohio (CO)1978–1985CO (into CSX in 1985)
Consolidated rail corp. (CR)1978–1998CR (split between CSX and NS in 1999)
CSX transportation (CSX)1986–2006CSX
Norfolk Southern (NS)1986–2006NS
Norfolk & Western (NW)1978–1985NW (into NS in 1985)
Seaboard system railroad (SBD)1978–1985SBD (into CSX in 1985)
Southern railway system (SOU)1978–1985SOU (into NS in 1985)
Western Maryland (WM)1978–1983WM (into BO in 1983)
Table A2:

Railroad Firms in the Western Area.

RailroadYears in dataAbbreviation (used in Figure A2)
Atchison, Topeka & Santa Fe (ATSF)1978–1995ATSF (into with BN in 1995)
Burlington Northern (BN); Burlington Northern Sante Fe (BNSF)1978–2006BN; BNSF
Canadian National Grand Trunk Corporation (CNGT)2002–2006CNGT (it incorporates all US activities of Canadian National Railroad, which included GTW activities)
Chicago & Northwestern (CNW)1978–1994CNW (into UP in 1994)
Colorado and Southern (CS)1978–1981CS (into BN in 1981)
Denver, Rio Grande & Western (DRGW)1978–1993DRGW (into SP in 1993)
Detroit, Toledo & Ironton (DTI)1978–1983DTI (into GTW in 1983)
Forth Worth and Denver (FWD)1978–1981FWD (into BN in 1981)
Grand Trunk & Western (GTW)1978–2001GTW
Illinois Central (Gulf) (IC)1978–1998IC (into GTW in 1998)
Kansas City Southern (KCS)1978–2006KCS
Milwaukee Road (MILW)1978–1984MILW (into SOO in 1984)
Missouri-Kansas-Texas (MKT)1978–1987MKT (into UP in 1987)
Missouri Pacific (MP)1978–1985MP (into UP in 1985)
Saint Louis and San Francisco (SLSF)1978–1979SLSF (into BN in 1979)
Saint Louis, Southwestern (SSW)1978–1989SSW (into SP in 1989)
SOO Line (SOO)1978–2006SOO
Southern Pacific (SP)1978–1996SP (into UP in 1996)
Union Pacific (UP); Union Pacific-Southern Pacific (UPSP)1978–2006UP; UPSP
Western Pacific (WP)1978–1985WP (into UP in 1985)

In the data, there are two problematic elements in the construction of merged firms, namely the merged firms CSX and NS in 1986. These two firms appear in 1986 and are the results of the mergers of several firms. The firms BO and CO were merged into the Chessie System, and that system was then merged into SBD in 1986. For NS, we assume that the merger parties have sold their assets to the firm with the highest market share before the merger.[30] Thus, we assume that the firm NW has sold its assets to SOU in 1986. This treatment of merger yields an unbalanced panel data with an attrition characteristic such that (see Wooldridge 2003: Chapter 17):

rj,t=1rj,τ=1,for all τt1.

This attrition characteristic of the data is an important technical issue for the demand estimation. (See Coublucq 2013.)

Appendix 2: Algorithm for Demand Estimation

This section is based on Coublucq (2013), where additional details are available.

First, we discuss the endogeneity of the variables included in the estimating Equation (16). The price, denoted pj,t, and the within market share, denoted lnsj,t|g, are endogenous. Thus, the variables Δpj,t and Δlnsj,t|g=1 are also endogenous. The discussion becomes more subtle for the variables Δkj,t and Δλj,t1. Using the structure of the model, we know that the variables kj,t and λj,t1 are weakly exogenous. Indeed, kj,t=ln(Kj,t), and the capital stock is constructed using the relation Kj,t=Kj,t1(1δ)+Ij,t1, where Ij,t−1 represents the investment in the network at date t − 1. From the dynamic model, we know that the investment is endogenous and it is a function of the previous state of the industry, wt1. This implies that the capital stock Kj,t, and thus the proxy for network quality kj,t, are a function of wt−1. By construction, the error term ej,t in Equation (14) is uncorrelated with the previous state of the industry wt1. Thus, the proxy for the network quality, kj,t, is weakly exogenous since it is uncorrelated with the contemporaneous and the future error terms, ej,s, st, and correlated with the past error term, ej,s, st − 1. This implies that in the estimating Equation (16), the variable Δkj,t=kj,tkj,t1 is endogenous since kj,t is correlated with Δej,t through ej,t−1. Nevertheless, we can instrument Δkj,t by using Kj,t−1 as instrument since the lag of the capital stock is a function of the state of the industry at date t − 2, wt2, and the error term Δej,t is uncorrelated with the state of the industry at date t − 2 (for the estimation, we have also added Kj,t−2 as an instrument). Lastly, we discuss the endogeneity of the first-difference of the Mills ratio, Δλj,t1=λj,t1(wt1)λj,t2(wt2). Like the stock of capital, the Mills ratio λj,t1(wt1) is also weakly exogenous since it is uncorrelated with ej,s, st and it is correlated with ej,s, st − 1. In the estimating Equation (16), Δλj,t1 is endogenous since λj,t−1 is correlated with ej,t−1 and thus with Δej,t. We instrument Δλt1=λj,t1λj,t2 by the second lag of the Mills ratio, λj,t−2.

To summarize, the choice of the instruments is guided by the structure of the model. Hence, during the estimation, accepting the over-identifying restriction may be interpreted as accepting the structure of the model as well.

We now provide the estimation algorithm to deal with the attrition issue due to concentration. This is important since attrition creates a bias in the price and the capital parameters. (See Coublucq 2013, for further details.)

In the estimating Equation (16), we have assumed that we know the previous state of the industry, wt2, since we use the condition E[Δej,t|zj,wt2,rj]=0. To make the estimation feasible, we need to use the following iterative algorithm:

  1. Start with an initial guess of the vector of demand parameters, denoted μ^=(θ^,α^,σ^).

  2. Using Equation (12), we compute an estimate of the unobserved state variable that represents the unobserved firm efficiency, ξ^j,t.

  3. We compute the probabilities of remaining in the industry as a function of the industry state, P^j,t(w^t), where w^t=(Jt;Kj,t,Kj,t;ξ^j,t,ξ^j,t), using a probit model, and Kj,t and 𝝃^j,t represent, respectively the sum of the observed and the unobserved state variable for the competitors.

  4. The threshold value Φ¯j,t1(w^t1)=F1(P^j,t1(w^t1)) is computed and we obtain the Mills ratio λ^j,t1(w^t1) as a correction term for attrition (see Equation (15)). We are also able to recover λ^j,t2(w^t2).

  5. We estimate Equation (16) by an instrumental variable regression using the instruments zj,t, zj,t1, Kj,t1, Kj,t2, and λ^j,t2.

  6. Using the new demand estimates μ^=(β^,θ^,α^,σ^), we repeat steps 2–5 until convergence of the demand estimates.

Appendix 3: Robustness Check on the Investment Policy Function

This part of the appendix provides the results of an open-access market structure when the algorithm allows for an increase in the capital stocks of the competitors (they were kept constant in the initial algorithm). We assume that the investment policy function is the same for each firm. Instead of solving one equation (see step 11 in the algorithm, Section 5), we solve for the parameter γ that minimizes the norm N(γ) = f(γ)′f(γ), where f(γ) denotes the set of first-order conditions for the investment of the seven active firms in 2006.

The results are very similar. The average price in the industry decreases by 6% and firm j carries less freight volume (see Figure A3). These two elements decrease the benefits from investing in the network. Indeed, the investment of firm j decreases by 10% per year (see Figures A4 and A5). Overall, the consumer welfare decreases (see Figure A6) and the difference between the two welfares is increasing over time to reach a gap of 10% after 30 years.

Figure A3: Market Share of Firm j and Open-Access.
Figure A3:

Market Share of Firm j and Open-Access.

Figure A4: Investment and Open-Access (in Thousands, $1982).
Figure A4:

Investment and Open-Access (in Thousands, $1982).

Figure A5: Capital stock and open-access (in Thousands, $1982).
Figure A5:

Capital stock and open-access (in Thousands, $1982).

Figure A6: Utility on Firm j Network and Open-Access (in Thousands, $1982).
Figure A6:

Utility on Firm j Network and Open-Access (in Thousands, $1982).

References

Association of American Railroads (1978–2006) Analysis of Class 1 Railroads. Annual Volumes.Search in Google Scholar

Berndt, E. R., A. F. Friedlaender and G. McCullough (1992) “Governance Structure, Managerial Characteristics, and Firm Performance in the Deregulated Rail Industry,” Brookings Papers on Economic Activity, Microeconomics, 23:95–186.10.2307/2534763Search in Google Scholar

Berndt, E. R., A. F. Friedlaender, J. S.-E. Chiang and C. A. Vellturo (1993a) “Cost Effects of Mergers and Deregulation in the U.S. Rail Industry,” The Journal of Productivity Analysis, 4:127–144.10.1007/BF01073470Search in Google Scholar

Berndt, E. R., A. F. Friedlaender, J.-E. W. Chiang, M. Showalter and C. A. Vellturo (1993b) “Rail Costs and Capital Adjustments in a Quasi-Regulated Environment,” Journal of Transport Economics and Policy, 27:131–152.Search in Google Scholar

Berry, S. (1994) “Estimating Discrete Choice Models of Product Differentiation,” RAND Journal of Economics, 25(2):242–262.10.2307/2555829Search in Google Scholar

Berry, S., J. Levinsohn and A. Pakes, (1995) “Automobile Prices in Market Equilibrium,” Econometrica, 63(4):841–890.10.2307/2171802Search in Google Scholar

Cardell, N. S. (1997) “Variance Components Structures for the Extreme-Value and Logistic Distributions with Application to Models of Heterogeneity,” Econometric Theory, 13:185–213.10.1017/S0266466600005727Search in Google Scholar

Christensen Associates (2008) “A Study of Competition in the U.S. Freight Railroad Industry,” Surface Transportation Board, November.Search in Google Scholar

Coublucq, D. (2013) “Demand Estimation with Selection Bias: A Dynamic Game Approach with an Application to the US Railroad Industry,” Working Paper No 94, Düsselddorf Institute for Competition Economics (DICE).Search in Google Scholar

Gallamore, R. E. and J. R. Meyer (2014) American Railroads: Decline and Renaissance in the Twentieth Century. Boston: Harvard University Press.10.4159/9780674369368Search in Google Scholar

Government Accountability Office (2006) “Freight Railroads: Industry Health has Improved, but Concerns about Competition and Capacity should be Addressed,” October 6. Available at https://www.gao.gov/products/GAO-07-94.Search in Google Scholar

Gowrisankaran, G. (1999) “A Dynamic Model of Endogenous Horizontal Mergers,” RAND Journal of Economics, 30(1):56–83.10.2307/2556046Search in Google Scholar

Hansen, L. P. and K. J. Singleton, (1982) “Generalized Instrumental Variables Estimation of Nonlinear Rational Expectations Models,” Econometrica, 50(5):1269–1286.10.2307/1911873Search in Google Scholar

Ivaldi, M. and G. J. McCullough (2001) “Density and Integration Effects on Class 1 U.S. Freight Railroads,” Journal of Regulatory Economics, 19:161–182.10.1023/A:1011149324027Search in Google Scholar

Ivaldi, M. and G. J. McCullough (2008) “Subadditivity Tests for Network Separation with an Application to U.S. Railroads,” Review of Network Economics, 7:159–171.10.2202/1446-9022.1143Search in Google Scholar

Judd, K. L. (1998) Numerical Methods in Economics. Cambridge: The MIT Press.Search in Google Scholar

Motta, M. (2004) Competition Policy: Theory and Practice. Cambridge: Cambridge University Press.10.1017/CBO9780511804038Search in Google Scholar

Nelson, R. A. (1975). Values of U.S. Class 1 Railroads. Association of American Railroads, Consultant Report.Search in Google Scholar

Organization for Economic Cooperation and Development (1997) OECD Policy Roundtables. Railways: Structure, Regulation and Competition Policy.Search in Google Scholar

Organization for Economic Cooperation and Development (2006) OECD Report on Experiences with Structural Separation. Available at https://www.oecd.org/daf/competition/sectors/45518043.pdf.Search in Google Scholar

Semykina, A. and J. M. Wooldridge (2005) “Estimating Panel Data Models in the Presence of Endogeneity and Selection: Theory and Application,” Journal of Econometrics, 157(2):375–380.10.1016/j.jeconom.2010.03.039Search in Google Scholar

Surface Transportation Board. (2009) Study of Railroad Rates: 1985–2007. Available at https://www.stb.gov/stb/industry/1985-2007RailroadRateStudy.pdf.Search in Google Scholar

U.S. Department of Commerce (1978–2006) Statistical Abstract of the U.S. Annual Volumes. Available at https://www.census.gov/library/publications/time-series/statistical_abstracts.html.Search in Google Scholar

Waters, W. G. (2007) “Evolution of Railroad Economics,” Research in Transportation Economics, 20:11–67.10.1016/S0739-8859(07)20002-2Search in Google Scholar

Wilner, F. H. (1997) Railroad Mergers. Simmons Boardman Pub Co. ISBN-13: 978-0911382198.Search in Google Scholar

Wooldridge, J. M. (2003) Econometric Analysis of Cross-Section and Panel Data. Cambridge, MA, USA: MIT Press.Search in Google Scholar

Published Online: 2019-10-07
Published in Print: 2018-12-19

© 2019 Walter de Gruyter GmbH, Berlin/Boston

Downloaded on 28.3.2024 from https://www.degruyter.com/document/doi/10.1515/rne-2018-0009/html
Scroll to top button