Skip to main content

Currently Skimming:

Appendix B: DEMONSTRATION OF COMPETITIVE RATE BENCHMARKING TO IDENTIFY UNUSUALLY HIGH RATES
Pages 225-259

The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.


From page 225...
... e shipment's competitive benchmark rate would be predicted on the basis of this information. Because of the impracticality of having complete information on all of a shipment's economically meaningful characteristics, perfect correspondence between the tested rate and the rate predicted from the benchmark group is unlikely.
From page 226...
... Such data can be compiled, as demonstrated in this appendix. Nevertheless, further refinement of the competitive rate benchmarking approach may reveal other data needs, including details on shipment characteristics that currently are not in CWS records or readily available through other databases.
From page 227...
... First, however, the reason for considering the development and introduction of a competitive rate benchmarking tool is recapped on the basis of the discussion in Chapter 3. RATIONALE FOR COMPETITIVE RATE BENCHMARKING e Staggers Rail Act of 1980 charged regulators with protecting shippers from unusually high common carrier rates when they have few competitive options.
From page 228...
... e benchmarking models developed and demonstrated next illustrate the types of statistical models that can serve this purpose. OVERVIEW OF A COMPETITIVE RATE BENCHMARKING MODEL Data on shipment characteristics and rates in effectively competitive markets are used to construct a predicted, or competitive benchmark, rate for any given rail shipment in a potentially noncompetitive market on the basis of key observable characteristics of the shipment.
From page 229...
... Once the effectively competitive benchmark model has been constructed for each commodity, a shipper could determine how close its common carrier rate is to the competitive benchmark rate for shipments having the same set of observable characteristics but in markets with effective competition. When such tests are performed, a significant fraction of rates tested will exceed the competitive benchmark rate even if pricing is not affected by the level of competition.
From page 230...
... If the CWS is used as the primary mechanism for gathering the data for estimating competitive benchmark price models of the type described in this appendix, quality controls must be in place to ensure that the sampling scheme for compiling shipment rates and characteristics is in fact random and stratified in a way that allows a valid estimate of the annual population joint distribution of rates and shipment characteristics to be computed.
From page 231...
... e user may need to enter only the shipment size, the railroads used, the commodity, and certain other shipmentspecific variables to find the benchmark price for its shipment. DETAILS OF THE MODELS DEMONSTRATED Benchmark and Nonbenchmark Samples To establish the pool of effectively competitive shipments to estimate the conditional quantile functions and the pool of shipments lacking effective competition to apply the models, CWS records from 2000 through 2013 are divided into two groups, as described below.
From page 232...
... e presence of effective competition is defined as one alternative rail option within 10 miles of the origin and the destination, water ports on the same waterway within 50 miles of both the origin and the destination, or both circumstances.2 e potentially noncompetitive group consists of all shipments that were moved by common carriage and the subset of shipments of nonexempt commodities that were moved by contract and have no effective rail or water competition. Data Sources As noted, the primary source of data for developing and testing the benchmark models was the CWS from 2000 through 2013.
From page 233...
... Finally, all rates from the CWS were adjusted to constant 2009 dollar values by using the gross domestic product price deflator available from Federal Reserve Economic Data through the Federal Reserve Bank of Saint Louis.6 Estimation Model and Variables In the approach illustrated here, shipment rates (rate) are modeled as a function of shipment distance (X1)
From page 234...
... Imposition of a linear functional form restriction on the conditional quantile function is unnecessary. is restriction is imposed for the current application as a means of simplifying the presentation.
From page 235...
... = α. e elements of the X-vector described above and all of the conditional quantile functions estimated are assumed to have the following parametric form: ∑)
From page 236...
... From 2009 forward, CWS has had a separate field for fuel surcharges. In the calculation for ton-miles, the variable "billed weight" was used for tons, and distance was calculated as the "total miles traveled for the shipment." e explanatory variables used in the model are based on past econometric studies, many of them cited in Chapter 1, that examine how rail rates relate to shipment characteristics such as distance, shipment size, and number of railroads involved in the shipment, as well as various measures of intramodal and intermodal competi
From page 237...
... Shipment distance, shipment size, the number of railroads involved in the movement, and the private cars dummy variable are directly observed in the CWS or are easily constructed from the data. Railroad competition is measured as the number of Class I railroads within 10 miles of the origin and of the destination.
From page 238...
... the validity and integrity of the random sampling scheme used by the CWS; (b) the criteria to be used in identifying the set of shipments to be included in the effectively competitive sample used to estimate the competitive benchmark rate function; (c)
From page 239...
... estimates are also reported in the tables as an informal specification test of the functional form for the linear conditional quantile functions.14 As noted, the models are applied with only the 2013 test group observations. Two graphs are provided showing the distribution of the actual-to-predicted rates for the 2013 test group.
From page 240...
... Important factors to consider in making such determinations are the number of likely excluded shipment characteristics that have economic meaning and the precision with which the conditional quantile function is estimated. However, the competitive rate benchmarking process is intended only to identify rates that are unusually high and deserving of further scrutiny; it is not intended as the final arbiter of rate reasonableness.
From page 241...
... Increases in shipment distance and shipment size tend to predict lower rates (revenue per ton-mile) , while increases in the number of railroads involved in the shipment tend to predict higher rates.
From page 242...
... OLS estimates are reported as an informal specification test of the functional form for the linear conditional quantile functions.
From page 243...
... e median regression model in Table B-2 is used to predict their competitive benchmark rates. e ratios of actual rate to predicted rate for the 6,319 shipments are summarized in Figures B-1 and B-2 and in Table B-3.
From page 244...
... The columns showing expanded percentages use the sample rate expansion factor associated with each observation.
From page 245...
... Coal e descriptive statistics for coal are provided in Table B-4. ere are 446,820 total observations, with 291,431 in the competitive benchmark sample and 155,389 in the nonbenchmark sample.
From page 246...
... Rail competition at the origin or destination predicts lower prices in all specifications. e presence of water competition and shorter distances to water from the origin and destination both predict lower rates, while the use of private cars predicts lower rates.
From page 247...
... −�.��� (�.�����) Observations ���,��� ���,��� ���,��� ���,��� ���,��� R� �.��� ����: Based on competitive benchmark data.
From page 248...
... 248 MODERNIZING FREIGHT RAIL REGULATION 0 .5 1 1.5 2 2.5 D en si ty 0 1 2 3 4 Actual Rate/Predicted Competitive Rate FIGURE B-3 Distribution of ratios of actual to predicted rates, nonbenchmark sample, coal, no ratios excluded. 0 1 2 3 D en si ty .5 1 1.5 2 2.5 3 Actual Rate/Predicted Competitive Rate FIGURE B-4 Distribution of ratios of actual to predicted rates, nonbenchmark sample, coal, ratios greater than 3 excluded.
From page 249...
... The columns showing expanded percentages use the sample rate expansion factor associated with each observation. in the competitive benchmark group and 198,469 in the nonbenchmark group.
From page 250...
... e magnitudes are also stable across columns. Increasing shipment distance and size both predict lower rates (revenue per ton-mile)
From page 251...
... �.���� (�.�����) Observations ���,��� ���,��� ���,��� ���,��� ���,��� R� �.��� ����: Based on competitive benchmark data.
From page 252...
... 252 MODERNIZING FREIGHT RAIL REGULATION 0 .2 .4 .6 .8 1 D en si ty 0 5 10 15 20 Actual Rate/Predicted Competitive Rate FIGURE B-5 Distribution of ratios of actual to predicted rates, nonbenchmark sample, chemicals, no ratios excluded. 0 .5 1 1.5 D en si ty 0 1 2 3 Actual Rate/Predicted Competitive Rate FIGURE B-6 Distribution of ratios of actual to predicted rates, nonbenchmark sample, chemicals, ratios greater than 3 excluded.
From page 253...
... e average distance traveled is 793 miles, and the average shipment size is 1.1 cars. Most shipments involve only one TABLE B-9 Chemicals Model: Distribution of 2013 Test Group Observations, Ratios of Actual Rate to Benchmark Rate Observations Expanded Group Contract No.
From page 254...
... ere is little difference across the two samples in most variables other than the water options and the distance to water for both origins and destinations. The estimation results are summarized in Table B-11 with the intercept effects (railroad dummies, annual dummies, and STCC TABLE B-10 Petroleum Summary Statistics, 2000–2013 Variable Combined Samples Benchmark Sample Nonbenchmark Sample Observations ��,��� ��,��� ��,��� Average revenue per ton-mile (���� dollars)
From page 255...
... �.���� (�.���) Observations ��,��� ��,��� ��,��� ��,��� ��,��� R� �.��� ����: Based on competitive benchmark data.
From page 256...
... A major problem for regulators has been 0 .5 1 1.5 D en si ty 0 2 4 6 8 10 Actual Rate/Predicted Competitive Rate FIGURE B-7 Distribution of ratios of actual to predicted rates, nonbenchmark sample, petroleum, no ratios excluded.
From page 257...
... of ARTM to the predicted 50th percentile. The columns showing expanded percentages use the sample rate expansion factor associated with each observation.
From page 258...
... For the most part, the tested rates were close to the competitive rates, but the procedure identifies traffic having rates that far exceed the competitive benchmark rate. ese rates might be candidates for further scrutiny for reasonableness.
From page 259...
... 1989. e Effects of United States Railroad Deregulation on Shippers, Labor, and Capital.


This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More information on Chapter Skim is available.