Order No. 2015-R-91

June 8, 2015

DETERMINATION by the Canadian Transportation Agency of the variable portions of railway company cost accounts for the Canadian National Railway Company and the Canadian Pacific Railway Company.

Case number: 


[1] The Canadian Transportation Agency (Agency) has the mandate to, among other things, set interswitching rates, provide costing information needed by parties involved in a Final Offer Arbitration process, and determine costs in the context of the maximum revenue entitlement program and disputes between railway companies and public passenger service providers. The Rail Costing System is a critical tool in the support of the Agency fulfilling those mandates.

[2] In 2009, the Agency launched a review of the variability factors used in its costing model which had been fixed based on regression analyses performed between 1992 and 1997. This review was due to concerns that as railway operations have changed significantly from that time, the fixed variabilities had become outdated and may no longer reflect the current relationships between cost and traffic volume. Furthermore, new statistical techniques were found that could be applied to the available railway company data. Further background on the review is set out in the Report – Development of Variabilities which is attached to this Order.

[3] The review was conducted by Agency staff, and consisted of the following steps: 

  • staff reviewed the past regression analyses and shared this with CN and CP staff.
  • Agency staff provided its preliminary results on the analysis to determine up-to-date variabilities for CN and CP cost accounts, to the railway companies. Responses to these preliminary results were filed by CN and CP.
  • Agency staff undertook an expansion of the analysis to consider additional alternative statistical analysis for each cost account.


Non–railway company party participation

[4] As part of its normal practice, the Agency regularly launches industry-wide consultations to seek input from industry stakeholders on issues of interpretation and methodology before the Agency makes its final determination on these issues.

[5] By holding these consultations, the Agency is generally seeking to obtain views of participants as to whether a proposed methodology or interpretation is adequate for the Agency to properly capture, analyze and take into account railway company confidential data in a manner that is consistent with the requirements of the CTA. Confidential railway company information, including the railway companies’ costing manuals, is not shared with the public or any other participant in these consultations, nor is it in the context of any proceeding in which the Agency staff’s unit cost determinations are used.

[6] In this case, the review of the variable portions of the prescribed railway companies’ cost accounts required the Agency to analyze railway company data to determine relationships between traffic volume-related quantities and costs for defined railway activities.

[7] This exercise did not involve purely methodological concepts that could meaningfully be considered separately from the data to which they will apply.

[8] For this reason, the Agency concluded that it was not practical or appropriate to hold an industry-wide consultation.


Does the Agency accept the recommendations of Agency staff set out in the Report – Development of Variabilities?


[9] CN and CP’s comments on the initial and revised variability estimates recommended by Agency staff raised the following issues.

Issue 1: Statistical methodologies employed

[10] The Agency notes CN’s concerns with the statistical methods employed are with regard to the use of a filter to remove time trend components from data which comprised time series observation, the effects of the lambda parameter in the filtering process, and the post-filter shift applied to the observations to permit statistical analysis.

[11] With respect to the need for a filter and the effects of the lambda parameter, the Agency is satisfied, based on the analysis performed by staff of relationships that rely on both filtered and unfiltered data, and relationships that employ different values of the lambda parameter, that using the filter and the lambda parameter is appropriate and allows to arrive at variability estimates that represent the best statistical properties in the underlying models as well as the most reasonable based on experience and understanding of railway operations.

[12] With respect to the post-filter shift in the observations, the Agency accepts CN’s concern that the variability estimate could be influenced by the size of the shift employed. However, as noted in Agency staff’s explanation to CN, only two shift locations are recognized in standard practice.  The Agency notes that its staff analyzed models that are based on both shift locations to select the variability estimates.  The Agency is of the opinion that this represents the best statistical properties in the underlying models as well as the most reasonable based on experience and understanding of railway operations.

[13] Furthermore, while Agency staff concluded that the shift to the origin would in principle provide the best vantage point for observing the variability, they nevertheless considered the alternative shift to the mean of the sample of the raw inflation-free data. While the a priori expectations of Agency staff were confirmed, in that the shift to the origin displayed in the vast majority of cases results that were more statistically robust and intuitively reasonable, for a few cost accounts Agency staff accepted and recommended variabilities derived from the alternative shift when warranted.

[14] The Agency finds, therefore, that the statistical methods employed by staff and staff’s proposed variability estimates represent the best empirical estimates of the relationships between the costs of specific railway activities and the traffic volume related variables explaining these costs, given the current costing structures employed by the Agency and the constraints of the data provided by CN and CP.

Issue 2: Use of third party experts

[15] The Agency notes CN’s suggestion that the Agency gets its review results confirmed by third party experts.

[16] The Agency and its predecessors have significant experience in rail costing, having performed that role as part of its legislative mandate for over 45 years.

[17] The Agency’s specialized expertise of the Agency in the interpretation of the CTA has been recognized by the Supreme Court of Canada in Council of Canadians with Disabilities v. VIA Rail Canada Inc., [2007] 1 SCR 650:

The Canada Transportation Act is highly specialized regulatory legislation with a strong policy focus.  The scheme and object of the Act are the oxygen the Agency breathes.  When interpreting the Act, including its human rights components, the Agency is expected to bring its transportation policy knowledge and experience to bear on its interpretations of its assigned statutory mandate: Pushpanathan, at para. 26.

[18] The Agency’s specialized expertise with respect to costing functions has also been recognized in Canadian National Railway Company v. Canadian Transportation Agency, 2008 FCA 363:

In my view, the determination of the Agency of the type of cost information that it required in order to make the Clause 57 adjustment is a matter that falls squarely within the expertise of the Agency, whether that determination is one of fact, discretion policy or mixed fact and law. Accordingly, I am of the view that this determination must be shown to have been unreasonable before any judicial intervention can occur.


In my view, these determinations are integral to the determination of the types of costs that are inherent in the determination of the Clause 57 adjustment. As such, they are squarely within the expertise of the Agency as determinations of fact or mixed fact and law, attracting the deferential standard of review. In my view, the railways have not demonstrated that any of the determinations made by the Agency in respect of these three matters is unreasonable. Accordingly, I am not prepared to intervene in relation to any of those matters.

[19] Section 20 of the CTA authorizes the Agency to retain the services of outside experts to provide advice. However, this is a discretionary power that the Agency will generally exercise when relevant internal expertise is not available. In this case, the Report – Development of Variabilities was prepared by staff working in the Economic and Cost Analysis unit of the Agency’s Industry Determinations and Analysis Directorate. Having reviewed the Report – Development of Variabilities, considered the level of analysis and the qualifications of the staff that prepared it, the Agency is satisfied that the relevant internal expertise required to complete the review of the of the variability factors used in the Agency’s costing model, and that no additional outside expertize was therefore warranted.

Issue 3: The use of empirical methods to determine the variabilities

[20] CN, in its comments on both the initial and revised staff variability estimates, submits that it does not consider an empirical approach to the estimation of the variable portions of cost accounts to be appropriate. Rather, CN recommends a conceptual approach in which each cost account would be judged to be either 100 percent or 0 percent variable “based on common sense, experience, and knowledge of railway operations”.

[21] The Agency notes, however, that CN’s proposed concept of variability is not the same as the variability used in regulatory cost determinations since 1969, as described by the Agency’s predecessor in Reasons for Order No. R-6313 Concerning Cost Regulations, pages 329-331, issued in August 1969 following the Cost Inquiry. CN appears to define variability as whether or not an activity is variable or fixed based on a general or vaguely-defined concept of railway operations. In other words, CN proposes an approach where a cost is presumed to be totally variable or totally fixed, based on purely subjective considerations, irrespective of whether the evidence supports this presumption or not.

[22] But, for regulatory purposes, variability is the portion of a cost activity that is explained by changes in selected traffic-volume related factors thought to cause that cost. This can only be determined by empirical investigation of the observed relationship between cost and the relevant traffic volume factors.

[23] The Agency also notes that CN presented a number of examples of cost accounts whose estimated variabilities it contends are illogical. The Agency is of the opinion that CN’s contention may be the result of this difference in the concept of variability.

[24] The Agency rejects CN’s conceptual approach to determining variability for regulatory purposes and finds that the empirical approach used by staff is the most appropriate for these determinations given the information currently available.

Issue 4: Implementation of new variabilities for cost determinations

[25] CP, in its letter to the Agency, maintains that a causal relationship must be established between an expense and the best-fitted driver before deriving a variability to determine how an expense varies in the long run with level of output. CP suggests that, for this reason, the variabilities should not be updated until there has been a proper review of the aggregation of activities into cost accounts and of the traffic volume related variables thought to influence each cost account.

[26] The Agency considers, however, that it is not appropriate for a review of the cost structures to proceed with the updating of the variabilities. Until such time that new aggregations of cost activities and the causal traffic volume related factors for each cost aggregation are established, the Agency must fulfil its regulatory mandates with the existing cost accounts and associated traffic volume‑related factors.

[27] Given that the existing cost structures must remain in place for the time being and given the data currently available, the Agency finds that the proposed new variabilities are far more statistically robust and far better reflect current railway operations than the existing variabilities that were determined based on 1992-1997 regressions. For this reason, the Agency rejects CP’s proposal to defer implementation of any new variabilities until an overhaul of the cost structures is completed.

[28] The Agency agrees with CP, however, that a more fundamental review of the existing aggregations of cost activities, the appropriate traffic-related variables influencing the cost activities, the appropriate data for analyzing the cost and traffic volume relationships, and the appropriate empirical methodologies for analyzing these relationships, are needed and would further improve the current model.

Issue 5: The variable/fixed ratio and contribution to fixed costs

[29] When the variable portions of all cost activities are calculated using the new estimated cost account variabilities and summed, the overall variable portion of total CN 2012 costs decreased from 84 percent to 64 percent, and the overall variable portion of total CP 2012 costs decreased from 83 percent to 62 percent. The higher proportions of fixed costs for both CN and CP than with the current variabilities is in better alignment with the nature of the railway industry as a high-fixed-cost capital-intensive industry.

[30] This means, for CN, the system average contribution to fixed (constant) costs that must be applied to system variable costs to cover system total costs increases from 19 percent to 56  percent, and for CP, it increases from 20  percent to 61  percent. These higher contribution levels are also in better alignment with U.S. practice, where by legislative mandate a contribution of up to 80 percent on the calculated variable costs is not eligible for rate disputes before the Surface Transportation Board.

[31] The Agency agrees with CP that the recommended new variabilities would result in a shift of the overall variable/fixed ratio for Canadian railway company costs from the 80/20 that has been the standard perception in the industry for decades to one closer to 60/40.

[32] CP expressed concerns about the impact of the new variabilities on the existing regulatory rates and on commercial negotiations. The Agency considers that whether the new variabilities will result in Agency rates being set at levels that are lower than the current ones is irrelevant to determining the proportion of the costs that are fixed and variable. What matters is that the most reasonable approach be adopted for determining the costs that are fixed and variable.

[33] The establishment of rates by the Agency, while clearly being informed by the determination of the railway company costs, is a distinct exercise. There is a long-established tradition, grounded in the legislative framework, where the Agency sets rates that cover at least the long‑run variable costs of a given railway activity, to which is added a certain contribution to cover the fixed costs of the activity. Rates set by the Agency must also be commercially fair and reasonable. 

[34] Since 2009, in exercising its discretion to set rates that are commercially fair and reasonable, the Agency has derived the contribution to fixed cost as the factor required to bump up the calculated variable cost to equal the pre-determined total cost. In the Regulatory Impact Analysis Statement published in the Canada Gazette as part of the Regulations Amending the Interswitching Regulations (2012) the Agency stated:

Since the last review in 2004, the Agency adjusted its methodology to be more robust in calculating the contribution to constant costs based on actual railway data to better reflect the difference between variable costs calculated pursuant to the Agency’s Regulatory Costing Model and the total costs incurred by the railway companies as part of information supplied to the Minister of Transport. The Agency finds that using contribution to constant costs based on actual empirical data is a valid practice and the proposed contribution of 20.3% is fair and reasonable to all parties.

[35] To date, the adjusted methodology to base the determination of the level of contribution to constant costs on the calculated railway company costs has been applied by the Agency in the following regulatory cost determinations:

  • Agency Decision No. 333-R-2009, Application pursuant to sections 152.1 and 152.2 of the Canada Transportation Act – For an order fixing the price to be paid by VIA Rail Canada Inc. for the use of certain railway, land, equipment, facilities, or services of Hudson Bay Railway Company;
  • Agency Decision No. 118-R-2011, Application pursuant to sections 152.1 and 152.2 of the Canada Transportation Act – For an order fixing the price to be paid by VIA Rail Canada Inc. for the use of certain railway, land, equipment, facilities, or services of Goderich-Exeter Railway Company Limited;
  • Regulations Amending the Interswitching Regulations (2012); and,
  • Regulations Amending the Interswitching Regulations (2014).

[36] The Agency has also applied the same methodology for the rates published in the Guide to Railway Charges for Crossing Maintenance and Construction for the years 2012, 2013 and 2014.

[37] The Agency does not agree with CP when it suggests that the new methodology would have the effect of lowering the existing rates, given how the Agency has chosen since 2011 to exercise its power in setting rates that are commercially fair and reasonable.

[38] Finally, the Agency notes CP’s concerns about this determination of the Agency being misinterpreted in commercial negotiations. The Agency considers that it should not refrain from making a methodology determination that it finds warranted for the Agency to continue to properly fulfill its regulatory mandate under the CTA, based on an alleged concern that someone may potentially adopt a given commercial negotiation strategy due to that person’s misinterpretation of the Agency’s decision. 

[39] Furthermore, the Agency has always been clear in its communications and decisions that the proportion of the total system costs that are variable and fixed form a whole that cannot be considered independently.  One could not therefore add to the costs that were deemed previously to be variable the proportion of the costs that are now estimated to be fixed based on the more robust statistical techniques. Conversely and for the same reasons, one could not use the current estimate of the variable costs based on the more robust statistical techniques, and then add the prior estimate of the proportion of fixed costs. 

[40] The Agency finds that it cannot, when determining the portions of railway company costs that are fixed or variable, take into account how the Agency would choose in the future to exercise its discretion to set rates that are commercially fair and reasonable to all parties, nor how commercial negotiations might be affected by its determination.


[41] The Agency has reviewed the Report – Development of Variabilities and accepts the findings and recommendations contained in that Report.

[42] Further the Agency directs staff to undertake a comprehensive indepth examination of the existing cost models and to report back their findings to the Agency by April 1, 2016. The examination should include:

  • Review of the aggregations of railway activities for cost determination purposes;
  • Review of the traffic volume related cost drivers for each cost aggregation;
  • Review of the appropriate data for determining the relationships between cost and causal factors;
  • Review of the appropriate empirical methods for deriving the causal relationships.

[43] The Agency will discontinue using the variabilities used to date.

  1. The Agency approves staff’s recommended variabilities for use in the determination of the 2012 and 2013 unit costs for CN and CP. The approved variabilities will be communicated separately to CN and CP.
  2. The Agency directs staff to update annually the variabilities to be used in subsequent annual unit cost determinations, pending the adoption of a new Rail Costing System.


Sam Barone
P. Paul Fitzgerald
Date modified: