APPENDIX B TO DECISION NO. 425-R-2011

Summary of Methodological Issues in the Review of the Methodology Used By the Canadian Transportation Agency to Determine the Cost of Capital for Federally-Regulated Railway Companies

1.0 Introduction

[1] This document provides a summary compilation of information that was integral to the consideration by the Canadian Transportation Agency (Agency) of the issues related to the methodology to be used within its legislative mandate to determine the cost of capital for federally-regulated railway companies. It also serves as a backdrop to the analysis, findings, and decisions of the Agency in respect of its cost of capital review.

[2] It includes a restatement of the issues raised for and during the methodology review and consultation, the context and relevance of those issues, a summary of participant submissions on each of the issues raised, as well as the practice of other regulators and excerpts from the Brattle Report, where applicable.

1.1 Cost of capital - General

[3] Cost of capital is defined as an estimate of the total return on net investment that is required by shareholders and debt holders so that debt costs can be paid and equity investors can be provided with an adequate return on investment consistent with the risks assumed for the period under consideration.

[4] There are many different ways to estimate cost of capital, none being deemed perfectly correct. According to the Brattle Group, "…it must be acknowledged that cost of capital estimation continues to be as much art as it is science".[1] Each input at each step may be altered to meet the purpose of the cost of capital rate. For example, an investor may use market values to calculate the firm's market structure or to calculate the firm's cost of debt in order to calculate a cost of capital rate which fully meets its opportunity costs. The problem in this case is whether buying, selling or holding a share at the current market price is consistent with the risk/reward expectations of a particular investor at a particular point in time. However, a regulator may choose to use book values for the same inputs for various reasons, including providing a fair and reasonable compensation for the investments actually made by the regulated party and maintaining a stable rate environment.

[5] The cost of capital is more commonly expressed as the weighted average cost of capital (WACC), expressed either on a before or after tax basis, depending on its application. The WACC is comprised of the cost rate of debt and the cost rate of equity, weighted to the relative components of each in an entity's capital structure. Of the two cost rates, estimating the cost of equity is the more controversial, least directly observable of the two. What is common though in all regulatory practices is that the cost of equity is always established on an expected basis. In reality, the equity market can and does generate negative returns in any time period, yet regulators would never set a negative return on capital for a given time period even if the return is expected to be negative in the time period for which the rate applies. The converse of this proposition is also true. Regulators do not attempt either to mimic the extremely high returns that the equity market can sometimes experience in a given time period. Intrinsic in regulatory practices is the notion of a return on equity that is sustainable over many different time periods.

[6] The two most commonly used methods for estimating a cost of common equity rate are the Capital Asset Pricing Model (CAPM) and the Discounted Cash Flow (DCF) Model. The CAPM is a method used to estimate the cost of common equity capital by comparing the return and the risk characteristics of an individual company's shares to the market average. For any individual security, it sorts risks into two categories: those that can be diversified away (and for which the investor is not compensated) and those systematic risks that the marginal equity holder is unable to diversify away (and for which the investor is compensated).

[7] The returns established by the CAPM fall into three components:

  • The risk-free rate, which provides the minimum returns available in the capital asset markets, and which is set to reflect the financial conditions in the target time period. This component therefore responds very rapidly to the conditions of the financial market and acts as the "elevator" that drives the return structure up and down;
  • The market risk premium (the difference between the market returns and the returns provided by the risk-free asset), which is set using information over a very long time period and is thus very stable; and
  • The beta, a company-specific adjustment factor to the market risk premium reflecting the systemic risk of an individual security relative to the market as a whole. Betas are set over a moderately long time period, and will thus react, though typically not drastically, to the changing circumstances of an individual security as new information comes in.

[8] The DCF Model measures the cost of common equity as the interest rate that equates the present value of the future cash flows investors expect to receive from the company over its lifetime, to the price investors are willing to pay for the company's equities.

[9] Variations of an equity risk premium (ERP) model are also sometimes used by regulators. The basic premise of an ERP model is that an investment in common equity carries greater risk than an investment in either debt, preferred shares or government securities, and therefore requires a higher return or premium over that required for those types of investments. Thus ERP models determine an entity's cost of common equity by assessing the reward that investors require to be compensated for the risk associated with holding equity compared to the return from those other lower risk instruments.

[10] Another approach to calculating cost of common equity is the comparable earnings method. The comparable earnings method involves selecting a group of companies of comparable risk; calculating the average return on book equity over an appropriate time period; and adjusting the result for any differences in risk between the target company and the comparable companies. This method requires selecting the group of companies of comparable risk, the time period over which to estimate the return on equity and the adjustment for any risk differences between the sample companies and the target company.

1.2 Regulatory application and evolution of the Agency's methodology

[11] The Agency determines cost of capital rates for three regulatory applications: as input into the determination of the maximum revenue entitlement (revenue cap) for the railway companies for the transportation of western grain; as input into the determination of rates for interswitching movements; and as input into regulatory cost determinations and other regulatory purposes. The rates determined for western grain and interswitching are forward-looking, aiming to forecast the cost of capital approximately one year into the future, while the rate determined for other regulatory purposes is retrospective, calculating the cost of capital rate for the previous year. Both the interswitching and other regulatory purposes calculations are closely connected to the Agency's costing model, which uses book values to measure rail assets.

[12] In the Agency's current cost of capital methodology, the cost of capital is determined by following four distinct steps. The foundation of the Agency's calculation is equating net rail investment with a company's capital structure. This equation allows theoretical justification for applying the company's cost of capital to its rail assets as done by the costing model.

[13] The current Agency methodology evolved through major Agency cost of capital decisions made in 1985, 1997 and 2004. The 1985 Decision was a comprehensive review including oral hearings, which set out a methodology, most of which is still being used today. The 1997 Decision was the result of the need identified by the Agency to review the 1985 Decision in light of the enactment of the Canada Transportation Act, S.C., 1996, c. 10, as amended (CTA) and developments in the railway industry, i.e., the privatization of the Canadian National Railway Company and the stated intention of Canadian Pacific Limited to restructure and create a publicly traded stock for its rail holdings in the Canadian Pacific Railway Company.

[14] On August 1, 2000, changes to the Agency's mandate resulting from amendments to the CTA enacted by Bill C-34 altered the manner in which cost of capital estimates were applied, particularly with respect to the movement of western grain, but the methodology for calculating these estimates remained consistent with those adopted in the 1985 and 1997 Decisions. The 2004 Decision was the result of a dialogue with interested stakeholders, which discussed certain recurring issues raised by the railway companies on the Agency's estimation of cost of common equity rates.

[15] Neither the 1997 Decision nor the 2004 Decision resulted in any significant changes to the cost of capital methodology, but more so, reviewed, reinforced and in some cases explained more explicitly the findings of the 1985 Decision.

1.3 Review of the Agency's methodology

[16] The current review of the Agency's methodology was conducted in two phases: a study phase and a hearing phase. In the study phase, the Brattle Group, an independent consultant retained by the Agency, examined existing cost of capital methodologies and principles, and reviewed the Agency's current cost of capital methodology, as well as the cost of capital methodologies used by other economic regulatory bodies. The study phase resulted in the Brattle Report, which identified and examined a variety of existing cost of capital methodologies and models. It includes an analysis of their strengths and weaknesses, as well as an assessment of any potential issues and implications associated with the implementation of such models.

[17] In the hearing phase, the Agency initiated a consultation with a broad list of stakeholders on certain recurring issues identified by the Agency for discussion, through the distribution of a Consultation Document. The Agency set a schedule and participants were given the opportunity to express their points of view on these issues and any others they thought pertinent to the examination, as well as to respond to the opinions expressed by other participants.

[18] This Appendix summarizes the issues that have been considered by the Agency during its cost of capital review, the positions of the participants (integrating the views of their chosen experts without necessarily identifying them separately), the analysis of the Brattle Group, and the practices of other regulatory bodies.

[19] The submitting participants whose positions have been summarized in this Appendix are:

  • Canadian National Railway Company (CN), with expert testimony provided by Ronald M. Giammarino, Ph.D. and Murray Carlson, Ph.D.;
  • Canadian Pacific Railway Company (CP), with expert testimony provided by Bruce E. Stangle, Ph.D. and George Kosicki, Ph.D.;
  • Canadian Canola Growers Association (CCGA);
  • Coalition of Rail Shippers (CRS), with expert testimony provided by Lawrence Kryzanowski, Ph.D.;
  • Western Canadian Shippers' Coalition (WCSC), with expert testimony provided by Lawrence I. Gould, Ph.D.;
  • Alberta Transportation, Province of Alberta (Alberta);
  • Manitoba Infrastructure and Transportation, Province of Manitoba (Manitoba); and,
  • Ministry of Highways and Infrastructure, Province of Saskatchewan (Saskatchewan)

[20] The full submissions of the participants are available on the Agency's Web site in the language in which they were submitted, and a copy of the Brattle Report is available in full, in English and French, upon request.

2.0 Capital Structure Weights - Book Value versus Market Value

Issue: Should the Agency use book values or market values when determining the relative weights of long-term debt and common equity in the railway companies' capital structure?

2.1 Context and relevance

[21] The capital structure of a firm refers to the combination of the various types of capital funds that a firm uses to finance its assets. In broad terms, funding can be achieved through borrowing or the issuance of debt, through deferred taxes, and through shareholders' equity. The capital structure shows the proportion of each of these financing instruments, and is based on the principle that a firm cannot have an asset without first funding it through debt or equity.

[22] Capital structure is an important component in the development of the cost of capital rate. Each type of funding has a cost rate consistent with its risk profile. The proportion of each type of funding in the capital structure is used to weight each respective cost rate and the sum of these weighted cost rates becomes the cost of capital rate, or the WACC, expressed in percentage terms, as shown below:

Mathematical equation  - see long description for explanation
Text alternative for the equation

where: is the weight of debt multiplied by the cost rate of debt; is the weight of equity multiplied by the cost rate of equity; and is the weight of deferred taxes multiplied by the cost rate of deferred taxes.

[23] This cost of capital rate, when applied to the asset base, (either net book value or market value of the assets) results in the cost of capital in dollar terms.

[24] It is apparent that the cost of capital rate depends not only on the cost rates of the capital structure components but also on the weights of the components.

[25] In the event of liquidation, the claims of debt holders are settled before those of shareholders. Furthermore, unlike shareholders, debt holders are guaranteed fixed payments by the company and have first claims of the free cash flows available. As a result, shareholders demand a higher return on equity once there is debt in the capital structure.

[26] Figure 1, based on the Modigliani and Miller Propositions I and II[2], shows that the cost of equity rises as the debt/equity ratio increases because of the risks associated with increasing debt levels in the capital structure.

Figure 1: Cost of Capital and Debt/Equity Ratio

Figure 1: Cost of Capital and Debt/Equity Ratio425-R-2011/app-b-exhibit1-en.gif" width="487" height="400" />

[27] The WACC decreases as the debt/equity ratio increases up to a certain level of debt beyond which the WACC starts to increase. The debt/equity ratio beyond which the WACC does not decrease is referred to as the optimal capital structure mix.

[28] At low levels of debt, the tax benefit of debt is higher than the associated costs (financial distress and bankruptcy) of debt. However, with an increasing debt/equity ratio beyond the optimal debt level, the associated costs of debt tend to offset the tax benefits of debt (trade-off theory) causing the WACC to increase.

[29] As the firm incurs more debt, the risk of default increases and the firm is required to pay higher rates of interest to bondholders. When the firm borrows more and more debt, the rate of increase in the expected return on equity will start to decrease. This occurs because holders of risky debt bear some of the firm's business risk. As the firm borrows more, more of that risk is transferred from stockholders to bondholders.

[30] For its regulatory purposes, the Agency currently considers only the portion of the railway company's assets that are used to provide railway transportation services under the Agency's jurisdiction, as shown in the company's books. Therefore, the asset base or net rail investment used for cost of capital determinations represents the book value of the railway company's Canadian rail assets (less accumulated depreciation plus an allowance for working capital).

[31] The three financing instruments available to the railway companies and used in the Agency's methodology are: long-term debt, deferred taxes and common equity. Long-term debt is comprised of long-term notes payable and capital lease obligations. Deferred taxes consist of deferred taxes, investment tax credits and deferred downsizing costs. Equity is the sum of share capital, contributed surplus, retained earnings and net investment in rail assets[3].

[32] To determine the relative weights for each type of funding in the capital structure, the Agency currently aligns the book value of the asset base with the book value of the liabilities and equities used to finance it, an approach that is based on the actual or acquisition cost of the asset. The sum of the three categories of financing instruments in the capital structure is equal to the net rail investment.

[33] In recent years, either one or both of the Class 1 railway companies have questioned this book value approach, advocating that an approach based on the current market value of the liability or equity be used to determine the relative weights of debt and equity instead.

[34] A pure market value capital structure has only two components, debt and shareholders' equity, with any deferred liabilities being considered as equity. The market value of debt is estimated based on the current dollar value in the market and the current market rate of yield of any marketable debt, and uses some form of proxy, or estimation model, in the absence of price data, for debt that is not traded. The market value of equity is the share price multiplied by the number of shares outstanding.

2.2 Positions of the participants

Canadian Pacific Railway Company

[35] CP takes the position that the Agency should be using market values when determining the relative weights of long-term debt and common equity in the capital structure. CP argues that market value weights are consistent with the CAPM and the multi-stage DCF methods for estimating the cost of equity, as both of these rely on market values for key inputs. CP also argues that using market value weights will eliminate the need to identify deferred tax liabilities as a component of CP's capital structure, suggesting that any economic effects associated with deferred taxes will be captured in the market value of equity.

[36] To support this position, CP cites references by Aswarth Damodaran, a noted finance professor, from his book, Damodaran on Valuation[4], which discuss cost of capital in terms of raising new funds, both debt and equity, at prevailing prices, in order to buy a firm today, and in terms of assessing the true value of a firm, which changes over time based on new information about the firm and the overall economy.

Canadian National Railway Company

[37] CN submits that the Agency should use market values and estimates of market values to determine the relative weights of long-term debt and common equity in the capital structure. CN submits that book values do not sufficiently cover the opportunity cost of the investors if book values depart from market values.

[38] CN provides a hypothetical situation to support this position, which is based on the contention that there is a direct correlation between any increases in the market value of a firm's assets and the market value of its equity, that an investor can liquidate its holdings in a company's assets at that market rate in order to reinvest in another company of comparable risk offering a higher rate of return, and that if a company was unable to match that higher rate of return it would liquidate its own assets and move to another line of business in order to avoid this.

[39] CN further submits that this scenario also illustrates the importance of using market values (replacement cost) for the asset base. It is CN's position that the use of market value based weights for capital structure is widespread by corporations, financial advisors, academia and regulators, as is, to a more limited degree, the use of market value/replacement cost for the asset base.

Western Canadian Shippers' Coalition

[40] WCSC suggests book values represent the correct measure for determining the capital structure. It advances that book values are much more stable than market values and that market value weights would dramatically increase the variability in the weighted average cost of capital.

[41] Rather than opportunity costs, WCSC speaks in terms of hurdle rates, indicating that a company's objective of maximizing the value of its common shares is linked to a company's decision to invest in projects that yield a return greater than a minimum acceptable hurdle rate; the riskier the project, the higher that minimum rate should be (i.e., the project's cost of capital).

[42] WCSC submits that the returns required on a firm's common stock are implied in the price investors are willing to pay to hold the stock. It indicates that as long as the return the company earns on its common equity is equal to the return investors require on the stock, existing shareholders neither gain nor lose, that is, their expectations are met. When the return on common equity is above the return investors require, each dollar of additional financing raises the value of existing shares.

[43] WCSC further submits that when firms raise incremental funds to maintain a target capital structure, they do it through book values. A firm would not use market values to maintain a target capital structure as it would not know with 100 percent certainty what the market value would be. As incremental funds are based on book values, the earnings to cover capital costs must be computed using book value weights as well.

[44] WCSC considers that the determination of the cost of capital using book value weights applied to the historical cost investment has been widely accepted in regulatory proceedings for Canadian companies, and has allowed them to raise the capital necessary to meet the demand for services without an adverse effect on current shareholder stock.

Canadian Canola Growers Association

[45] CCGA indicates that the Agency's current approach for calculating the weights of long-term debt and common equity seems reasonable, though it appears that CCGA is under the impression that the Agency's current methodology is based on the use of market values for equity, which is a misinterpretation.

[46] With respect to the use of book values to determine long-term debt, CCGA notes that determining the market value of thinly traded debt instruments requires extensive analysis with respect to credit ratings, debt instrument duration and coupon rates, as well as an adjustment to reflect real and/or perceived factors affecting the railway company specifically, relative to its peers.

Coalition of Rail Shippers

[47] CRS also recommends the use of book values. CRS examines the issue through the price (P) to book value (BV) ratio, where price is the market value of the stock. If a firm earns its fair return on new assets, then the P to BV ratio will be equal to one as price will still be equal to book value. CRS submits that the notion that each regulated entity should maintain a market value above book value is contradictory, as it suggests that each regulated entity should plan to earn a return on new investment above the required rate of return. CRS considers that an examination of P to BV ratios provides a test of reasonableness of any cost of equity estimates.

Province of Alberta

[48] Alberta submits that the Agency should continue determining the capital structure using book values, which are more transparent. Alberta also cites difficulties in calculating the market value of long-term debt, which result in the use of book value in any event, and thus redefines the issue as whether a market value estimate of common equity is appropriate.

Province of Manitoba

[49] Manitoba submits that the Agency should use the book value of the components of capital structure. Because the market value of common equity is driven by investors' perceptions of prospective profitability and the Agency's findings as to cost of capital affect that profitability, it considers that the use of market values makes the regulatory process circular, and therefore not reasonable. It describes the circularity as: a generous cost of capital results in increased profitability, which in turn increases the market value of equity, which in turn increases the Agency's estimate of the cost of capital.

[50] Manitoba submits that book values are, in contrast, unaffected by the Agency's decision because they are based on the historical record of investment. It considers book values to be reliable and pragmatic because they are stable and transparent, whereas market values vary daily, if not hourly. Manitoba further submits that it is both inconsistent and incorrect to estimate the cost of capital based on market value and apply that estimate to a book value asset base, as would be the case for the investments to which the Agency's cost of capital findings are applied.

Province of Saskatchewan

[51] Saskatchewan did not comment on whether the Agency should use market values or book values to determine capital structure.

2.3 Rebuttal comments

Canadian Pacific Railway Company

[52] CP reiterates its position that the capital structure weights should be based on market values, citing Damodaran to the effect that as the cost of capital is a forward-looking measure and captures the cost of raising new funds to buy the firm today, and as new debt and equity have to be raised in the market at prevailing prices, the market value weights are more relevant. CP again cites Damodaran to counter arguments that the WACC based on market values is too volatile, quoting the statement that market value, even with its volatility, is a much better reflection of true value than is book value.

Canadian National Railway Company

[53] CN suggests that WCSC does not provide any theoretical support for its contention that book values should be used to determine the relative weights of long-term debt and common equity. CN also states that the illustrative example presented by WCSC is vague and unclear as to how it supports the conclusion as there is no mention of asset or liability values, either book or market.

Western Canadian Shippers' Coalition

[54] WCSC notes CN's contention that if book values depart from market values, market value weights should be used in calculating the WACC, and submits that this position is based on an incorrect assumption that, under regulation, the market-determined opportunity cost of capital should be applied to the market value of the assets under regulation. WCSC asserts that this incorrect assumption extends across all the examples that CN provides.

[55] WCSC quotes CN's conclusion that "This example illustrates the importance of using market values for the asset base and market values for the weights used in the WACC calculations." It notes that CN does not explain how the Agency could obtain the market value of these assets, without which the computation is not possible, and asserts that the process would be very complex and contentious. WCSC also notes that according to CN, the market-determined cost of equity capital would be constantly changing to reflect the changing market value of equity, yet CN provides no guidance on how the Agency should make this adjustment. WCSC suggests that CN makes the mistake of trying to use financial theory that applies to unregulated companies to apply to regulated companies.

[56] WCSC uses examples to illustrate its position that the expected return is a direct link between the book value investment and share price, and that as long as a firm is expected to earn its required return on the book value of its assets, it will be able to raise capital for new investment and maintain its financial viability. WCSC notes that this view has been widely accepted in regulatory proceedings for Canadian companies, and has allowed the regulated companies to raise the capital necessary to meet the demand for services without an adverse effect on current shareholder stock. WCSC asserts that the basis of this rate-setting process is the determination of the cost of capital using book-value weights applied to the historical cost investment.

2.4 Practice of other regulators

[57] The United States Surface Transportation Board (STB) uses a market­-value approach to determine the capital structure for its cost of capital purposes.

[58] For its annual cost of capital rate determination, STB creates a composite railroad using Class 1 railroads that meet certain criteria. The criteria are that the Class 1 railroad: (1) be listed on the New York Stock Exchange or the American Stock Exchange; (2) paid dividends throughout the year; (3) had rail assets greater than 50 percent of its total assets; and (4) had a debt rating of at least BBB (Standard & Poor's) and BAA (Moody's).

[59] The market value of equity is determined by multiplying stock prices of each railroad by the number of outstanding shares filed with the U.S. Securities and Exchange Commission.

[60] Since 1988, STB has estimated debt costs as the "current cost of debt" for both capital structure and cost estimates.[5] It attempts to measure market yields of debt instruments, and apply these yields to the market values of the debt. In recent years, due to the absence of sufficient new debt being issued, the debt cost has consisted of the market values of publicly traded debt instruments (including bonds, notes and debentures) and estimated costs of other forms of debt financing.[6] STB estimates are based on a joint submission by the American Association of Railroads (AAR) on behalf of U.S. Class 1 railroads subject to cost of capital proceedings.

[61] The debt instruments used by STB to determine the portion of debt in the capital structure include: (1) bonds, notes and debentures; (2) Equipment Trust Certificates (ETC); and (3) Conditional Sales Agreements (CSA). The yields of these instruments are weighted based on their market values. STB does not include the cost of capital leases and miscellaneous debts in the average cost of debt, as it considers that the costs of these instruments are not directly observable in the open market. These debt instruments are explained further below.

[62] All bonds (including notes and debentures) used to finance a railway company, whether traded in that year or not, are included in the AAR submissions. For bonds traded during the year, the current market value is developed on the basis of monthly prices and yields. For bonds not traded during the year, (approximately 40 percent of all bonds), the face value is used. Information on market prices and yields is sourced from an independent third party (such as Standard & Poor's Bond Guide in 2009).

[63] Other forms of debt financing include ETC, CSA and capitalized leases and miscellaneous debt. ETC and CSA are both debt instruments where a company takes possession of an asset without fully paying for it, and the asset title is used by lenders to secure the asset until it is fully paid for. ETC are not actively traded on secondary markets, so government securities with similar maturities are used as surrogates for developing yields with an addition for greater risk. This additional risk factor is approximated from the difference between ETC and government bonds at the time of the most recent ETC issue (which may be in a previous year). The value of ETC issued at floating interest rates rather than fixed is not modelled, nor are ETC which do not have "all of the characteristics typical of an ETC."[7]

[64] The cost of CSA is estimated by adding an additional factor to the spread between government bonds and ETC. This additional CSA factor is measured as the spread between the CSA and ETC in the year the CSA was issued. This CSA factor is added to the ETC factor described above. The book value of capitalized leases and miscellaneous debt is added to the overall debt structure, but no debt rate is applied to these factors. Similarly, transaction costs of issuing the debt (called flotation costs) are also added to the debt structure without applying a rate to the debt.

[65] In 2009, complaints were made by the Western Coal Transportation League, which claimed that no spreadsheets were provided for cost-of-debt calculations, and that the AAR's calculations could not subsequently be replicated. However, STB found the AAR's cost of debt calculations generally correct, with some adjustments.

[66] In direct contrast to its use of market values in calculating the capital structure weights for cost of capital determinations, STB uses the historical cost approach (book value) to calculate net investment for the determination of return on investment (ROI). ROI is used in conjunction with the cost of capital rate to make the annual revenue adequacy determination. A railroad is considered to have adequate revenue for that year if its ROI exceeded the cost of capital and inadequate revenue if its ROI was less than the cost of capital rate.

3.0 Treatment of Deferred Taxes

Issue: Should the Agency continue to include and give weight to deferred taxes in the capital structure, and if so, what rate should be assigned?

3.1 Context and relevance

[67] Deferred income tax balances arise because a company may, through the claiming of capital cost allowances, depreciate fixed assets for income tax purposes at a faster rate than it depreciates the same assets for accounting purposes. This is due to a variation in the reporting requirements of income and expense for regulatory (accounting) purposes, as set out in the Canada Business Corporations Regulations, 2001, SOR/2001-512, and for income tax purposes, as set out in the Income Tax Act, R.S.C., 1985, c. 1 (5th Supp.), where a company may recognize an item (income or expense) for one reporting purpose and the same item may not be required to be reported in the same manner for the other.

[68] Regulatory reporting is based on Generally Accepted Accounting Principles (GAAP). A fundamental principle of GAAP is to match revenues and expenses in the period in which they occur. One item in the income statement that is affected by the variation in the reporting requirements and is significant to the books of the railway companies is amortization expense. The difference in the reporting requirements therefore results in a variation in the taxes payable depending on whether the financial statements were prepared for regulatory or for income tax purposes. As such, businesses create an account called accumulated deferred income taxes on their balance sheets to show the income tax balances in order to better reflect their financial position. The Uniform Classification of Accounts (UCA) recognizes deferred taxes as either an asset or liability depending on whether the reporting variation results in income tax benefit or income tax payable[8].

[69] The treatment of the balance of the accumulated deferred income taxes account has an impact on the cost of capital rate. There are generally two acceptable approaches for the treatment of deferred incomes taxes as far as capital structure and the cost of capital rate are concerned: the normalization approach and the flow through approach[9].

[70] The normalization approach can be applied in two ways. Using normalization type 1, deferred taxes are included in the asset base, and in the capital structure with a zero cost rate applied to them. This is the approach currently used by the Agency. At a zero cost rate, the inclusion of deferred income taxes in the capital structure takes up some weight and results in a lower cost of capital rate.

[71] Using normalization type 2, deferred taxes are deducted from the asset base (net rail investment), and both the asset base and the capital structure are adjusted accordingly. The exclusion of deferred taxes under this approach leaves the capital structure to be divided only between debt and equity, and also results in a lower cost of capital rate.

[72] Under the flow through approach, deferred taxes are applied dollar for dollar to reduce the income tax expense in the year or years in which the tax effects are realized, resulting in a higher net income, higher retained earnings and higher equity. That is to say, deferred taxes are included in the asset base and increase the value of equity in the capital structure. The flow through approach produces the highest cost of capital rate among the three methodologies.

[73] The Agency's current treatment of deferred taxes was determined in the 1985 Decision, which viewed deferred taxes as a cost free source of capital for the railway companies. In this regard, the Agency concluded in the 1985 Decision:

...from a cost viewpoint, accumulated deferred taxes are in essence an interest-free loan and as such must be considered as a zero cost source of capital, since the objective is to determine a fair level of compensation....to allow the cost of equity rate on these balances, would provide excess returns to the shareholders....

[74] In recent years, either one or both of CN and CP have questioned this treatment of deferred taxes.

3.2 Positions of the participants

Canadian Pacific Railway Company

[75] CP submits that using market value weights, its recommended option, eliminates the need to include deferred taxes as a component of capital structure, as any economic effects associated with the depreciation policies that generate deferred tax liabilities will be captured in the market value of equity. CP adds, however, that even though using the market values of equity is the best method for determining the capital structure weighting, regulators sometimes use book values as an alternative proxy for market values. CP submits that, in cases where book values are used, neither operating liabilities nor its equity equivalents, such as deferred tax liabilities, should be considered part of the capital structure and should instead be included in equity.

[76] CP submits that the ability to defer taxes was created by the government to encourage investment and that this incentive should remain with the investor making the investment decision, otherwise investment incentives are strongly diminished. CP further submits that any benefits from deferred taxes accrue to equity holders and, therefore, should be part of equity value. CP states that the Agency's current treatment of deferred taxes contributes to an unreasonably low cost of capital and counteracts the investment incentives associated with the government's accelerated depreciation policy.

Canadian National Railway Company

[77] CN argues that when the market value of assets is used as the asset base, then deferred taxes are irrelevant and they are not part of the market value. CN adds that if, however, book values are used to determine capital structure and book value is to be an unbiased estimate of market value, then deferred taxes must be included in the total asset value and must be allowed a return that ensures the WACC is earned on the entire book value of assets. It considers that under some conditions this will be equivalent to the cost of equity.

[78] CN considers deferred tax to be an accounting adjustment to reconcile two different sets of books, and cites an unnamed academic reference as stating that deferred taxes should not be regarded as a source of financing or an element of the WACC formula. It maintains that the liability for deferred taxes is not a security held by investors, but a balance sheet entry created to serve the needs of accounting. CN considers the problem compounded when deferred taxes are considered a zero-cost source of financing, finding it difficult to imagine conditions under which suppliers of capital would see zero as the rate of return they could earn on an alternative investment of equal risk.

[79] CN further argues that deferred tax is not a contribution to capital, but simply an account which recognizes that actual after tax cash flows from operations are different from the after tax cash flows from operations reported in the accounting statements, if the allowed tax depreciation for tax purposes differs from depreciation used for reporting purposes, and actual taxes differ from reported taxes.

Western Canadian Shippers' Coalition

[80] WCSC states that in establishing rates for regulated business, the rates charged to customers are set so that the income tax falls on the company's customers and not on the company and its shareholders. A rise (fall) in tax raises (reduces) customer charges but leaves the return in investor capital unchanged.

[81] WCSC provides the following explanation of the various treatments of deferred taxes. When companies were allowed to adopt accelerated depreciation for tax purposes, the taxes they had to pay were reduced. Due to accelerated depreciation, what normally occurs is that companies pay less tax in the early years of the asset and higher taxes in the distant future. The reduction in taxes due to the use of accelerated depreciation is known as flow-through accounting. The alternative method called normalization involves reporting what the tax would have been if the accelerated depreciation had not been allowed for tax purposes. The difference between the statutory tax rate and the taxes the company actually paid are put into an account called deferred taxes.

[82] According to WCSC, the accounting profession certified tax normalization as the proper and correct accounting procedure as, in this way, the proper matching of expense and revenue is accomplished. WCSC states that under normalization, rates paid by customers are not reduced by the reduction in taxes paid by the railway companies (in other words, customers are paying rates associated with the statutory tax rate rather than rates associated with the effective tax rate), forcing on customers periodic loans equal to the provision for deferred taxes. WCSC submits that no interest is paid on these loans, and customers never recover the higher charges paid.

[83] WCSC further advances that customers, however, are compensated for the higher rates in the present by lower rates in the future, if deferred taxes are included in the WACC at their book value weight and assigned a zero cost rate. It states that, in effect, tax normalization with no return allowed on the provision for deferred tax provides customers as a group with the WACC on the provision for deferred taxes – the money they have loaned the company.

Canadian Canola Growers Association

[84] CCGA submits that deferred taxes should be assigned a zero cost rate as deferred taxes are temporary, given that they only exist due to a difference between two financial reporting methodologies. At some point in the future, the deferred taxes will be eliminated as the two values will eventually converge.

Coalition of Rail Shippers

[85] CRS did not comment on the issue of deferred taxes.

Provinces of Alberta and Manitoba

[86] Alberta and Manitoba express the same position on the question of deferred taxes. It is their position that deferred taxes are a non-cash expense that provides the railway companies with a source of free cash flow. The deferred taxes are a result of CN and CP depreciating their assets at a higher rate for tax purposes than for accounting purposes. CN and CP are able to defer the tax liability indefinitely as they continue to invest and upgrade their capital infrastructure.

[87] Manitoba views deferred taxes as an allocated amount during the period to cover tax liabilities that have not been paid, a non-cash expense that provides a source of free cash flow, available to the railway companies at no cost. Manitoba also submits that in actual practice, the deferred tax liability is usually postponed indefinitely because financially prudent companies continue to invest and upgrade their capital infrastructure.

[88] Alberta and Manitoba are of the view that the Agency should continue its practice of giving weight to deferred taxes at zero cost in the capital structure. They submit that the alternative, which is to subtract deferred taxes from the asset base (investment) to which the cost of capital is applied, fails the test of pragmatism. They state that this treatment is feasible only when the asset base corresponds to the entirety of the company's capital. They state that this is not the case in any of the applications of cost of capital by the Agency. For example, in assessing the railway companies' grain transportation costs, the Agency applies a cost of capital factor to the investment used to transport grain, which is only a portion of the total investment of the railway companies.

Province of Saskatchewan

[89] Saskatchewan did not comment on the treatment of deferred taxes.

3.3 Rebuttal comments

Canadian Pacific Railway Company

[90] CP reiterates its opinion that using market value weights also eliminates the need to identify deferred tax liabilities as a component of CP's capital structure, and that any economic effects associated with the depreciation policies that generate deferred tax liabilities will be captured in the market value of equity, a connection that it says is ignored in all the initial submissions of the shippers and provinces. CP suggests that the Agency's continued treatment of deferred tax liability as a separate component of the capital structure and assigning it a zero cost in the calculation of the WACC contributes to an unreasonably low cost of capital and counteracts the investment incentives associated with the government's accelerated depreciation policy.

Canadian National Railway Company

[91] CN suggests that the assertion by WCSC that deferred taxes are somehow an interest-free loan is difficult to follow, and that in its initial submission it had provided a specific model to show that this is not the case. CN cites an unnamed academic reference to support its view that deferred taxes are not a source of capital. According to CN, the standard textbook treatment makes it clear that asset value depends on after-tax free cash flows, and accounting accruals, like deferred taxes, that do not affect after-tax free cash flows are irrelevant. Furthermore, continues CN, the CAPM required rate of return applies to the market value of the assets, making deferred tax entries or their timing irrelevant. CN concludes, therefore, that it is inappropriate and irrelevant to treat deferred taxes as an interest-free loan

Western Canadian Shippers' Coalition

[92] WCSC suggests that both CN and CP are confused about the differences of deferred taxes in unregulated companies and the effect of deferred taxes under regulation. WCSC asserts that a basic principle of regulation is to allow a return on capital provided by investors sufficient to attract capital; investors other than the railway companies' customers do not provide the funds represented by deferred taxes. Therefore, in its view, any return on these funds represents an excess return on the common equity over and above the return required. Submitting that tax normalization under regulation represents compulsory loans from the railway companies' customers, WCSC concludes by saying that CN's recommendation on deferred taxes is the equivalent of being charged interest on the money you lend to, as well as the money you borrow from a bank.

3.4 Practice of other regulators

[93] STB captures deferred taxes in ROI but not in the capital structure for determination of the cost of capital rate. ROI is calculated by dividing net rail operating income by an investment base that consists of the firm's net investment in rail property, plus working capital, less accumulated deferred income tax credits. ROI is used for the determination of revenue adequacy by comparing it to the cost of capital rate. A railroad is revenue adequate if its ROI exceeds the cost of capital rate as described in the STB 2008 Decision[10]. The Decision adds, "by comparing ROI and cost of capital, the Board seeks to ensure that a rail carrier is able to continue to invest in its infrastructure and provide a reasonable return to its investors". It is STB's position that deferred income tax credits are a zero-cost source of capital and hence are subtracted from the investment base (normalization type 1).

4.0 Asset Base

Issue: An issue raised in the course of the consultations is whether the Agency should continue using an asset base (net rail investment) that is based on book value, or, instead, adopt a market value asset base.

4.1 Position of CN

[94] CN is the only party that directly advocated for the use of market values (replacement cost) for the asset base (net rail investment), as part of its proposal for the use of market values in determining the firm's capital structure. CN, using an example to illustrate the importance of market-based measures of opportunity cost, submitted that even if a market value capital structure were used, in the absence of a market value asset base, an investor's opportunity cost would not be met. CN's argument goes as follows:

Using book values can be distorting. If, for example, the Agency offers REV [expected freight revenue] based on the book value of the rolling stock and used book values for the weights in the WACC calculation, they would only allow REV of $25, not the required $59 as identified above. Clearly, in such a situation the company would sell the rolling stock for $200 and not participate in the regulated market.

If, as an intermediate step, the regulator recognized the market value of the rolling stock had increased to $200 but used the book value for the weights in WACC, they would multiply the pretax WACC of 25% by the market value of $200 to offer pretax REV of $50 but this is below the breakeven pretax REV of $59 so again the rolling stock would be sold for $200 and the firm would not participate in the regulated market.

This example illustrates the importance of using market values for the rate base and market values for the weights used in WACC calculations. The use of market value in estimating the cost of capital is both academically appropriate and widely used in practice... The Brattle Report reports widespread use of market value based weights in the calculation of WACC as well as more limited use of market value/replacement cost for the rate base.

4.2 Rebuttal comments

Western Canadian Shippers' Coalition

[95] WCSC notes CN's position that because the cost of capital should reflect opportunity cost, it is difficult to imagine conditions under which suppliers of capital would see zero as the rate of return they could earn on an alternative investment of equal risk, as well as CN's argument, based on this position, that deferred taxes are irrelevant when market values are used as the asset base.

[96] In response, WCSC asserts that, on the contrary, the argument merely shows why market value of assets should not be used as the asset base. WCSC then goes on to say that based on CN's assumption about using a market value asset base, CN is only able to provide an analysis when book values are an unbiased estimate of market values (that is, book values are approximately equal to market values). WCSC asserts that as both CN and CP have market-to-book ratios greater than 2.0, CN's analysis has no relevance.

5.0 Cost Rate of Debt

Issue: Should the Agency determine the cost rate of long-term debt by using the historic cost of debt in the railway companies' financial statements for the most recently completed fiscal year? If not, how should the Agency determine the cost of long-term debt?

5.1 Context and relevance

[97] The Consultation Document raised the possibility of calculating the cost of debt rate on the basis of the cost of debt currently available in the market or projected costs of potential future debt, as opposed to the present methodology based on actual interest payments for existing debt. The examination of a market value capital structure also brings forward the issue of market-based measurements of the yield on long-term debt.

[98] This section examines the differences between existing debt and potential future debt, and evaluates how best to measure the cost of debt.

[99] Long-term debt is one of the sources of capital used to finance a company. There are many types of debt issues, such as bonds, notes, debentures, commercial paper, revolving credit facilities, capital leases and purchase agreements, all with individual rates, maturities and terms of agreement.

[100] The Agency's current methodology for determining the cost rate of long-term debt is to calculate the interest rate from the railway companies' most recent financial statements. It allocates interest payments on debt to the responsible debt and computes the cost of debt as the weighted average of these interest payments. This method has the benefits of consistency, robustness, simplicity, transparency and reproducibility. It is retrospective, in that the railway companies are compensated based on the cost rate of debt they have already incurred, rather than on the current price of debt available in the market.

[101] The Agency reviews cost of capital estimates annually and any debt incurred in a given year is included in Agency calculations the following year. This new debt is weighted in relation to its share of total debt, which is in turn weighted by the company's capital structure. So, although it results in a lag of one year, the current approach precisely captures the existing debt obligations of a company.

[102] Due to its retrospective nature, this approach does not necessarily measure the requirements of potential investors for any new debt issued by the railway companies. For example, if a company incurred most of its debt obligations in an economic context significantly different from the present one, the cost of debt (interest payments) paid by the company for existing debt may not be representative of the cost of new debt available on the market. This raises the question of how best to reflect the cost of debt in the Agency's regulatory context, as debt obligations to existing investors differ from anticipated debt requirements of potential investors, which in turn differ from the market value of all outstanding debts.

5.2 Positions of the participants

Canadian Pacific Railway Company

[103] CP argues strongly to use market values for the cost of debt. It maintains that they are a better estimate of CP's current cost of borrowing than coupon rates. CP contends that when the market cost of debt is higher than book value costs, using book values can "severely understate" the cost of capital. For this reason, CP recommends using market values to estimate the cost of debt, stating that "the yield to maturity reflects the current market cost of debt faced by the firm."

[104] CP asserts that when market risks are elevated and the cost of debt is high, the use of book value can severely undervalue the cost of capital. It adds that the coupon rate approach used by the Agency does not take capital appreciation or depreciation into account. CP recommends, instead, calculating the yield-to-maturity on the average yield for an index of similarly rated corporate bonds with a weighted average time to maturity consistent with CP's existing debt instruments. CP included with its submission a market-based cost of debt, "calculated as average of month-end yields to maturity during the first quarter of each calendar year from the 10- and 15-year Bloomberg Fair Value Curve for BBB Canadian Corporate Bonds."

Canadian National Railway Company

[105] CN provides an extensive example in support of its view that market values should be used to estimate the cost of debt. The essence of CN's argument is similar to CP's, that in the context of rising interest rates, the historical cost of debt will not suffice to raise funds in the market. CN also maintains that mixing book and market values will "cause distortions".

[106] CN recommends therefore that the cost of debt should be based on current yields extracted from market values of debt, either from recent trades in CN debt issues or from issues that are comparable in terms of default risk and duration.

[107] CN acknowledges that the market for railway companies' corporate debt may be too illiquid to provide meaningful estimates, and suggests development of an estimation model:

Although the market values of some debt issues might be available from recent trades, corporate debt trades much less frequently than equity. As a result, debt market values must be estimated using a "model," [sic] that could consist simply of an average of a set of equivalent rating and duration comparables for which market values are available.

Canadian Canola Growers Association

[108] CCGA views the use of the historic (i.e., book value) cost of debt published in the railway companies' annual reports as reasonable. CCGA advances that a company is more likely to issue debt in a low-interest environment, and that projecting debt costs in volatile markets would be difficult. CCGA also indicates that it may not be clear how much debt companies may issue in a particular year, or the duration or even currency relating to the debt. CCGA concludes by pointing out that annual cost of capital calculations should minimize differences between historic and current costs of long-term debt.

Western Canadian Shippers' Coalition

[109] WCSC recommends that the cost of debt be determined by the actual embedded interest cost in the railway companies' financial statements, a method it indicates is equivalent to the coupon rate. WCSC supports its viewpoint by indicating that the contractual obligation to pay a stated amount of interest and repay the principal at maturity is not affected by market yields. WCSC argues that if bond yields are calculated using market rates, there would be a windfall profit captured by the companies' shareholders:

...when market yields rise above the embedded cost, no additional payment is made to the bondholders because of the contractual agreement on interest payments, so the difference in rates is captured by the common shareholders.

[110] WCSC relates in its proposal to use the railway companies' actual cost of debt with using book values to determine the capital structure weights:

[In] using the embedded interest cost with book value weights in calculating the cost of capital, common shareholders are allowed to earn the market determined cost of common equity capital without receiving windfall gains or losses when interest rates change.

Coalition of Rail Shippers

[111] CRS recommends determining the cost of long-term debt "based on the historic cost of debt in the railway companies' financial statements." CRS made this statement in the context of using book values (rather than market values) as the appropriate measure for the capital structure. CRS suggests that if debt costs were to be forecasted instead, any forecasting errors could be captured in a deferral account set up for that purpose.

[112] CRS agrees with WCSC that switching to current market yield on debt instead of the actual debt interest rate would create windfall gains for shareholders. This is because when forecasted yields rise, no additional payments are made by the company to the debt holders. Therefore, any increase in the Agency's cost of debt measure will be captured instead by equity holders.

Province of Alberta

[113] Alberta submits that the yield on long-term debt should be calculated using the yield-to-maturity method.

Province of Manitoba

[114] Manitoba suggests that the Agency measure the cost of debt using the most-recent available "embedded cost of debt." Manitoba adds that this cost should be estimated using the yield-to-maturity method, which "factors in the costs of debt issuance and any discount or premium that the railway may have received when it initially sold the bond issues" but not any "current discounts or premiums on the outstanding debt."

[115] Manitoba argues strongly against the use of forecasted interest rates to represent the cost of debt, indicating that these forecasts may not be relevant for the railway companies' actual debt expenses, and will be less transparent and will likely vary among analysts and over time.

Province of Saskatchewan

[116] Saskatchewan did not provide explicit commentary on the appropriate measurement of the cost of debt.

5.3 Rebuttal comments

Canadian Pacific Railway Company

[117] CP notes WCSC's submission that using the current market yields on debt instead of the historic cost of debt in the railway company's financial statements would create windfall gains or losses to shareholders, and suggests that this concern is based on a conceptual misunderstanding of the cost of capital. CP explains that although a railway company may be contractually obligated to make periodic interest payments on some of its debt instruments, the economic cost of those debt instruments at a given point in time is determined by economic conditions in debt markets, which are not reflected in the historic cost of debt shown in the railway company's financial statements. CP states that when market interest rates change, the prices of outstanding debt instruments change such that the market yield to maturity reflects the return that investors require to bear the risks associated with a given debt instrument.

Canadian National Railway Company

[118] CN argues that its initial submission shows how using the embedded cost of debt could distort the weighted average cost of capital, and also maintains that the embedded cost of debt does not reflect the true cost that debt holders would demand to supply capital for new investments. According to CN, the WCSC submission fails to provide any analysis of how to avoid the distortion of the weighted average cost of capital that would arise from the use of the embedded cost of debt, and also avoids any discussion of the inappropriateness of using the embedded cost of debt for the financing of new investments.

Western Canadian Shippers' Coalition

[119] WCSC suggests that CP's recommendation that the Agency use the yield to maturity of an index of Canadian bonds with the same time to maturity and debt rating as the railway company for which the debt cost is being determined is complex and results in costs of debt for CP which do not seem plausible.

5.4 Practice of other regulators

[120] Other regulators that use market forms of debt include STB and the Canadian Radio-television and Telecommunications Commission (CRTC). The practice of STB was discussed in more detail in the earlier sections on capital structure, but is briefly summarized here for ease of reference.

[121] STB calculates the cost of debt based on the existing market yields. For this purpose, STB's methodology applies the current yield formula on those bonds in the composite sample considered by STB that have been publicly traded (approximately 50 percent in 2010). For those securities that are not traded, the methodology estimates the yield based on the spread between the debt issue and Treasury securities issue contemporaneously. There are also some long-term debt securities that are included at their par value (i.e., employing the coupon rate). The final yield on long-term debt is a weighted average of the estimated yield on bonds and the yield on non-traded securities plus an estimate of the floatation costs.

[122] The CRTC uses the "current cost of debt financing" to measure the cost of debt.[11] This market cost of debt is estimated as the yield on 10-year government bonds plus a company specific risk premium and issuance costs. The company-specific risk premium is estimated as an average of estimates obtained from investment banks. The cost of debt is adjusted (reduced) in relation to equity to account for the corporate tax "shield" related to debt interest. The cost of debt is applied to the company's current capital structure

5.5 Brattle Report

[123] The Brattle Report did not explicitly discuss methods for calculating the yield on long-term debt. The response of the Brattle Group partially addresses the issue by pointing out that the yield-to-maturity is preferred over the current yield as it takes all payouts from the bond issue into account.[12]

6.0 Cost Rate of Common Equity

Issue: What method or combination of methods should the Agency use to estimate the cost rate of common equity?

6.1 Context and relevance

[124] Equity is one of three components of a firm's capital structure along with debt and deferred income taxes. These three components represent the three sources of financing available to the firm to fund its capital investments. Each component carries its own cost of financing. The weighted average of these costs (with weights determined by each component's proportion of the capital structure) make up the firm's weighted average cost of capital. As equity makes up a large portion of the railway companies' capital structures, the method used to determine the cost of equity can have a significant impact on the overall weighted average cost of capital. The Agency currently considers three market driven models in estimating the cost of equity.

Capital Asset Pricing Model (CAPM)

[125] The CAPM measures the cost of common equity for an investment as the sum of the risk-free rate of return available in the economy plus an additional return that is a function of the return available in the market and the systemic risk in the equity relative to the overall market. The CAPM is shown mathematically as:

Re = Rf + ß (MRP)

where:

Re is the expected or required cost rate of equity;

Rf is the risk-free rate;

MRP is the average returns from the market minus the average returns from the risk-free asset over a specified period of time

ß is the systemic or non-diversifiable risk in the equity relative to the market.

[126] This relationship, generally referred to as the traditional or unconditional CAPM, is the result of a number of simplifying assumptions regarding investor behaviour and the working of capital markets. In particular, the traditional CAPM assumes that all the non-diversifiable risk associated with the equity is reflected in the single term ß. Under certain circumstances, additional terms may be added to the traditional CAPM equation to reflect other risks not captured by the ß, including currency exchange rate risks, inflation risks, future uncertainties, etc. These extensions of the traditional CAPM model are reviewed in greater detail in section 8.

[127] The Brattle Report summarizes the appropriateness of the CAPM as a cost of equity methodology as follows (pages 47-48):

The CAPM has a strong theoretical foundation and fits with the intuition of a risk-return tradeoff. The data necessary for its implementation are widely available at low cost. And its calculations are relatively simple. In the model, the risk-free interest rate reflects current market conditions, but the estimated beta relies on historical data, so the model is neither forward-looking nor completely backward-looking. The model is robust to violations of its underlying assumptions albeit not necessarily to changes in economic conditions.

As was revealed in the above discussion, the primary source of debate for the CAPM is estimating parameters, particularly the MRP, but the appropriate method to estimate beta and what is the appropriate measure of the risk-free interest rate are often controversial as well. It is important to recognize this lack of consensus in the academic literature and among practitioners when employing the CAPM in a regulatory setting. Although it is perhaps more well-rooted in economic theory than other methodologies, it is also more subject to technical debate and disagreement.

Equity Risk Premium (ERP)

[128] There are various ERP methods used for estimating the cost of common equity, all of which are based on the premise that an investment in equity carries greater risk than an investment in debt and therefore requires a premium over that required for bonds. The ERP method used by the Agency is shown mathematically as:

RE = Rf + MRP

where:

RE is the cost rate of equity;

Rf is the risk -free rate;and

MRP is the market risk premium.

[129] As can be seen from the above equation, the Agency's ERP model is a special subset of the traditional CAPM, where the company-specific systemic risk, ß, is assumed to be the same as that for the market as a whole. The Agency currently takes the average of the market risk premiums for the five most recent years to arrive at the market risk premium variable for this calculation.

[130] The Brattle Report summarizes the appropriateness of the ERP as a cost of equity methodology as follows (pages 56-57):

The risk premium model is a derivative of the CAPM so the comments that apply to the CAPM also apply to the Risk Premium Model; however, the Risk Premium Model does not have the same level of theoretical support. The tie between theory and implementation is weakened because the interest rate in the Risk Premium Model is not necessarily equal to the risk-free rate and the risk premium is not explicitly based upon the product of the investment's beta and the MRP. However, the calculations are simple and the model is based upon the risk-return trade-off underlying the CAPM. The model is forward looking because the benchmark interest rate is a current rate, and the data necessary for the model is generally widely and cheaply available, depending upon how the risk premium is estimated.

Discounted Cash Flow Model

[131] The DCF Model is based on the premise that the price of an equity is the present value of all the future cash flows accruing from the equity over the lifetime of the equity. If it is assumed that dividends are the only cash flows accruing from the equity, and it is also assumed that the dividend grows at a constant rate to the end of time, then the standard DCF formula estimates the expected return on an equity as a relationship between the current dividend yield, the current price of a company's common shares and the expected future dividend growth rate. This simple single-stage DCF Model is shown mathematically as:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-131.gif" width="194" height="47" />
Text alternative for the equation

where:

RE is the rate of return on equity;

D0 is the current dividend;

P is the current price of the stock;

D0/P is the current dividend yield;and

g is the dividend growth rate.

[132] If the assumption that dividends grow at a constant rate to infinity is thought unrealistic, and various estimates of the dividend growth rate over time are available, then the expected return for the equity can be estimated as a solution for Re in the equation:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-132.gif" width="590" height="74" />
Text alternative for the equation

where:

D1 ... DT represents the dividend cash flows in the various periods;

T represents the long run period when growth levels off; and

g represents the equilibrium growth rate in the long term.

[133] The implementation of this multi-stage DCF Model requires forecasts of the growth rates for each stage of the model along with the length of each stage, in addition to the current dividend and price, and is discussed in greater detail in section 13. The Agency currently uses the single stage version of the model.

[134] The Brattle Report summarizes the appropriateness of the DCF as a cost of equity methodology as follows (pages 52-53):

In summary, the reliability of the DCF model hinges on the appropriateness of its assumptions — whether the basic present value formula works for stocks, whether option pricing effects are important for the company, whether the right variant of the basic formula has been found, and whether the true growth rate expectations have been identified.

Most of the data necessary for its implementation is widely available at low cost. The exception is that there is no source of data on the long-term growth rate of dividends. Its calculations are relatively simple, and the logic of the model is intuitive in that the expected return on an investment is equal to the expected amount of current income (i.e., the dividend payment) and the expected amount of capital gain (i.e., the growth in the price or dividend payments).

The major source of debate for the DCF model is determining the dividend growth rate, particularly for the long-term. There is generally no publicly available data on forecast growth rates for periods longer than 5 years. Unfortunately, the forecast growth rate has a major effect on the cost of equity estimated by the DCF method.

The DCF approach is conceptually sound if its assumptions are met, but can run into difficulty in practice because those assumptions are so strong, and hence unlikely to correspond to reality.

6.2 Evolution of the Agency's current cost of equity approach

[135] The cost of equity methodology currently used by the Agency was determined in the 1985 Decision:

Three market-based methods were put forward in this hearing to measure CP Ltd.'s cost of common equity: the discounted cash flow method (DCF), the capital asset pricing model (CAPM) and the risk premium approach. The Committee finds that these three methods have some theoretical merit, but that all three are very difficult to implement. In future annual determinations of the cost of capital, the Committee will have regard to the outcomes of all three methods, tempered by informed judgement. In addition, if new analytic tools are developed in the future to a point where they are reliable, the Committee will give consideration to their use in addition to the three mentioned above.

[136] Since the 1985 Decision, the Agency has revisited its cost of capital methodology twice. In the 1997 Decision, after re-examining its cost of common equity methodology, the Agency concluded:

....an assessment of the three market-based methods (i.e. the DCF, the CAPM and the Equity Risk Premium) will continue to be performed to estimate CPL's cost of common equity. As in the 1985 decision, the outcome of all three models will continue to be assessed annually and weight given to the most appropriate model or a combination of models. The Agency notes that there is judgement involved in estimating the results of each model.

[137] In light of regulatory changes, the Agency reviewed its cost of equity methodology again in 2003 and 2004, at which time, responding to concerns expressed by CN about the zero weight being assigned by the Agency to the DCF method and other matters, it concluded:

...no change is required in the methodology ascribed in the 1985 and 1997 Decisions regarding which market driven models to use and how they are to be assessed by the Agency in cost of equity appraisals. Thus, the Agency will continue to assess all three recognized models and give the appropriate weight to each model or combination of models that best reflect the state of the capital markets.

[138] The Agency has continued to evaluate a railway company's cost of equity for regulatory purposes, using the CAPM and the ERP and DCF Models. After calculating the cost of equity using each model, the Agency determines which model or combination of models is most suitable given the current economic environment. Since the 1992-1993 crop year, the Agency has given 100 percent weight to the CAPM, finding that the CAPM produced results that better reflect the capital markets within which the railway companies operate. In the 2004 Decision, it was noted that the Agency has given preference to the CAPM in the past due to the fact that the:

CAPM is widely known and accepted in regulatory and financial practice and it is not subject to the degree of conjecture required to estimate an expected growth rate or risk premium as is the case with the DCF and EPR methods respectively. CAPM allows for a transparent quantifiable projection, that incorporates reconcilable data from the market as a whole, as well as a company specific factor (beta), to arrive at a cost of common equity forecast.

[139] Since the 2004 Decision, either one or both of the Class 1 railway companies have continued to express concerns and have advocated that the Agency use a combination of the results of the CAPM and the DCF to determine cost of equity.

6.3 Positions of the participants

Canadian Pacific Railway Company

[140] CP notes that the Brattle Report states that "looking at the evidence from a number of models continues to be best practice, because different models may be better at capturing different aspects of pricing." In respect to this statement, CP proposes that the Agency rely on an average of the cost of equity estimates from the CAPM and the multi-stage discounted cash flow (MSDCF) Model.

[141] With reference to the appropriate CAPM, CP indicates that the Agency's current model has some major shortcomings in that: (i) the risk-free rate should be the long horizon risk-free rate of return; (ii) it does not rely on all of the available data (which go back to 1936) to determine the market risk premium; and, (iii) the risk-free rate of return used in the calculation of the market risk premium is measured using total returns, as opposed to income returns. CP concludes that if the Agency were to adopt the combined methodology it proposes, it would produce a more stable estimate of CP's cost of equity than if it were to rely on the CAPM methodology alone.

Canadian National Railway Company

[142] CN is of the opinion that the CAPM methodology best meets all of the Agency's criteria of being reasonable, reliable and pragmatic. It does submit, however, that the DCF methodology, while based on different inputs than the CAPM, is conceptually sound and should be used to provide a consistency check on the estimates that result from the CAPM methodology. CN also states that it would be comfortable if the Agency chose to use a simple average of the two methodologies.

[143] With respect to the ERP approach, CN considers that it might have merit when market prices for a regulated company are not available, but because CN is an actively traded public company, application of the CAPM is superior.

[144] As for the DCF approach, CN submits that consensus earnings growth rates are obviously not sustainable in perpetuity, and the direct use of even the five year growth rate will lead to an overstated cost of equity. CN further submits that the DCF method presents challenges but concludes that using the DCF as a consistency check on the CAPM could be helpful in that a large divergence between the two estimates might indicate a problem with the CAPM.

Canadian Canola Growers Association

[145] CCGA supports the use of the CAPM. CCGA notes that when determining the cost of equity, the Agency reviews the results from the DCF and ERP methodologies as well. CCGA also expresses the opinion that reviewing these results provides reassurance that the CAPM results are reasonable and provide an additional means for the Agency to detect potential issues with either the CAPM calculation methodology or the inputs used in the calculations. CCGA submits that the Agency's current calculation process is robust. CCGA does not discuss the DCF Model and its variations, or submit any opinion on inputs to the DCF Model.

Coalition of Rail Shippers

[146] CRS's submission, which is in the form of a critique of the Brattle Report, comments on several cost of equity models discussed in that report.

[147] CRS states that adding cost of equity estimates derived from inferior estimation methods to those derived from superior estimation methods, as it submits is advocated in the Brattle Report, only increases estimation error and potential bias and does little to produce a more accurate (fair) estimate of the cost of equity. With respect to the Agency's current methodology of reviewing three models and using its judgement to select the best model, CRS submits that a convincing case has not been made in the Brattle Report for why the Agency should not retain this flexibility.

[148] CRS also submits that the ERP models [ERP models include, amongst other models, the CAPM] are the models of choice for regulators considered in the Brattle Report. It submits that the DCF models are a distant second, and the Comparable Earnings Model is at the end of the choice line with only one regulator partially using it.

[149] As for the DCF Model, CRS submits that economic literature documents that the earnings estimates of analysts exhibit substantial optimism and overconfidence biases, that revisions in analysts' forecasts cause variability in stock prices, and that the use of these upwardly biased estimates without removing the bias leads to an upwardly biased estimate of the required ERP and cost of equity. CRS further submits that if regulators set allowed cost of equity rates based on the DCF Model using expected growth rates that are consistently higher (or lower) than what is subsequently achieved, then the allowed cost of equity will consistently be higher (or lower) than the fair cost of equity.

Western Canadian Shippers' Coalition

[150] WCSC submits that the Agency should continue to estimate the cost rate of common equity using the CAPM without giving weight to the DCF Model. WCSC questions the accuracy of the estimated growth rates used in the DCF and discusses three possible ways in which the growth rates can be estimated. WCSC does not, however, recommend any of the three methods.

[151] The first method discussed by WCSC is to take an average of the return on common equity times the fraction of earnings retained by the company over a pre-specified period in the past. WCSC rejects this method because it considers it subject to considerable error. The second method is for an expert to take all of the historical information that is considered relevant and arrive at an estimate; however, WCSC submits that different judgements by different persons would result in different numbers, and the Agency would be compelled to look at the reasoning behind each judgement in deciding which number to accept. The third approach WCSC discusses is to take averages of the estimates made by security analysts who are employed by firms that provide investment advice. WCSC points out that estimates such as this are biased to the extent that the return on investment and the fraction of earnings retained by the company are expected to change over the period.

[152] WCSC expresses the view that the DCF method for estimating the cost of equity capital can be very accurate when used by an expert who takes all of the relevant historical and current information into account, but it is less accurate when used as a rigid formula. WCSC concludes that without annual hearings and the opportunity to examine the reasoning behind the estimates, the DCF method would be unlikely to produce accurate results.

Province of Alberta

[153] Alberta submits that the Agency should maintain the CAPM to determine the common equity return portion of the cost of capital. It believes that, based on the Brattle Report, a clearly superior model does not exist and that no improvements could be made to the CAPM.

[154] Alberta does not discuss the DCF Model and its variations, or submit any opinion on inputs to the DCF Model, other than to submit that using the DCF would adversely impact western grain shippers and unreasonably enhance the earnings of CN and CP from the hauling of western grain.

Province of Manitoba

[155] Manitoba reviewed the CAPM, DCF, ERP and Comparable Earnings approaches. Based on its review, Manitoba submits that neither the ERP nor the Comparable Earnings approaches should be used to calculate the return on equity, but that it considers the CAPM reasonable, reliable and pragmatic. Manitoba considers that the DCF method has problems of reliability and pragmatism, as well as reasonableness, submitting that the current forecasted growth rates are unsustainable in the long run.

Province of Saskatchewan

[156] Saskatchewan sees no reason for the Agency to change from the use of the CAPM as its basis for finding the cost of equity. It concludes that no "other methodology has met the high standard of being "clearly superior" to the C[A]PM methodology currently being used by the Agency, and as a result the Agency should continue the use of the C[A]PM methodology."

[157] Saskatchewan also performed an analysis of the impact of using the DCF to calculate the cost of equity used in the Western Grain Maximum Revenue Entitlement Program. Saskatchewan submits that had the Agency used the DCF methodology for calculating the cost of capital in the revenue cap in the 2009/2010 crop year, it would have resulted in an increase of approximately $88 million to the revenue cap.

6.4 Rebuttal comments

Canadian Pacific Railway Company

[158] CP submits that CRS raises a number of theoretical issues about the estimation of the cost of equity, but does not put forward a specific alternative to the Agency's current methodology, and appears to conclude that the Agency should retain the flexibility to use multiple methods for estimating the cost of equity. CP also submits that, at the same time, CRS suggests it is possible to identify a single method for estimating the cost of equity that is more precise and, therefore, superior. CP disagrees with CRS's assessment of these issues. CP explains that the cost of equity is an implicit cost that cannot be directly observed, and there is no way to measure the cost without relying on economic and financial models, each of which have positive and negative attributes that must be considered before relying on their results. CP suggests that no single model is always considered to be best, and that a widely accepted method for increasing the reliability of results generated from these models is to combine individual estimates from multiple models into a single cost of equity estimate.

[159] CP notes CCGA's concern about potential year-to-year fluctuations that may occur if the methodology used to calculate the cost of capital was frequently altered. CP agrees with CCGA's concern about potential fluctuations in the regulated cost of capital rate, but disagrees that the consistent use of a single methodology is beneficial over the longer term in reducing these fluctuations. CP is of the opinion that CCGA's conclusion appears to be based on the premise that the use of multiple methods would involve the switching of methods over time, not the consistent use of multiple methods in a systematic way. CP suggests that its proposal to average the results of the two models (the CAPM and the multi-stage DCF) should alleviate the concerns raised by CCGA. CP asserts that its proposed methodology will allow the Agency to obtain a cost of equity estimate that is more likely to reflect the fundamental economic conditions of the Canadian railway industry, and the estimate is less likely to be unduly affected by specific judgements that are made in implementing the models.

Canadian National Railway Company

[160] CN submits that the WCSC submission provides no support or justification for its affirmation that the Agency should continue to estimate the cost rate of common equity using the CAPM. In addition, CN disagrees with WCSC's rejection of the DCF approach on the basis that model inputs must be selected using either historical estimates that are subject to error, or forecasts that rely on judgement. CN submits that as no statistical estimate can be error-free it does not consider the first argument to be a valid reason to reject a method, and as the application of any model requires judgement, including the CAPM approach advocated by WCSC, the second argument also fails to provide a reasonable basis for the submission.

Western Canadian Shippers' Coalition

[161] WCSC states that it has no theoretical objection to using a multi-stage DCF Model to estimate the cost of equity capital, but it does have objections to the specific multi-stage DCF Model that CP proposes, the source of the estimates, and the way in which the model will be used. WCSC agrees that a DCF Model could be used in some regulatory proceedings to obtain a very accurate estimate of the cost of equity capital, but submits that the multi-stage DCF Model recommended by CP has severe deficiencies. First, says WCSC, the model itself has not been designed to estimate the cost of equity capital for either CN or CP, and the one-size-fits-all model will not produce accurate results for CN or CP. Second, continues WCSC, the procedure used by CP for estimating the inputs for the model, particularly the growth rates, is flawed. And finally, WCSC asserts that multi-stage DCF Models in general are unlikely to provide accurate results when used in a rigid formula, as employed by CP.

6.5 Practice of other regulators

[162] The issue of cost of equity methodology was discussed in the STB's most recent review. During its annual determination of the cost of capital rate for regulated railroads for the year 2005, the single stage DCF Model was challenged by the Western Coal Traffic League (WCTL). STB began a review of the methodology it uses to calculate its cost of equity in September of 2006.

As a result of that examination, in January 2008, we replaced our single-stage DCF model with a Capital Asset Pricing Model (CAPM) to determine with better accuracy the railroad industry's cost of capital". In this decision, the STB stated that it would also "initiate a separate proceeding aimed at gathering additional information on the cost-of-capital-estimating methodology . . . [that would] be focused on detailed multi-stage DCF proposals that could be used in conjunction with CAPM in the future.[13]

[163] After this review, STB came to the conclusion

....that using a simple average of CAPM and the Morningstar/Ibbotson multi-stage DCF model to calculate the cost of equity will yield a more precise determination than relying on CAPM alone.[14]

[164] STB offered the following rationale for its decision to average the two models:

We saw merit with using both models to estimate the cost of equity. As we observed, "[w]hile CAPM is a widely accepted tool for estimating the cost of equity, it has certain strengths and weaknesses, and it may be complemented by a DCF model. In theory, both approaches seek to estimate the true cost of equity for a firm, and if applied correctly should produce the same expected result. The two approaches simply take different paths towards the same objective. Therefore, by taking an average of the results from the two approaches, we might be able to obtain a more reliable, less volatile, and ultimately superior estimate than by relying on either model standing alone."

6.6 Brattle Report

[165] The Brattle Report[15] summarizes the use of different models in calculating the cost of equity by the Alberta Utilities Commission (AUC), Ontario Energy Board (OEB), Australian Energy Regulator (AER), STB, CRTC, United Kingdom Competition Commission (UKCC), Western Australia Economic Regulation Authority (ERA), and the New Zealand Commerce Commission (NZCC). All of the regulatory bodies studied in the report at least partly use the CAPM to arrive at a cost of equity estimate. OEB also uses the risk premium approach and partly looks at the comparable earnings approach.

[166] The Brattle Report also notes that while neither AUC nor OEB made detailed comments on the implementation of the DCF Models, both regulators indicated that they relied in part on results from the DCF Models to determine the allowed ROE or risk premium. The Brattle Report notes that both regulators considered both single-stage and multi-stage DCF Models; however, their decisions did not indicate a preference for the inputs into the models.

[167] The equity risk premium model employed by OEB adjusts a base ROE (based on various cost of equity models) using three quarters of the difference between the current Long Canada Bond forecast for the year and the Long Canada Bond forecast from the base year.

7.0 Grain Risk Premium

Issue: Should the Agency continue to make an annual assessment of whether the cost of common equity should be adjusted to reflect the risk of carrying grain, and, if so, on what basis should it be established?

7.1 Context and relevance

[168] The Agency has annually considered the application of a risk adjustment to the cost of equity rate for the movement of grain since 1983, when it was introduced in the Western Grain Transportation Act (WGTA). Paragraph 38(2)(b) of the WGTA stated, at that time:

…the Commission shall…compute the costs of capital in accordance with paragraph 276(3)(b) of the Railway Act and adjust that cost by any amount it deems justified in light of the risks associated with the movement of grain.

[169] In applying this adjustment, the Agency has reflected the differential risk associated with the movement of grain over the movement of other commodities.

[170] In the 1985 Decision, the Railway Transport Committee of the Canadian Transport Commission (RTC, predecessor of the Agency) found that grain traffic, for the purposes of the WGTA, was less risky than freight traffic in general. The RTC found that the circumstances under which grain was carried may have been significantly affected by the passage of the WGTA, specifically that the revenues guaranteed by way of government subsidies made the movement of grain less risky than the movement of other commodities[16]. As the allowable rates to grain shippers under the WGTA were, unlike any other commodity, fixed but insufficient, the subsidies guaranteed the railway companies sufficient revenue to cover their total economic costs from moving grain, including their cost of capital and depreciation. This led the RTC to conclude that the risk and consequently the cost of common equity rate for the movement of grain was one percentage point below that determined for CP's operations as a whole. CN was still a government-owned entity at the time.

[171] Following the repeal of the WGTA in 1995 and the enactment of the CTA in 1996, the 1985 Decision was reviewed and subsequently amended in the 1997 Decision. Where previously there had been significant subsidies provided to CN and CP by the federal government, with the introduction of the CTA subsidies were eliminated, rate structures were revised and grain producers were now required to pay the entire cost of moving grain from their farms to a terminal point.

[172] Taking this legislative change into account, the Agency found in the 1997 Decision that, although the carriage of grain is subject to varying volumes resulting from the circumstances of the weather in Canada, the carriage of grain was no less risky than the carriage of other commodities. As a result, the Agency raised the risk adjustment for the movement of grain from -1 percent to zero.

[173] At the same time the Agency also acknowledged that with the repeal of the WGTA there was the potential for additional risk for the railway companies. The Agency noted that there had been a number of attendant changes to the way the business of growing, handling and transporting grain was conducted. It found that producers had been given a greater incentive to select the lowest cost combination of transportation by truck and rail or truck to a local processor.

[174] The Agency acknowledged that there was a significant risk that quantities of grain that customarily moved by rail would not do so in the future. Grain might be trucked to a competitive railway company, or directly to market, or to a local facility for value-added processing, such as crushing of oilseeds or livestock production using feed grains. The Agency also concluded that there was a risk that the total quantity of grain produced might decline as producers withdrew marginal land from grain production, as their products might no longer support the unsubsidized costs of transportation to market.

[175] Further, the Agency acknowledged that grain producers might themselves diversify into other lines of business, such as livestock production, thereby reducing the quantity of grain moved by rail. The Agency noted evidence that some of the value-added production was moving by truck and that other products would be subject to truck competition in the future.

[176] In the 1997 Decision, the Agency also recognized that the announced, but not as yet completed, sale of the government hopper-car fleet would add a new element of risk to the business of transporting grain. The railway companies were facing the uncertainty of who would own and control the operation of these cars, which constituted approximately fifty percent of the available grain car fleet. Potential car allocation problems and conditions of use for cars were also considered elements that contributed to the uncertainty.

[177] For these reasons, the 1997 Decision left the door open for possible adjustments as required, concluding that the Agency would continue to monitor the situation and, based on its informed judgement, make a determination on an annual basis about whether an adjustment to reflect the risk of carrying grain was required.

[178] In all subsequent years, the Agency has heard arguments from the railway companies and others either for or against a grain risk adjustment, but has found in its cost of capital determinations that the reasons given in the 1997 Decision for a zero percent grain risk adjustment are still valid, and has maintained that level.

7.2 Positions of the participants

Canadian Pacific Railway Company

[179] CP comments that many agronomic and geopolitical factors play a role in the volatility of grain volumes produced and shipped from year to year, but acknowledges that much of the uncertainty and risk identified in the 1997 Decision is no longer applicable today. It submits that the risk now faced is a broader public policy risk that if railway companies cannot earn a sufficient return on capital, the sustainability of their networks and their ability to serve their customers would be compromised. CP considers that this is a risk not specific to grain, but rather to the broader supply chain. CP maintains that "the inclusion of a grain risk adjustment would be contradictory to the Agency's objective to have a methodology that is reasonable, reliable, and pragmatic."

[180] In its response to initial submissions, CP refutes the positions of Alberta and Manitoba which favour a 100 to 180 basis point grain risk discount to the cost of equity, stating that the Agency acknowledged that the repeal of the WGTA added to the risk faced by the railway companies in the transportation of western grain. CP also submits that the risks and uncertainties to which grain is subject impact not just the grain community, but also translate into risks for CP due to its investment in equipment, rail cars and facilities to support the production and distribution of grain.

Canadian National Railway Company

[181] CN submits that there have been no changes in the situation that led to the Agency's 1997 decision on this matter, and that any suggestion that the risk is different would be a totally subjective and discretionary assessment, and, therefore, inconsistent with the principles established by the Agency that the cost of capital methodology be reasonable, reliable and pragmatic. Responding to those submissions calling for an adjustment for lower risk associated with the transportation of grain, CN submits that no evidence was provided to support this position, and that it is impossible to measure the systemic risk of the non-traded investment in carrying grain or the covariance between the return for carrying grain and the return on all assets.

Western Canadian Shippers' Coalition

[182] WCSC defines the systemic risk of hauling grain as the covariance between the return for carrying grain and the return on all assets. WCSC submits that, while the systemic risk of hauling grain may or may not differ from the railway companies' overall cost of capital, if that systemic risk differs from the overall systemic risk, it is practically impossible to measure, because market prices do not exist for the non-traded investment in carrying grain.

Canadian Canola Growers Association

[183] CCGA considers the argument for a grain risk to be based on the railway companies' grain revenues being more volatile than their overall revenues measured on an enterprise basis, and submits that a diversified traffic base will produce an overall revenue stream that is less volatile than any individual traffic segment, because the segments are likely to be uncorrelated to one another. CCGA submits that it is not warranted to include a risk premium for grain, as the cost of capital is determined on an enterprise basis.

Provinces of Alberta and Manitoba

[184] Alberta and Manitoba submit that the Agency should continue its practice of making an annual assessment of grain risk, and a determination of whether the cost of common equity should be adjusted to reflect the risk of carrying grain. Should the Agency consider an alternative option, Alberta and Manitoba submits that the Agency should consider that the carriage of grain is a risk reducing factor for the railway companies, for two reasons.

[185] First, because the railway companies have a virtual monopoly in the transportation of grain over long distances, which reduces the risk for transporting grain relative to other commodities. Manitoba illustrates the risk reduction attributed to the monopoly effect with studies comparing the cost of equity of regulated and unregulated companies with comparable safety and financial strength ratings. Manitoba submits that the consistently lower cost of equity rates exhibited by the regulated entities demonstrate that investors regard monopoly companies as significantly less risky than competitive companies, and the same risk differential applies to monopoly versus competitive segments of the railway companies' transportation market.

[186] Second, unlike other commodities transported by rail, the volume of grain is not tied to the general economy. Manitoba supports this reasoning using the 2009 recession as an example in which most rail traffic experienced significant declines in volume due to the recession, while grain volumes increased (due to good growing conditions and healthy beginning inventories of grain stock), working to lessen the overall drop in freight revenues for the Class 1 railway companies. It also cites the 1985 Decision in which the RTC ruled that grain traffic reduced the overall risk for railway companies because grain revenues were not correlated with business cycles in the economy as were revenues for non-grain traffic.

[187] Consequently, Alberta and Manitoba submit that there should be a downward adjustment in the rate of return on equity of investment associated with grain transportation. Alberta submits that the reduction should be 100 basis points. Manitoba considers that a reduction should be in the range of 100 to 180 basis points to recognize the risk reducing characteristics of this traffic.

Province of Saskatchewan

[188] Saskatchewan did not state a position on grain risk adjustment.

8.0 Capm Version

Issue: In the course of the consultation certain participants proposed alternative versions to the traditional CAPM for consideration.

8.1 Context and relevance

[189] The CAPM determines the required rate of return on an equity in a well-diversified portfolio as the return on a theoretically risk-free asset plus a return commensurate with the non-diversifiable risk of the asset relative to the market as a whole. This is usually represented by the equation:

Re = Rf + ß(MRP)

where:

Re is the required return on the equity;

Rf is the risk-free rate of return in the economy;

MRP is the average returns from the market minus the average returns from the risk-free asset over a specified period of time; and

ß is a measure of the risk in the equity relative to the market as a whole.

[190] This basic relationship, known as the traditional or unconditional CAPM, is intuitively appealing, and suggests that investors will accept higher returns to compensate for higher risk in an asset. The market risk premium reflects the premium that the equity market, as a whole, provides over the risk-free rate; the ß denotes the non-diversifiable risk associated with a specific equity, with an increasing ß denoting increasing risk and resulting in higher expected returns. However, the simplicity and intuitive appeal of the traditional model is based on an idealized and simplified version of financial markets, reflected in a number of technical assumptions, which are not always found in real life. These lead to technical weaknesses in the traditional CAPM to correct for which various alternative specifications of the CAPM are prescribed, which in turn result in more complex and more general formulations that are more difficult to test empirically and to apply. Some problematic assumptions and their effects are discussed below.

Static time periods

[191] The traditional CAPM does not allow an opportunity to consume and rebalance portfolios repeatedly over time in response to changes in interest rates and other factors. If the assumption that the interest rate (and hence the risk-free return) remains constant over time is dropped, a version of the model called an Inter-temporal CAPM can be developed, where investors hedge against potential shortfalls in consumption or changes in the set of investment opportunities available in the future. The projected risk factors over time for which investors may hedge would include changes in inflation, employment opportunities, future stock market returns, etc. The generalized CAPM is therefore formulated to include a core investment tied to the market portfolio, along with other portfolios that mitigate against currently-perceived risks, and may be stated as:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-191-en.gif" width="362" height="70" />
Text alternative for the equation

where:

?i for i = 1, 2, ..., n, referred to as risk loadings, measure the sensitivity of the asset to each of the risks that individuals care about and vary across assets; and

(Ri – Rf) are risk premia which measure the expected return compensation an individual must receive to bear one unit of the relevant risk.

Closed economy

[192] The traditional CAPM assumes individuals make investments within a closed economy, with portfolio choices taken from a stock market which represents all the investment choices available to individuals. If the assumption of a closed economy is relaxed, and it is assumed that investors will make investment decisions in an international environment where they can buy assets from any financial market in the world, the result is what is called an International CAPM, which is generally stated as:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-192-en.gif" width="363" height="24" />
Text alternative for the equation

where:

Rf,DC is the domestic risk-free rate of return;

ßwm is a measure of the risk in the equity relative to the world market (or a larger market than the domestic market);

?DC is the sensitivity of domestic currency return to a change in the foreign currency;

WMRP is the world (larger than domestic) market risk premium; and

FCRP is the foreign currency risk premium.

[193] When only one foreign currency is involved, and it is assumed that the covariance between world market returns and foreign currency risks is zero (a very unrealistic assumption), the International CAPM may be stated more simply as:

where the world market risk premium (WMRP) is estimated as the difference between the world market return and the domestic risk-free rate, and the foreign currency risk premium (FRCP) is estimated as the difference between the domestic risk-free rate and the foreign currency risk-free rate.

[194] It can be shown that the International CAPM is a special case of the more generalized CAPM, and that under certain additional restrictive assumptions (specifically, that the hedging of foreign currency risk is zero or negligible, another very unrealistic assumption), the third term in the International CAPM formulation above (the term relating to currency risk premia) disappears. In such a case, the simplified International CAPM reduces to the domestic risk-free rate, excess returns in the integrated international market and the sensitivity of the company beta to the returns in the international market.

Stock index represents all investment options

[195] Theoretically, the market portfolio on which market returns are determined (Rm) should include all types of assets that may be held by a person as an investment, including real estate, works of art, etc. However, in practice, such a portfolio is impossible to observe and a stock index is usually substituted as a proxy for the true market portfolio. It has been observed (first by Richard Roll in 1977) that a stock index does not always serve as a true substitute for the full range of investment opportunities available in an economy, and that the substitution may lead to false inferences about the validity of the CAPM. It is proposed by some critics that as the true market portfolio is not observable, a true empirical test of the CAPM is not possible.

Efficient market hypothesis

[196] The efficient market hypothesis asserts that financial markets are "informationally efficient". This means it is impossible to consistently achieve higher than average market returns on a risk adjusted basis, given the publicly available information at the time the investment is made. This is one of the assumptions underlying the development of the traditional CAPM. However, the traditional CAPM has been found in numerous studies not to adequately explain the variation in stock returns. Specifically, research showed that low beta stocks tend to offer higher returns than predicted by the traditional CAPM, which tends to cast doubt on the validity of either the efficient market hypothesis or the CAPM development. Fama and French argued that two additional factors should be added to the traditional CAPM to account for the empirical observation that small capitalization stocks provide higher returns than high capitalization stocks, and that value stocks (those with high book to market ratios) provide higher returns than growth stocks. The Fama-French Three-Factor Model has been shown to explain 90 percent of diversified portfolio returns compared to the average of 70 percent explained by the traditional CAPM, and is stated as:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-196-en.gif" width="388" height="24" />
Text alternative for the equation

where:

SMB (or Small Minus Big) is a factor representing the historical excess returns of two categories of stocks, as measured in terms of market capitalization;

HML (or High Minus Low) is a factor representing the historical excess returns of two categories of stocks, as measured in terms of high and low book to market ratio;

ßFF is analogous but not equal to the traditional beta due to the presence of the two other factors;

Re, Rf and MRP retain the same meanings as for the traditional CAPM; and

bs and bv are determined through multiple regression analysis.

8.2 Positions of the participants

[197] CN and Manitoba propose alternative versions to the traditional CAPM for consideration.

[198] CN submits that while the traditional CAPM meets the Agency's objectives, international versions of the CAPM could be used to provide additional estimates to complement its results because CN competes for capital in international markets:

The empirical issues involved in implementing the CAPM have been widely studied during the past 40 years, however, and there is broad consensus regarding how to estimate unconditional (time-invariant) costs of equity capital. In light of this, an appropriate implementation of the unconditional CAPM has the potential to meet the three criteria outlined by the Agency (i.e., it is reasonable, reliable, and pragmatic).

We emphasize the CAPM can be applied in many forms. Canadian companies compete for capital from Canadian and international investors in a global financial market. It is clearly of utmost importance to account for this fact in a regulatory environment designed to provide fair returns in exchange for the capital required to improve and maintain railway assets. Versions of the CAPM that utilize a North American or World market can be easily implemented, and estimates of costs of capital from such models provide valuable additional information for the regulatory process.

We recommend that the unconditional CAPM be the primary determinant of the cost of capital and that international versions should be estimated to provide additional complementary rates of return.

[199] CN also asserts that the traditional CAPM represents a static framework which posits a fixed investment horizon and does not reflect the dynamic nature of the investment environment in which the railway company operates. Therefore, CN proposes that the more appropriate version to use is an Inter-temporal CAPM where the third (usually unspecified) term is replaced by a "term premium" which value can be approximated by replacing the single risk-free rate in the traditional CAPM with short-term and long-term risk-free rates:

The current state-of-the-art asset pricing models explicitly account for an unspecified investing horizon and provide implementable extensions of the CAPM. Most notably, the Intertemporal CAPM developed by Robert Merton has been empirically implemented by a number of researchers to provide expected equity returns in a dynamic environment. The authors Michael Brennan, Ashley Wang, and Yihong Xia ("Estimation and Test of a Simple Model of Intertemporal Capital Asset Pricing," Journal of Finance, 2004) show that in a setting where interest rates vary over time, risk premia will depend not only on market risk as measured by CAPM beta but also on the duration of the investment's cash flows. Roughly speaking, their Intertemporal CAPM specifies costs of capital as:

[200] Manitoba recommends the use of the Fama-French Three-Factor CAPM over the traditional CAPM:

In addition to the conventional application of that [traditional CAPM] methodology, the Agency should consider the results of the Fama-French Three Factor Model, described in Appendix B (page 117) of the Brattle Group report. This model is based on the findings of E. Fama and K.R. French that company size and the market-to-book relationship, in addition to the beta, strongly influence equity returns. This model enhances the CAPM by avoiding the implicit assumption that the beta is the sole measure of a company's risk relative to the market.

8.3 Practice of other regulators

[201] The use of any of the alternative versions of the CAPM other than the traditional CAPM in a regulatory application is not apparent. As noted earlier, discussions of the Inter-temporal CAPM and International CAPM appear to be confined mostly to academic literature. With respect to the use of the Fama-French Model by other regulators, the Brattle Report states (page 119):

The Fama-French model is not commonly relied upon in North American rate regulation, although it has been submitted for consideration in some instances. For example, in 2007, the Régie de l'énergie in Québec considered the Fama-French approach and found that the model had not been sufficiently examined to date to be used as a basis for setting the rate of return for a gas distributor. Similarly, Fama-French evidence has been presented at times for U.S. and U.K. regulatory proceedings. However, we are not aware of a U.S. decision that primarily relied on the Fama-French model although the U.K. Competition Commission used the model to determine whether a small company premium should be included in the cost of capital.

9.0 Relevant Market Data

Issue: Should the Agency continue to use Canadian data or use some combination of Canadian and U.S. data to establish the variables for each of the methods used to calculate the cost of common equity?

9.1 Context and relevance

[202] The discussion of the use of Canadian or U.S. data revolves around the cost of equity models. In particular, market data is required for the determination of the main inputs in the CAPM model, the risk-free rate, the market risk premium, and the beta coefficient. In the 1997 Decision, the Agency examined the issue and concluded that primary weight should be given to Canadian data and studies.

[203] However, in view of concerns raised by CN during the cost of capital review in 2004, and considering the changing investment environment, in particular the fact that the shares of both CN and CP were traded on the New York Stock Exchange (NYSE), and that both companies raise significant amounts of capital in U.S. markets, the Agency conducted a thorough re-examination of its position of exclusive use of Canadian data in the cost of common equity calculation, and concluded:

With regards to the appropriate benchmark to be used for cost of equity determinations, the Agency has taken all of the above elements into consideration. Mindful that the primary determining factor in this issue is not shareholder mix, but rather the relevance and relativity of the United States experience on a cost of equity rate that has a focused application on a specific segment of the railway companies' Canadian market, the Agency determines that cost of equity estimations should continue to be based on Canadian data.

[204] In the years subsequent to the 2004 Decision, either one or both of CN and CP have continued to object to the Agency's use of exclusively Canadian data.

9.2 Positions of the participants

Canadian Pacific Railway Company

[205] CP submits that both railway companies trade on the NYSE and operate very significant parts of their networks south of the border, and that both Canadian railway companies compete with other Class 1 railway companies in a North American environment for equity. CP submits, therefore, that a combination of Canadian and U.S. data could be used to establish the variables for the methods it advocates using to calculate the costs of debt and of equity. However, CP does not suggest a methodology for how the combined data could be used in the calculations.

Canadian National Railway Company

[206] CN submits that Canadian investors have greatly increased their exposure to investment opportunities outside of Canada over the past decade. CN states that the opportunity cost to a Canadian investor holding CN shares depends on the choices available in forming his or her portfolio. If there are no barriers to international investing, then the market portfolio includes assets from around the world, and the opportunity cost of any security would be determined by the systemic risk of the security relative to the global portfolio, reflecting the investors' global or integrated market opportunities. CN advocates that, therefore, international opportunities must be included in cost of capital estimates, and suggests that the simple North American CAPM is best suited for this purpose given its simplicity.

[207] CN acknowledges the potential impediments to capital mobility, which include explicit barriers such as legal restrictions, discriminatory taxes and transaction costs, as well as implicit costs like foreign currency risk. However, CN points out that legislative changes occurred to increase access since 1990, including reductions in limits imposed on foreign investment content in pension plans as well as the Multi-Jurisdictional Disclosure System introduced in 1991. It brought evidence suggesting an increased level of integration between the Canadian and U.S. markets, including a study in 2003[17] which found evidence that Canadian and U.S. markets are integrated to some extent, and changes to the the Canadian Pension Plan investment portfolio over the past 10 years, which is now dominated by investments in foreign markets.

[208] CN introduces two basic methods of including international opportunities:

  1. Replace the Canadian market portfolio with a North American market portfolio and conduct the standard CAPM analysis on this data.
  2. A more complex approach/estimate, the International CAPM (ICAPM), that includes an adjustment for unhedged currency exposure.

[209] CN indicates that the North American market portfolio it is recommending above refers to the U.S. capital market. CN submits that as the largest in the world, the U.S. market is the dominant international consideration. CN suggests that estimating the standard CAPM with the North American market portfolio would be better and simpler than estimating the full ICAPM, as the omitted currency hedging risk is typically quite negligible. CN does not provide any theoretical or empirical support for this assertion.

Western Canadian Shippers' Coalition

[210] WCSC submits that the Agency should continue to use only Canadian data to establish the variables used to calculate the cost of common equity, and indicated several factors that led to this conclusion.

[211] WCSC submits that, in theory, the market portfolio includes all risky assets that can be held by investors, which is far more extensive than the S&P/TSX Total Return Index. It includes assets in all countries, not just Canada, or Canada and the U.S., and comprises bonds private companies, real estate, human capital and more. WCSC submits that, in practice, a proxy is always used for measuring the market return, typically the S&P/TSX Total Return Index in Canada and the S&P 500 Index in the U.S. It emphasizes that there are major differences in the Canadian and U.S. capital markets, differences in taxation and exchange rate risk, and submits that it would not be correct to use U.S. data in Canada without adjusting for these differences. Furthermore, WCSC submits that if it were decided not to use only Canadian data, there are many other countries' data that should be considered.

[212] WCSC also submits that the regulatory applications for which the Agency determines a cost of capital rate are exclusive to Canada, and notes that CN and CP's operations in the United States are regulated by STB, which does not use Canadian capital market data to estimate the cost of capital for the railroads that it regulates. WCSC asserts that the Agency should not use the U.S. data to estimate the cost of capital for railway operations in Canada that it regulates.

Canadian Canola Growers Association

[213] CCGA submits that within the context of the Agency regulating companies that are incorporated and conducting business within Canada, it considers it appropriate that the Agency rely on Canadian data in calculating the cost of common equity, and suggests that Canadian capital markets are sufficiently large and liquid to facilitate effective arbitrage with global financial markets.

Coalition of Rail Shippers

[214] CRS suggests that the International CAPM is not an appropriate model for estimating the cost of capital for regulated Canadian railway companies, on the grounds that the Canada-U.S. markets are not sufficiently integrated. CRS cites a conclusion drawn from a paper authored by its expert that the U.S. market does not make a statistically significant contribution to explaining the portion of the return of Canadian utilities that is not explained by the Canadian market. CRS also notes that in January 2009, the U.S. Federal Energy Regulatory Commission (FERC) refused to include a Canadian company in the proxy group it used to evaluate U.S. equity returns, based on FERC's reasoning that Canadian pipelines are subject to a significantly different regulatory structure that renders them less comparable to domestic pipelines under its regulation.

Province of Alberta

[215] Alberta submits that the Agency should use either a combination of Canadian and U.S. data or conduct separate analyses using Canadian and U.S. inputs, because the stocks of the two Canadian railway companies are traded on both the Toronto and the New York exchanges, and, therefore, the return should reflect investors' requirements on both sides of the border. Alberta does not suggest what weight should be attributed to each market.

Province of Manitoba

[216] Manitoba submits that the Agency should continue to use Canadian data because the investments in grain transportation to which the cost of capital is applied are entirely within Canada.

Province of Saskatchewan

[217] Saskatchewan submits that the Agency should continue using the CAPM methodology to calculate the cost of equity, but does not specifically address the issue of the Agency's use of exclusively Canadian data.

9.3 Rebuttal comments

Canadian National Railway Company

[218] CN notes that the WCSC submission starts by pointing out that central to the CAPM is the concept of a market portfolio that includes all assets available to investors, but then goes on to argue, among other things, that the regulated operations are in Canada only. CN points out that the fact that the regulated operations are in Canada is irrelevant to whether or not investors have access to investments in other countries. CN reiterates its initial position that Canadian investors have significant exposure to international investments, and that ignoring their ability to invest in other capital markets does not make this fact go away, nor does it mitigate the importance of U.S. returns to Canadian investors. CN concludes that the important consideration in defining the market portfolio is the access Canadian investors have to global security markets, following from Modern Portfolio Theory established in the 1950's, and that the location of the company's assets is irrelevant to this consideration.

Western Canadian Shippers' Coalition

[219] WCSC notes that CP recommended that the Agency use Canadian data, and that while CP noted that its methodology could be modified to incorporate U.S. data, it provided no explanation as to how or why this should be done.

[220] WCSC also notes that CN discusses the use of Canadian data, a combination of Canadian and U.S. data, and international data using an ICAPM. WCSC maintains that CN's supporting analysis does not arrive at a conclusion, though CN states in summary that it considers the simple North American market CAPM as best suited at this time given its simplicity and the fact that the omitted hedging risk is typically quite negligible. WCSC suggests that the problems CN sees in using the ICAPM are also applicable to incorporating U.S. data, and the simplicity argument is more applicable to continuing the Agency's current practice of using only Canadian data.

9.4 Practice of other regulators

[221] The Alberta Utilities Commission (AUC), in its 2009 Generic Cost of Capital Decision[18], found that the regulatory risk faced by the U.S. utilities in general remains materially higher than the regulatory risk of Alberta utilities. As a consequence, the returns awarded by U.S. regulators for U.S. local distribution companies (LDCs) would be expected to reflect this materially higher level of risk, leading the Commission to conclude that the U.S. allowed returns should not be used in determining a fair return for Alberta.

[222] In October 2009, the National Energy Board (NEB) issued a decision where it concluded that the formula-approach to the determination of a generic cost of capital for companies under its jurisdiction, which was implemented in 1995 via Decision RH-2-94, is no longer appropriate in today's financial environment. Decision RH-1-2008 arose in the context of Trans Quebec & Maritimes Pipelines Inc.'s (TQM) application for approval of the cost of capital to be used in calculating final tolls. Along with many changes to their cost of capital methodology that are not relevant to this discussion, RH-1-2008 marked the first time that NEB considered U.S. information in its calculation. NEB accepted the evidence of TQM that (i) Canadian and U.S. financial markets are integrated and, as a result, Canadian pipelines compete for capital with their U.S. counterparts; (ii) U.S. pipelines serve as comparables to Canadian pipelines; and (iii) U.S. LDCs serve as comparables to Canadian pipelines such as TQM[19].

[223] The CRTC, in deciding a benchmark rate of return for all telephone companies subject to a rate cap[20], stated that increased integration of world capital markets has a potential impact on the overall Canadian equity market risk premium, as it should, in theory, bring the Canadian market risk premium closer to that experienced in the U.S. equity market. The CRTC further stated that in view of the trend of national markets' convergence towards integration, weighted consideration should be given to U.S. data. The decision does not specify the weight, but states that equal weighting would be inappropriate.

[224] The Australian Competition & Consumer Commission, in dealing with pricing considerations related to the commercialization of national networks (rail lines and airports) through third party access to publicly owned infrastructure[21], stated that care should be taken when comparing financial results from other countries, because of inherent differences in factors affecting market risk, such as taxation systems.

9.5 Brattle Report

[225] Appendix B of the Brattle Report discusses the importance of selecting a proxy for the market that comes closest to the true market return, while at the same time mitigating potential data problems. Further, because capital markets in the U.S. and Canada may be becoming increasingly integrated, it questions whether a strictly Canadian market index is the appropriate market choice, or whether a portfolio of Canadian and U.S. indices is more appropriate.

[226] The Brattle Report indicates that for companies that have significant cross-border operations and which raise capital in both markets, it may be preferable to use the merged market return as opposed to a domestic market return only. The Brattle Report notes in its Appendix that it is generally recognized that the market risk premium in Canada will be affected by the investment opportunities in the U.S. and the rest of the world.

10.0 Risk-Free Rate

Issue: Bearing in mind the stability and responsiveness implications and given that the Agency develops cost of equity rates annually, should the Agency develop risk-free rates using short-term or long-term bonds, or a combination of both, and what time horizon should it use for each?

10.1 Context and relevance

[227] The risk-free rate, which is integral to the CAPM, is the rate of return on a risk-free investment available to investors. The influence of the risk-free rate on the estimated cost of equity rate is illustrated in the estimation process of the CAPM:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-227.gif" width="332" height="82" />
Text alternative for the equation

where:

(Rt)^ is the estimated rate of return on equity for the current year t;

t)^ is the estimated beta factor obtained through a regression model;

R(m,j) is the composite rate of return on the market in historical year j;

R(f,t) is the risk-free rate of return for the current year t;

R(f,j) is the risk-free rate of return in historical year j; and

T is the number of years over which the market risk premium is estimated.

[228] This equation simply says that the expected return on the asset at any time is the return on a risk-free asset plus a factor (ß) of the market risk premium over the risk-free asset. Thus, selection of a risk-free rate is crucial to the expected cost of equity. If, as is typically the case, a government debt instrument (such as a bond or Treasury bill) is used as a proxy for the risk-free asset, a complication arises, because the rate of return (or yield) provided by the risk-free asset is then determined by the maturity period of the debt instrument.

[229] Under normal economic circumstances, the longer the maturity period of a bond, the higher the interest rate (or yield) required to persuade investors to buy the bond. Two explanations are generally advanced to account for this phenomenon. First is the market's expectation of future inflation and its effects on the value of money. A higher rate compensates investors for the willingness to wait a longer period for their money. Second, a higher rate also compensates investors for their willingness to accept greater uncertainty and the risk of unforeseen events that might impact the investment, as the period of maturity increases. The positive relationship of longer maturities with higher yields is known in Finance as a normal yield curve.

[230] Under some unusual economic circumstances, for example when investors anticipate a sustained deflationary period ahead, the situation described above reverses, and shorter maturity bonds are assigned higher yields than longer maturity bonds. This negative relationship of longer maturities with lower yields is known in Finance as an inverted yield curve. It has been shown that strongly inverted yield curves have historically preceded economic depressions.

[231] The preceding discussion therefore means that, so long as the current yield curve is normal, using the yield from a shorter maturity bond results in a lower risk-free rate and, consequently, a lower cost of equity. The reverse is also true; a longer maturity bond results in a higher risk-free rate and a higher corresponding cost of equity. Yield curves continually adjust to reflect action in the bond market, but all the rates on the curve typically adjust in the same direction so that the whole curve shifts up and down as interest rate levels rise and fall. The government bond yield curves for Canada and the U.S., based on annualized monthly yields as of August, 2011, are shown in Figure 2.

Figure 2: Canada and U.S. government bond yield curves, August 2011

Figure 2: Canada and U.S. government bond yield curves, August 2011425-R-2011/app-b-exhibit2-en.gif" width="590" height="394" />

Sources: Bank of Canada, U.S. Federal Reserve Bank

[232] In the 1985 Decision, the RTC reasoned that the maturity period chosen should correspond to the investment horizon of investors, and that some investors have a long-term horizon while others invest for short-term purposes. The Agency determined that an average of short term and long term takes into account both kinds of investors. Since then, the Agency has used Government of Canada marketable bonds to represent risk-free assets, and has assessed them from the perspective of both short-term and long-term maturities, with 1-3 year bonds for short-term and 10+ year bonds for long-term maturities.

10.2 Positions of the participants

Canadian Pacific Railway Company

[233] CP proposes that a long-term maturity period be used to develop risk-free rates as it is consistent with the railway industry's long-term investment horizon. In addition, CP states that combining a short-term and a long-term is underestimating the cost of equity. CP submits calculations to demonstrate that choosing a long-term time period for both the calculation of the market risk premium and the risk-free rate maturity leads to higher values as proof of its position.

[234] CP also submits that if the Agency were to adopt the practice of obtaining the market risk premium from Morningstar/Ibbotson as it proposes (see market risk premium issue for more details), it should update its risk-free return to be the return on long-term Canadian bonds to harmonize with the long-term bonds used in this publication.

Canadian National Railway Company

[235] CN discusses the traditional formulation of the CAPM and points out that the risk-free rate plays two potentially distinct roles, as can be seen if the CAPM equation is written as:

[236] CN points out that Rf1 in the CAPM equation provides the forward-looking risk-free return to apply to a zero beta asset, while Rf2 is required to calculate risk premium to an asset with non-zero beta. CN disagrees with the Brattle Report recommendation that the same debt instrument should be used to provide the two risk-free rates Rf1 and Rf2. CN argues that a weakness in the traditional CAPM is that it assumes only one fixed investment horizon, and that dynamic asset pricing models have been developed and implemented in the academic literature to relax this restrictive assumption, with Merton's Inter-temporal CAPM as an example. CN cites a study by Brennan, Wang and Xia (2004), which it claims "show that in a setting where interest rates vary over time, risk premia will depend not only on market risk as measured by CAPM beta but also on the duration of the investment's cash flows." CN then presents the Inter-temporal CAPM used in that study, which specifies the cost of equity as:

[237] CN notes that in this dynamic asset pricing model the risk-free rate clearly is provided by short-term riskless investments which, ignoring inflation, would be represented by short-term government bills. CN then cites some research which suggests that the term premium represents the difference between short and long bond yields, which can be estimated from the yield curve. Based on this, CN suggests that dynamic asset pricing for CN can be captured by replacing (Rf1 + Term Premium) in the above equation with a long bond return. In effect, CN submits that two risk-free rates could be used simultaneously in the traditional CAPM to produce the same result as an Inter-temporal CAPM: a long bond yield to represent the risk-free rate plus time-related risks, and a short government bond to represent the true risk-free rate. CN suggests that this implementation of the CAPM has solid justification in economic theory.

[238] CN explains why the potential exists for a negative correlation between unanticipated changes in interest rates and railway company cash flows, which would act to increase the effective duration of cash flows and to increase the cost of capital. CN concludes:

To summarize, neither the short- nor long-run CAPM should be expected to provide an appropriate cost of capital in an environment with uncertain future interest rates. Recent research suggests that long-duration cash flows warrant a risk premium in excess of that predicted by the short-run CAPM. Adding a "term premium" to the short-run CAPM is justified. The CAPM, with rf1 estimated from long-run government bond yields and rf2 provided by historical short-run government bond returns, can be implemented to provide costs of capital while accounting for the additional risks associated with the long-run cash flows generated by CNs regulated and unregulated operations.

Western Canadian Shippers' Coalition

[239] WCSC submits that, in theory, the risk-free rate should match the period of regulation; that is, if the cost of capital is being determined for one year, the one-year Government of Canada bond yield should be used. However, as short-term bond yields are more volatile than long-term yields, WCSC considers it understandable that a regulator might consider stability of rates to be a goal. Noting that the Agency monitors bond yields to assess their reasonableness, WCSC submits that the Agency's current practice represents a reasonable approach to accommodate the goal of rate stability.

[240] WCSC considers the use of the Inter-temporal Model proposed by CN to be a violation of accepted practice and notes that under perfect regulation, the return allowed each year is equal to the investors' required return, regardless of the length of the railway industry's investment horizon.

Canadian Canola Growers Association

[241] CCGA notes that short-term rates can be affected by current and projected monetary policy and near term inflation rates and longer term rates reflect expected future inflation and economic growth rates. It considers a blended rate to be more appropriate and better reflective of the railway companies' capital financing activities and related costs, and that the Agency's choice of 1-3 year and 10+ year terms to reflect short and long-term respectively is appropriate.

Coalition of Rail Shippers

[242] CRS submits its conclusion that, based on empirical findings for the tests of the CAPM, accuracy problems of the CAPM related to allocational efficiency have occurred. CRS suggests that using long Canada bonds as the risk-free proxy accounts for such accuracy problems. CRS specifically cites the use of the normalized yield forecast for 30-year Canada bonds or other less or no risk proxy to be correct.

Provinces of Alberta and Manitoba

[243] Alberta and Manitoba both submit that the yields on 1-3 year short-term government bonds are the best indication of the risk-free rate. Because yields on shorter terms (e.g. six months) reflect fluctuations in liquidity and are likely to be unstable, and future inflation may erode the value of the underlying long-term bonds, neither can be considered fully risk-free.

Province of Saskatchewan

[244] Saskatchewan submits that the Agency should continue its current use of the CAPM but did not comment specifically on this subject.

10.3 Rebuttal comments

Canadian Pacific Railway Company

[245] CP submits that while some of the submissions advocate that a short-term risk-free rate be used in implementing the CAPM, the Brattle Report pointed out, in CP's view correctly, that the risk-free rate used directly in the CAPM formula should be the long-term risk-free rate because equity can be viewed as a long-term claim on the firm's assets, and therefore the relevant alternative risk-free investment is a long bond. CP submits that the risk-free rate used in the CAPM formula should also have the same time horizon as the risk-free rate that is used in calculating the market risk premium, and that in general, medium- to long-term risk-free rates (10 years or more) are appropriate when calculating the market risk premium for the railway industry because such durations are consistent with the railway industry's long-term investment horizon. CP maintains that because the Morningstar/Ibbotson estimate of the Canadian market risk premium is measured relative to a long horizon risk-free rate, using the Morningstar/Ibbotson estimate of the market risk premium resolves any uncertainty over the appropriate risk-free rate to use in the CAPM formula.

Canadian National Railway Company

[246] CN observes that investors in CN are claimants to cash flows produced by long lived assets, assets that last more than a single regulatory period, and accordingly, that they are exposed to changes in the term structure of interest rates. CN notes that the WCSC submission claims that, in theory, the risk-free rate should match the period of regulation, yet WCSC does not make clear what theory is being referred to, though it seems to be referring to the single period CAPM. CN claims that academics have long recognized the shortcomings of a single period CAPM, which assumes the investment is made for one period and the relevant default free rate matches this single period, but in reality, investment in rail transportation facilities involves many overlapping long- and short-term investments.

[247] CN submits that if the investment is made in a regulated context where the regulator resets cash flows periodically, it does not change the fact that the life of the asset is long lived and that investors face significant interest rate risk. CN suggests that recent research has added to understanding of the "term premium" that investors require on long lived assets, and submits that the duration of the cash flows, even if regulated, is the primary determinant of the term premium in the default free rate in the CAPM, not the regulatory review period. CN submits that for this reason, yields on long-term bonds, that incorporate the term premium, should be used for the default free rate.

Western Canadian Shippers' Coalition

[248] WCSC notes that CN proposes neither the short-run nor long-run versions of the CAPM be used, and instead recommends using the long-term government bond rate for the risk-free rate, and the short-term bond return to obtain the expected risk premium. WCSC maintains that although CN attempts to justify this by claiming the long-term rate represents a short-term rate plus a term premium, the effect of this is the same violation of accepted practice. WCSC submits that in theory, the risk-free rate should match the period of regulation; that is, if the cost of capital is being determined for one year, the one-year Government of Canada bond yield should be used.

[249] WCSC also disagrees with CP's recommendation to adopt the Morningstar/Ibbotson long-horizon market risk premium. WCSC submits that once the decision on how the calculation should be done is made by the Agency, the data for the S&P/TSX Total Return Index and the various Government bond series are all easily obtainable and the calculations are not difficult, and that there is nothing magical about data services such as Morningstar/Ibbotson that put their calculations beyond question.

10.4 Practice of other regulators

[250] STB indicated that the parties (shipper and carrier associations and other interested parties) responding to its Notice of Proposed Rule-Making (NPRM) all agreed that the 20-year Treasury Bond is a more appropriate measure of the risk-free rate of return than the 10-year Treasury rate proposed in the NPRM. STB adopted this measure, stating that it believes this modification will have a very small effect on the final calculations[22].

10.5 Brattle Report

[251] The Brattle Report suggests that, while short-term bond yields are closer to a truly risk-free rate, other considerations have been used to justify selection of longer term bond yields. The report states (pages 22-23):

The CAPM is typically implemented using a long-term risk-free interest rate or a short-term risk-free interest rate. Using short-term Treasury bills as the risk-free asset seems most in line with the traditional CAPM – the return on Treasury bills is the closest to a truly risk-free rate of return, and the shorter horizon is closer to the 2-period nature of the CAPM. However, it has become common in many regulatory settings to implement a long-term version of the model using a long-term government bond yield as the risk-free rate and an MRP relative to the long-term bond yields. There are several justifications given by analysts and regulators for doing so. One is that regulated rates are set periodically, which means that current cost of capital estimates will determine rates for a potentially long period (potentially years). Estimates must therefore be set to be reasonable (on an expected basis) over that period. Short-term rates are a tool of monetary policy, however, and are much more affected by efforts of a country's national bank to alter economic activity other than long-term rates. As a result, short-term rates are more volatile. Cost of equity estimates based on short-rates could therefore change rapidly over the course of a few months. The recent financial crisis has demonstrated that short-rates may be significantly affected by short-run economic considerations, making their use in setting allowed rates over a five or six year horizon questionable. Another reason given for using a long rate to estimate cost of equity is that equity can be viewed as a long-term claim on the firm's assets, and therefore the relevant ‘alternative risk-free investment' is a long bond.

10.6 Implementation

[252] For implementation purposes, it is important to note that the risk-free rate appears twice in the CAPM and in different capacities, as may be seen in the following estimation equation:

[253] The first risk-free rate, Rf, is a forward-looking risk-free rate in the economy, while the second, Rf,j, describes the historical risk-free rate in year j during the time period T, and is used in estimating the market risk premium.

[254] The Brattle Report discusses common practices in Canada in selecting the risk-free rate (page 19):

Selection of the risk-free rate is generally one of the least controversial elements of the implementation process. The usual approach is to use the current yield or a forecasted yield on the home country's government debt, since government debt is generally considered to be free of risk, at least in countries such as Canada with well developed financial markets.

[255] The Brattle report then goes on to provide additional guidance with respect to selection of the risk-free rate (page 24):

After deciding on whether it is more appropriate to use a short-term or long-term risk-free rate, the next decision is whether to use a current yield or a forecast yield as the interest rate. In using a current yield or a forecast yield on long-term government bills or bonds as the risk-free rate in the CAPM, it is essential to ensure that the rate used is meaningful and not unduly influenced by either a single day or forecast. For example, if a current risk-free rate is used, the analyst can use either an average over a short period (often 10-15 trading days) or check for unusual changes in the yield around the day or period of interest. If a forecast rate is used, it is important that it is a consensus of market expectations rather than a forecast developed by a single analyst or entity. In addition, if a historically developed MRP is relied upon, it is important that the security relied upon as the risk-free rate (e.g., the 10-year government bond) has a consistent series of historical data available over a relatively long period for analysis. Otherwise, it would not be feasible to develop the historical MRP that corresponds to the relied upon risk-free rate.

11.0 Market Risk Premium

[256] The market risk premium (MRP) is an estimate of the premium provided by historical stock market returns over the risk-free rate.

11.1 Averaging period

Issue: What averaging period should the Agency use to develop the market risk premium used in the CAPM?

11.1.1 Context and relevance

[257] The time period over which the MRP is estimated can have a significant influence on the estimate, and hence on the cost of equity, as illustrated in the estimation equation presented earlier and reproduced below for convenience.

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-257-en.gif" width="336" height="106" />
Text alternative for the equation

where:

(Rt)^ is the estimated rate of return on equity for the current year t;

t)^ is the estimated beta factor obtained through a regression model;

R(m,j) is the composite rate of return on the market in historical year j;

R(f,t) is the risk-free rate of return for the current year t;

R(f,j) is the risk-free rate of return in historical year j; and

T is the number of years over which the MRP is estimated.

[258] In the 1985 Decision, the RTC stated:

The Committee also believes that the market risk premium is not constant over time, but changes in response to market conditions.....In addition, the Committee has reservations regarding the use of very long term historical premiums which may not reflect changes in income tax legislation as well as other factors.

[259] In the 2004 Decision, the Agency reaffirmed its methodology and made it more specific:

In this regard, the Agency acknowledges the commonly accepted principle that an average based on a long data series minimizes the distorting impact of an unusual year and incorporates a range of outcomes. The Agency further notes that long-term averages over periods of 30 years or more do tend to produce stable results. However, building on its earlier conclusions, the Agency finds that a very extended time period is inappropriate because it puts emphasis equally on recent and early historic data. The Agency is of the opinion that in forecasting the future, the distant past cannot be considered as relevant as the current past. The Agency considers market data reflecting the business practices, investor behaviour and expectations, government policies, accounting practices, taxation rules and the politics of modern times to be a better predictor of reasonable future returns on equity. The Agency is also of the opinion that a moving time period keeps these reflections contemporary.

[...]

The Agency concludes that it will continue to assess the market risk premium on an ongoing basis, by use of a time period that has sufficient length to incorporate many business cycles, periods of low and high performance, periods of volatility and stability, as well as to reflect the impact of unusual events and significant changes that the modern world has undergone. The Agency further concludes that the time period assessed will be one that will allow for the influence of these factors not unduly diluted by earlier historic data, that will produce results that are relatively stable, year over year, yet realistically reflective of any trends.

[260] Since the 1997 Decision, the Agency's method for developing MRP has been to use a 45-year moving historical analysis period. One or both of the Class 1 railway companies have consistently submitted that a longer analysis period should be used.

11.1.2 Positions of the participants

Canadian Pacific Railway Company

[261] CP submits that a long time period should be used to calculate the MRP. It views a long-run MRP as the best representation of the estimate of future performance because many types of historical events recur over time. CP also submits that a long-run MRP tends to be stable and that a stable cost of capital rate is essential for efficient planning and investment decisions, considering the railway industry's exceptionally long-lived physical assets.

[262] CP advocates a risk premium that incorporates all readily available market data, considering it to be more informative than one based on the 45-year period currently favoured by the Agency. CP references the STB determination, in Ex Parte No. 664, that using data from 1926 is the superior and more standard approach.

[263] CP submits that by excluding any periods of high and low returns that occur outside of the 45-year window currently used by the Agency, the non-diversifiable risk faced by equity investors is understated. CP further submits that there are no concerns about data reliability for the period from 1926 to the present.

[264] In conjunction with this and other factors, CP strongly submits that using a reputable and independent source such as Morningstar/Ibbotson's annual publication of MRP estimates, Canadian Risk Premia Over Time Report, which uses data dating back to 1936, would be reasonable due to its transparency, reliable as it is consistently produced and pragmatic as it is based on a readily available report that is simple to access and easy to use. Another factor contributing to this position is the issue raised by CP of the use of income returns instead of total returns, which is discussed further in this Appendix.

[265] CP rejects CRS's suggestion that MRPs based on historical averages should be adjusted downward because of an upward bias evident in some financial economics research, considering it to be the unproven product of evolving research. CP supports the estimates in Morningstar/Ibbotson's report as being squarely within the range of estimates reported in financial economics literature.

Canadian National Railway Company

[266] CN submits that in estimating the risk premium, more observations provide more confidence, as long as data is not used from a period where fundamental risk determinants were different than they are today. It is CN's position that there is no evidence of any such structural breaks to the fundamentals and as much data as is available should be used.

[267] CN further submits that because short-run changes in the risk premium are volatile, the use of a moving time period may result in volatile and inaccurate estimates. CN does however state, from observations of the actual historical MRP based on Canadian and U.S. returns from the period 1935 to present, that it seems that market returns relative to the risk-free rate in the past few years are similar to returns observed long ago.

Western Canadian Shippers' Coalition

[268] WCSC submits that the averaging period used to develop the MRP should be a long period for which high quality data is available, because it is objective and avoids the risk that a weighting scheme will introduce subjectivity and bias.

[269] WCSC indicates that at present there are 54 years of reliable data available to assess for MRP, with data from 1957 for the S&P/TSX Total Return Index and from 1936, 1950 and 1951 for 91-day Treasury bills, 1-3 year Canada bonds and 10+ year Canada bonds respectively. WCSC indicates that a wide range of market conditions, periods of double digit inflation, deflation, oil price shocks, booms, recessions and financial crises have occurred within the period from 1957-2010, and it is reasonable to believe that these types of events will continue to occur.

[270] WCSC states that a longer period is not better if the quality of the data deteriorates, referencing that pre-1957 only spliced data of poorer quality are available for the S&P/TSX Total Return Index, and some bond series are not available or have poorer quality data before 1950. WCSC also submits that important changes in the Canadian capital markets, such as the introduction of the dividend tax credit, reforms by the Bank of Canada in 1953-1954 and structural changes in the bond market have occurred and affect the period before 1957.

Canadian Canola Growers Association

[271] CCGA considers that the 45-year MRP averaging period currently used by the Agency is a statistically valid sample size and is sufficiently long to reduce the impact of any single year. It submits that it is not evident that a longer averaging period would produce a materially better measure of risk premium in spite of further lessening the impact of any single year within the period.

Coalition of Rail Shippers

[272] CRS considers the method of developing MRP by measuring historical realized risk premiums, using long periods of time to be based on the invalid underlying assumptions that returns are independently, identically and normally distributed and markets are allocationally efficient over long periods of time. CRS submits that there is emerging evidence that the use of this method produces an estimate that while statistically unbiased is overstated and should be adjusted downward.

[273] CRS advocates the use of survey evidence from estimates of investment professionals and economists for prospective market returns, citing two Canadian sources for such forecast survey publications, Mercer (Fearless Forecast), and Towers Watson (Economic Expectations).

Provinces of Alberta and Manitoba

[274] Alberta and Manitoba expressed concern over the use of a prospective, rather than historic approach to determining risk premium, However, this viewpoint was not expressed in the context of developing the MRP input to the CAPM, but rather in the context of the use of the equity risk premium approach to determining the cost of equity. Alberta and Manitoba did not express specific views on the method of calculating the equity risk premium for the CAPM.

Province of Saskatchewan

[275] Saskatchewan submits that the Agency should continue its current use of the CAPM but did not comment specifically on this subject.

11.1.3 Rebuttal comments

Canadian Pacific Railway Company

[276] CP submits that conceptually, the MRP is the excess return over the risk-free rate that investors demand for holding a well-diversified portfolio of investments, and that, according to the widely used textbook by Damodaran, the MRP is most often based on a very long time series of historical data. CP also notes that Morningstar/Ibbotson provides an estimate of the Canadian long-horizon market risk premium that uses Toronto Stock Exchange and Canadian government bond income return data beginning in 1936. CP submits that the Morningstar/Ibbotson report is publicly available and eliminates the need for the Agency to continually update the necessary data and estimate its own version of the market risk premium.

Canadian National Railway Company

[277] CN agrees with WCSC that using all available data avoids debates about subjectivity and bias that would result if a sub period were used. CN notes that WCSC also recognizes that data is available as far back as 1923, yet it recommends that only data from 1957 to the present should be used. CN considers this conclusion inconsistent. CN maintains that, for the reasons set out in its initial submission, as much data as possible should be used to estimate the risk premium.

Western Canadian Shippers' Coalition

[278] WCSC reiterates its original position.

11.1.4 Practice of other regulators

[279] STB uses a time window from 1926 to the present to estimate the MRP through a historical average. It states[23]

We are now persuaded that basing the equity-risk premium on returns dating from 1926 is the superior and more standard approach. We are cognizant of the literature, cited by several parties, indicating that some experts believe that the forward-looking equity-risk premium should be lowered to reflect the impact of higher price/earnings ratios. For example, the expert for the AAR directed the agency to an adjusted equity-risk premium published by Morningstar/Ibbotson that seeks to reflect the upward trend in price-earnings ratios and reduces the forward-looking equity-risk premium. We acquired the cost-of-capital book published by Morningstar/Ibbotson so that we might carefully review that alternative figure. But while Morningstar/Ibbotson does report such a figure, which falls in the 6% range, the company itself continues to rely on returns dating from 1926 in its own CAPM calculations. Moreover, WCTL [Western Coal Traffic League] submitted evidence showing that most commercial vendors of cost-of-capital information use this same figure in their CAPM calculation. Accordingly, we will follow the standard approach and use the historical average from 1926.

11.1.5 Brattle Report

[280] The Brattle Report states (page 25-26):

Some argue that returns over more recent periods are likely to be a better measure of investor expectations going forward, because the economy and capital markets have evolved so much over time. Alternatively, some argue that using the historical arithmetic average of this excess going back as far as possible provides data spanning many different economic environments and therefore provides the best measure on an unconditional basis (albeit not necessarily on a conditional basis). This is the approach that Ibbotson Associates take in their estimates of the historical MRP and the approach adopted by the Surface Transportation Board in STB Ex. Parte 664.

11.2 Averaging methodology

Issue: An issue raised during the consultation is whether the MRP should be calculated using arithmetic or geometric averages.

11.2.1 Context and relevance

[281] The most common methodology for estimating the MRP is to derive excess returns of the stock market over bond returns for a historical period, then average the results. There is some debate in finance literature over whether the arithmetic average or the geometric average is the more appropriate. The Agency currently uses the arithmetic average.

[282] The arithmetic average (or arithmetic mean) of a series is the sum of all n observations in the series divided by the number of observations, n, and in the case of annual observation of market returns, would be written as:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-282-en.gif" width="330" height="72" />
Text alternative for the equation

where Rm,t is the observed market return in year t and T is the total number of years.

[283] The geometric average of n observations, on the other hand, is the nth root of all n observations multiplied together, and, in the case of annual observations of market returns, would be written as:

11.2.2 Position of CRS

[284] Only one party, CRS, commented on the averaging methodology. CRS suggests the common argument that the best forward-looking predictor of market returns is an arithmetic and not a geometric average is based on the contentious assumption that the underlying distribution of market returns is normal and remains unchanged over the measuring period. Such a distribution is referred to in the academic literature as independent and identically distributed (IID) normal. CRS submits that there is emerging evidence that the use of the arithmetic instead of geometric averages can lead to misleading estimates (page 27):

The BG Report (page 32) provides a quote from Mehra and Prescott (2003), who are the authors who first identified the equity premium puzzle. Mehra and Prescott acknowledge that they reported arithmetic averages in their original article, since the best available evidence at that point in time indicated that (multi-year) stock returns were uncorrelated over time. They now acknowledge that the arithmetic average can lead to misleading estimates when returns are serially correlated, and that the geometric average may be the more appropriate statistic to use. [Emphasis added]

[285] CRS adds that long-term bond yield is a geometric, not arithmetic, type of average yield if the bond is held to maturity.

11.2.3 Brattle Report

[286] The reasons for the debate are summarized in the Brattle Report (page 27):

Generally speaking, the geometric mean is a backward looking measure of performance – that is, it provides a measure for comparing past performance across different securities or portfolios. Many economists therefore find that from a forward-looking cost of capital perspective, the arithmetic mean is more appropriate, since it reflects the expected value of future returns. Specifically, compounding the arithmetic return over a number of periods gives the expected compound return over those periods, but compounding the geometric mean does not.

Some financial economists, however, have suggested that this line of reasoning is flawed when returns are mean reverting; i.e., exhibit negative correlation between consecutive periods. When such is the case, the expected return may differ from the historical return and the arithmetic mean no longer provides an accurate measure of the expected return. The reason for this conclusion is that negative correlation introduces a degree of path dependence – above average returns one year are more likely to be followed by below average returns the following year – and vice-versa.

[287] The Brattle Report then presents a standard methodology for adjusting the MRP in cases where negative serial correlation exists in the returns (page 27):

In such situations, using a value between the geometric mean and the arithmetic mean is technically a more accurate estimate of the unconditional MRP. The amount of weight given to the geometric average will depend on both the forecast horizon and the degree of correlation, or memory, in the return series. Blume (1974) suggested a weighting of:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-287-en.gif" width="500" height="40" />
Text alternative for the equation

where H is the return horizon (over which we are forecasting an average return) and T is the size of the sample used to estimate the arithmetic average (e.g., for 1926 to 2009, T = 83).

[288] For a 1-year return horizon, H = 1, and the adjusted MRP becomes the same as the arithmetic average. This suggests that, even in cases where negative serial correlation is found to exist, the arithmetic average would be used if the cost of equity rate is being set for a 1-year return horizon.

11.3 Income versus total returns on bond rates

Issue: An issue raised during the consultation is whether the historical risk-free return used in the MRP should be the product of the income return or the total return on the bond instruments examined.

11.3.1 Context and relevance

[289] Although not explicitly highlighted as an issue in the Consultation Document, the method by which the historical risk-free return in the MRP calculation is determined was raised by CP in its submission.

[290] In theory, the risk-free return required for the CAPM calculation represents the return on a security that bears no risk. While the risk-free return cannot be directly observed in the marketplace, returns on Government of Canada debt instruments can be considered as good proxies. Each bond provides two types of returns, total and income. Total returns include three potential sources of return to the bond investor: the coupon payments, income from reinvestment of the coupon payments, and capital gains (or losses). Income returns comprise the cash flow received by the bondholder, and include only the coupon payments.

[291] The Agency currently estimates the risk-free bond returns in the MRP as the arithmetic average of the historical total returns on short (1-3 year) and long-term (10+ year) Canadian federal bonds over a 45-year period.

11.3.2 Positions of the participants

Canadian Pacific Railway Company

[292] CP suggests that one of the two major shortcomings of the Agency's methodology in estimating the MRP is that the Agency uses total instead of income return of bonds (the other being the length of the averaging period, as discussed earlier). CP proposes that the Agency should use the income return, which is the truly risk-free portion of the bond returns, and indicates that Morningstar/Ibbotson uses bond income return data. CP proposes that the Agency should use the annual MRP for Canada published by Morningstar/Ibbotson in its Canadian Market Risk Premia Over Time report. CP explains that the Morningstar/Ibbotson report provides an estimate of the long-horizon market risk premium that uses the TSE and government bond income return data beginning in 1936. CP submits that using the Morningstar/Ibbotson report would eliminate the need for the Agency to continually update the necessary data and estimate its own version of the MRP.

[293] CP also explains that Morningstar/Ibbotson uses bond income return data because only that portion of the return associated with the coupon payments can be considered truly risk free, and that the use of income return data helps to avoid any potential upward bias in the estimate of the MRP based on historical returns. CP notes that one of the reasons CRS cites for the potential upward bias in the historical MRP is "one major unfavourable accident, (a prolonged period of unanticipated inflation) for bondholders". CP submits that a prolonged period of unanticipated inflation would not have an impact on the MRP if income return data were used, as that data does not reflect changes in bond prices

Western Canadian Shippers' Coalition

[294] WCSC counters that the expected [market] risk premium is taken to be an average of the difference between the realized stock index return and the realized government bond return from each period, and that it is the realized total return on these long-term bonds that should be used in the calculation, not just the income return.

11.3.3 Brattle Report

[295] The Brattle Report notes that the use of income returns to calculate the risk-free rate is widespread in the industry. It states (page 21):

Standard practice in the academic literature and among valuation practitioners is to use the income return (i.e., bond coupon payment dividend by bond price) on long term bonds as opposed to the total return, because the total return includes capital gains or losses which are not risk-free. However, this was more of an established practice than a prescription by the traditional CAPM.

[296] In response to comments by other industry participant the Brattle Report again noted (page 5-6):

Morningstar SBBI 2011 Valuation Yearbook is clear on their recommendation to use income returns as an unbiased estimate of the risk-free rate for estimating the MRP.

12.0 Beta

[297] Beta is a measure of the systemic or undiversifiable risk in a given investment relative to the market as a whole. It is estimated as the non-intercept parameter in a simple linear regression of the investment returns against excess market returns, as shown in the equation:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-297.gif" width="266" height="24" />
Text alternative for the equation

where:

Re represents observations of historical returns associated with the investment;

Rf represents observations of historical returns from the risk-free rate;

Rm represents observations of historical returns associated with the market; and

a and ß are regression parameters.

[298] In reference to the actual beta estimation, the Brattle Report states (page 21):

The most common methodology to estimate betas is to use the most recent five years of weekly or monthly return data, and then adjust the raw estimate towards one as an adjustment for sampling reversion that was first identified by Professor Marshal Blume (1971, 1975).

[299] The issues that arise in beta estimations concern the type of returns and period over which the returns are observed, the type of adjustment for sampling reversion, and the use of levered or unlevered betas.

12.1 Return interval and analysis period

Issue: An additional issue examined by the Agency during the review is the appropriate return interval and analysis period to be used in estimating beta.

12.1.1 Context and relevance

[300] Return interval refers to daily, weekly or monthly observations of stock and market returns, and analysis period refers to the number of years over which the observations are taken. These issues were not raised by any party in the consultation but are nevertheless addressed here, as they are two decision elements affecting the implementation of the CAPM.

[301] In the 2004 Decision, the Agency found that an appropriate data series for beta calculation should cover a number of different economic scenarios and allow sufficient data points for statistical precision without introducing irrelevant historical information or daily "noise" that may bias the regression results. The Agency determined that a five-year measurement period should be used when possible. It considered two years to be the minimum acceptable measurement period, but anything less than five years should be considered only when a five-year history does not exist, unless company or industry specific events, such as a change in the company's focus, dictate that a shorter period should be used. The Agency was also of the opinion that the frequency of data measurement should be weekly or monthly and that betas should be adjusted for the mean-reverting tendency.

12.1.2 Brattle Report

[302] The Brattle Report explains the effects of the choices of return interval and analysis period on the beta estimates (page 35):

The choices for the interval for the return data and the length of the beta estimation window involve a tradeoff between obtaining more observations through the choice of a longer window and/or more frequent return data and ensuring that no structural change has occurred during the estimation window as well as that the return data are based on sufficient trading activity. For example, monthly data provides fewer observations unless a long enough estimation window is chosen (i.e., 5 years of monthly data gives only 60 data points – by contrast, a weekly horizon provides 260 observations over a 5 year period). Daily data is noisy with potentially few trades in any particular day.

Structural changes means that the risk of the asset relative to the market could change over the estimation period, so that the resulting beta estimate would be a "blend" of the risk of the asset over the historical estimation period instead of representing the forwarding-looking risk of the asset. The choice of a very long-run horizon (say, 10 years) introduces a potential problem for beta estimation. Specifically, many economic relationships shift in fundamental ways over a 10 year period. Asset risk relative to the market may not be stable over such shifts, so that return data from early in the estimation period represents a risk relationship with the market that is no longer applicable. In addition, the longer the estimation period, the more likely it is that data issues arise. For example, Canadian Pacific has less than 10 years of trading data.

12.2 Beta adjustment

Issue: An issue raised during the consultation is whether beta estimates should be adjusted for the mean-reverting tendency.

12.2.1 Context and relevance

[303] The beta calculated from a regression analysis as described above is an unadjusted beta, which, as per the 2004 Decision, should be adjusted for the mean-reverting tendency. The 2004 Decision notes that these measurement criteria are consistent with calculations done by commercial beta services. Thus, the Agency adjusts for convergence to unity, and uses the Blume methodology to make the adjustment. The Blume adjustment is the most widely used method of adjusting beta. Its formula is as follows:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-303.gif" width="198" height="29" />
Text alternative for the equation

where:

ßBA is the Blume Adjusted Beta; and

ßUA is the unadjusted Beta.

12.2.2 Position of CRS

[304] Only CRS filed comments on the beta adjustment, and it calls for the Agency to revisit its 2004 Decision that the beta should be adjusted for a mean-reverting tendency. CRS examines the case where the beta of a regulated entity is estimated from a portfolio or sample of regulated entities. CRS argues that if there is evidence that the beta of the regulated entity is mean reverting to the sample of regulated entities, then the appropriate adjustment for the mean-reverting behaviour of the regulated entity should be towards the mean beta of the regulated entities, not towards 1.0. CRS suggests that a cursory examination of the rolling window beta for both CN and CP suggests little evidence that their betas, which are generally below one, exhibit mean reversion to the market beta of one.

[305] CRS then argues that two main reasons are advanced for adjusting betas for a mean-reverting tendency, both of which are not valid. First, CRS submits, if interest rate risk is a return determinant then the appropriate way to deal with this is the specification of a two-factor asset pricing model where the beta of each factor is estimated and the risk premium of each factor over the long Canada yield is measured.

[306] CRS argues that the second rationale, based on empirical studies in the U.S. in 1971 and 1975 that provide evidence that betas for all firms in the U.S. market, on the average, revert to a market average of 1.0 over time, is equally fallacious. CRS submits that whether this result still applies, or whether it applies to specific stratified samples of firms is debatable and untested. Furthermore, CRS notes that there appears to be no similar tests for the Canadian market.

[307] Finally, CRS submits that statistical estimation errors in the beta forecast is the third reason generally advanced for adjusting the betas. CRS argues that for statistical estimation errors the Vasicek-shrinkage method and not the Blume-type method should be used. CRS describes the Vasicek method as a weighted average of the standard beta of one member of a sample of regulated entities, and the standard beta of the entire sample of regulated entities, where the weights are based on relative estimation errors. However, CRS maintains that, for CN and CP the betas are not estimated by adjustment from the combined betas of a sample of regulated entities, but directly. In effect, in terms of the Vasicek method the "member of the sample" is the same as the "sample", and the adjustment factor reduces to 1.0. Therefore, CRS claims that no Vasicek-type adjustment is needed for CN or CP.

12.2.3 Brattle Group response

[308] The Brattle Group responded to the issue of beta adjustment raised by CRS (as prepared by its expert, Professor Lawrence Kryzanowski)[24]:

A larger point raised by the Kryzanowski Report is whether to use adjusted betas or not. This is one of many topics in estimating the cost of capital on which reasonable arguments are available on both sides. However, we note up front that this issue is much less relevant for the railroad industry than for regulated utilities as the beta estimates for railroads are closer to 1.0 than are those of regulated utilities.

The Kryzanowski Report notes that there is no evidence that the betas of regulated companies drift to a value of 1.0 used in the standard beta adjustment procedure. We agree. However, use of adjusted betas is common in regulatory settings as well as by providers such as VL and others. The real issue is whether adjusted betas are a better estimate of the relative risk of regulated utilities than unadjusted betas. Regulated utilities are different than other companies in that the way they are regulated makes them particularly sensitive to changes in interest rates. Based upon first principles and previous tests of interest rate sensitivity, utilities were shown to be sensitive to interest rate changes in a statistically significant way. Recent tests have not demonstrated the statistical significance, but based upon first principles, the interest rate sensitivity should still be there in spite of our inability to detect it. Betas estimates for regulated companies (albeit not railroads) over the last 10 years have varied dramatically and have frequently been at or below zero which would imply a risk-free investment. Although regulated companies are less risky than the average stock, they are not risk-free so betas estimates close to zero are meaningless. In any case, the issue is of less importance for railroads because their betas are closer to 1.0 than are the betas of most regulated industries. The closer a beta is to 1.0, the smaller is the effect of using adjusted betas.

In summary, some analysts, such as The Brattle Group, rely upon adjusted betas for rate regulated entities because of the sensitivity of rate regulated companies to interest rates not because of a drift of the risk of regulated companies to a beta value of 1.0. Some analysts rely on adjusted betas because they believe that investors rely on adjusted betas because vendors typically report adjusted betas. Other analysts point to the lack of significance in recent tests of interest rate sensitivity and the lack of any evidence suggesting that the risk of regulated companies tends to increase toward the risk of an average stock in the market as reasons not to use adjusted betas.

12.3 Unlevering and relevering betas

Issue: An issue raised during the consultation is whether the Hamada equation for unlevering and relevering beta estimates was valid.

12.3.1 Context and relevance

[309] Unlevering is a process for separating the financial risk of a firm from its business risk. Only in special circumstances, where the company of a beta cannot be directly calculated because it is not publicly traded, does unlevering and relevering beta become an issue for the Agency. In cases where, for example, a passenger service rail complaint is filed with the Agency, and the railway company involved is not a publicly traded company and, therefore, market information to calculate a beta is unavailable, the Agency develops a peer group of traded railway companies, obtains their levered (market) betas and unlevers them to arrive at an average unlevered beta for the group. That average unlevered beta is then relevered, based on the debt to equity ratio and tax rate of the railway company in question. The Agency uses the equation developed by Professor Hamada for this purpose. Currently, the Agency uses company specific market betas in the cost of equity calculations for CN and CP and the issue of unlevering and relevering beta is not applicable to CN and CP.

12.3.2 Position of CRS

[310] The issue was raised by CRS in its comments on the Brattle Report's analysis of levered betas. CRS submits that using the Hamada equation to unlever and relever betas is fraught with problems. According to CRS, the assumptions that arise from the Hamada equation imply that the levered value of the firm increases indefinitely with leverage if the unlevered values of the firm remain constant along the WACC curve.

12.3.3 Brattle Report

[311] The Brattle Report describes a general methodology for calculating a fair return on equity for a regulated company using a sample of comparable securities. According to the Brattle Report, the market for corporate debt is generally not very liquid, and, as a result, it is not clear that reliable estimates for a given company's debt are available. A common practice is to assume that the beta of debt is zero, implying that the cost of debt should be the risk-free rate.

[312] Once a decision on the debt beta is made, the cost of equity for each company in the sample can be computed on an unlevered basis, and averaged to produce an estimate of the industry's unlevered beta. To estimate the cost of equity for the regulated company, this estimate of unlevered beta can be relevered to the regulated company's capital structure[25].

[313] The Brattle Report points out that the problem in using a set of comparable securities to estimate the cost of equity of the regulated company is that it ignores the fact that the underlying asset risk in each company is typically divided between debt and equity holders. Even though the risk of the underlying assets may be comparable, a different capital structure splits that risk differently between debt and equity holders, making the equity in one firm potentially more risky than equity in another. Increased leverage generally adds financial risk to a company's equity.[26]

13.0 Elements Of The Dcf Model

Issue: During the course of the consultation certain participants advocated for the use of the DCF methodology, or one of its variations, as a means to estimate the cost rate of common equity.

[314] The DCF Model is based on the principle that the price investors pay for an equity is the sum of all discounted cash flows investors expect to receive in perpetuity on the equity. The expected future cash flows, in turn, reflect the expected growth profile of the company into the future. Hence, the major elements that influence the cost of equity estimates using the DCF approach are the expected growth profile of the company, which is reflected in the choice of a one-stage or multi-stage model formulation, the growth rates assumed within the growth profile, the measure of the expected cash flows and, finally, the observed price. These issues are discussed in the following sections.

13.1 The growth profile

13.1.1 Context and relevance

[315] This issue describes whether the company's growth is assumed to be constant into the foreseeable future (single-stage model), or changes over future time periods (multi-stage model).

Single-stage or constant growth DCF Model

[316] In this model, the stock price is assumed equal to the sum of the future cash flows, where each amount is discounted for the time and risk between the present and the time the cash amount is expected to be received, with the company assumed to grow in perpetuity at a constant rate. The model states that the cost of equity is equal to the expected current dividend yield (dividend divided by price) allowing for the constant rate of growth, plus the expected future growth rate of the dividends. The model is shown mathematically as:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-316.gif" width="185" height="44" />
Text alternative for the equation

where:

RE is the return on equity;

D0 is the current dividend;

P is the current price of the stock;

D0/P is the current dividend yield;and

g is the dividend growth rate.

[317] Implementation of the single-stage DCF Model requires estimates of the constant growth rate, along with observations of the current price and the current dividend. The Agency currently estimates the one-stage DCF Model.

Two-stage DCF Model

[318] The assumption that the company will grow at a constant rate into the future in perpetuity is unrealistic and is considered to be a fundamental weakness of this Model. If there is reason to believe that the growth will vary in the future, the single-stage DCF Model can be expanded to allow for different growth rates to be used for different stages in the company's growth. The two-stage model assumes that a company will grow at one rate for a pre-determined number of years, and then will grow at a terminal growth rate in perpetuity after that time. The terminal growth rate is typically the nominal long-run rate of GDP growth. The two-stage DCF Model is shown mathematically as:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-318-en.gif" width="327" height="109" />
Text alternative for the equation

where:

g1 is the dividend growth rate that is assumed to apply for the first n years;

g2 is the terminal dividend growth rate;

Dn is the dividend in year n;

n is the number of years in the first stage;and

all other variables are defined as before.

[319] In the two-stage growth model, the first term is simply the present value of the first n years over which the dividends are expected to grow at rate ‘g1'. The second term is the present value of all dividends from year n to infinity.

Multi-stage DCF Model

[320] The DCF Model can also be extended to a multi-stage DCF Model, where more than two stages of growth are considered. The number of years in each stage is not fixed. While any number of stages of growth could be used, the most common multi-stage DCF Model is the three-stage model, where the first stage is assumed to be the company's own growth, which will settle down into the average industry growth rate at some point in the future, and finally settle down to the average growth rate of the overall economy (GDP) in perpetuity. The three-stage DCF Model is shown mathematically as:

Mathematical equation  - see long description for explanation425-R-2011/app-b-eq-320-en.gif" width="561" height="109" />
Text alternative for the equation

where:

Dn1 is the dividend in year n1;

Dn2 is the dividend in year n2;

g2 is the dividend growth rate that is assumed to apply for the second stage of the model;

g3 is the terminal dividend growth rate;

n1 is the number of years in the first stage;

n2 is the total number of years in the first two stages of growth;and

all other variables are defined as before.

[321] Implementation of the multi-stage DCF Model requires assumptions regarding the growth rate in and the length of each period (stage), along with observations of the current dividend and price.

13.1.2 Positions of the participants

Canadian Pacific Railway Company

[322] CP proposes that the Agency use a multi-stage DCF Model based on that presented in the Morningstar/Ibbotson Cost of Capital Yearbook, submitting:

The specific financial formula used to implement the MSDCF model of the cost of equity proposed here is based on that presented in the Morningstar/Ibbotson Cost of Capital Yearbook. Shannon Pratt's widely cited textbook on the cost of capital describes the Cost of Capital Yearbook as "a comprehensive source of industry-level financial data" that presents "[c]ost of equity, cost of capital, capital structure ratios, growth rates, industry multiples, and other useful financial data" on over 300 industries.

[323] CP indicates that the proposed methodology is consistent with the formula contained in the STB's January 23, 2009 decision (Service Date - January 28, 2009) in STB Ex Parte No. 664, Sub-No.1.

Canadian National Railway Company

[324] CN submits that the assumption in the one-stage DCF Model that dividends are a constant proportion of earnings and growth, on the average, at a constant rate "is unlikely to be true for most companies, including CN," adding that:

The currently reported "next year" (5-year) consensus earnings growth rate according to Thompson Financial, available from Yahoo Finance, is 14.8% (8.77%) for CN and 25.90% (8.25%) for CP. These are obviously not sustainable in perpetuity, and the direct use of even the five year growth rate will lead to an overstated cost of equity.

[325] CN further notes that the DCF Model presents challenges but concludes that using the DCF Model as a consistency check on the CAPM could be helpful in that a large divergence between the two estimates might indicate a problem with the CAPM. CN does not submit any position with reference to the single-stage DCF Model versus the multi-stage DCF Model, nor does it submit any opinion on inputs to the DCF Model.

Province of Manitoba

[326] Manitoba presents the median earnings growth rates currently forecast for CN and CP, and suggests that these growth rates are unsustainable in the long run. Manitoba then presents forecasts of the long-term growth in nominal Canadian GDP forecast by the Office of the Parliamentary Budget Officer, and concludes:

The predicted earnings growth rates for the railways are multiples of this forecast national GDP growth. The projection of these railway growth rates into the indefinite future would imply that the railways consume more and more of the national output over time – a totally unreasonable assumption. Ultimately, the sustainable rate of the railways' earnings growth is constrained by the growth of the economy in which they operate.

[327] Manitoba suggests that STB selected a three-stage implementation of the DCF Model because of this problem, and submits that if the Agency uses the DCF Model at all, it should use that of STB or a similar variant of the DCF application.

13.1.3 Practice of STB

[328] STB uses a three-stage DCF Model as one of two methodologies it uses to estimate the cost of equity for U.S. railroads. In its 2009 Decision (Ex Parte No. 664, Sub-No. 1, January 2009), STB noted that shipper groups had argued against the one-stage DCF Model that STB had previously used to estimate railroads' cost of equity on the grounds that the constant growth rate assumption was unreasonable, and stated (page 3):

Accordingly, we began this proceeding to explore in depth an appropriate multi-stage DCF that could be used in the Board's cost-of-equity determination. In the Advance Notice of Proposed Rulemaking (ANPRM), we identified the following four requirements that a multistage DCF model should satisfy: (1) the DCF model should be a multi-stage model; (2) it should not focus on dividend payments only; broader measures of cash flow or shareholder returns should be incorporated as well; (3) it should be limited to those firms that pass the screening criteria set forth in Railroad Cost of Capital – 1984, 1 I.C.C.2d 989 (1985) (Railroad Cost of Capital -1984); and (4) when combined with CAPM, it should enhance the precision of the resulting cost-of-capital estimate.

[329] STB concluded that it can improve its cost of capital determination by using a multi-stage DCF Model in conjunction with the CAPM to estimate the cost of equity for the railroad industry.

13.2 Measure of expected cash flows

13.2.1 Context and relevance

[330] An important aspect of the DCF Model is the measure of earnings. The DCF Model is based on the idea that the current value of a company is the present value of its future earnings. Typically, either dividends or cash flows are used to represent earnings. This input enters into the DCF Model as "D" in the various DCF formulas presented above.

13.2.2 Position of CP

[331] CP is the only party that proposed a detailed methodology for estimating cost of equity using the DCF Model. As noted above, CP proposed the use of the Morningstar/Ibbotson three-stage DCF Model. In CP's application of the Model, the measure of expected cash flows (CF) is defined as income before extraordinary items (IBEI), minus capital expenditures (CAPEX), plus depreciation (DEP) and plus deferred taxes (DT). That is:

CF = IBEI – CAPEX + DEP + DT.

[332] CP computes IBEI by deducting extraordinary items from net income, then constructs weighted averages over five years of the financial measures described in the equation.

13.2.3 Practice of STB

[333] STB also adopted the Morningstar/Ibbotson three-stage DCF Model, and provides the rationale for using cash flows and a description of the methodology:

The cost of equity in a DCF model is the discount rate that equates a firm's market value to the present value of the stream of cash flows that could affect investors. These cash flows are not presumed to be paid out to investors; instead, it is assumed investors will ultimately benefit from these cash flows through higher regular dividends, special dividends, stock buybacks, or stock price appreciation. The incorporation of these cash flows and the expected growth of earnings are the essential aspects of the multi-stage DCF we are adopting here.

The Morningstar/Ibbotson model defines cash flows (CF), for the first two stages, as income before extraordinary items (IBEI) minus capital expenditures (CAPEX) plus depreciation (DEP) and deferred taxes (DT), or

CF = IBEI – CAPEX + DEP + DT.

An average cash flow figure is used as the starting point of the analysis under the Morningstar/Ibbotson model. To find the average cash flow, the model uses the 5-year period leading up to the year being analyzed, and the total cash flows for that time period are divided by total sales, which determine the 5-year cash-flow-to-sales ratio. The ratio is then multiplied by the total sales for the year being analyzed to obtain the average cash flow estimate for that year. For the third (and final) stage of the Morningstar/Ibbotson multistage DCF model stage, Morningstar/Ibbotson uses two additional assumptions: that there is no depreciation or deferred taxes. Therefore, in the third stage, cash flows are based solely on income before extraordinary items.

13.3 Growth rates

13.3.1 Context and relevance

[334] Whether using a single-stage or multi-stage model, it is necessary to estimate a growth rate for each of the stages. The projected future growth rate is the most controversial input into the DCF Model, because it is not directly observable. A common approach is to take an average of analysts' growth rate estimates. The Agency's current practice for its single-stage DCF Model is to use consensus analyst estimates (from Yahoo Canada Finance) of five year growth rates for CN and CP. Alternatively, historical growth rates can be estimated for each company by calculating an average company growth over a number of years, using some appropriate measure of growth such as changes in revenue or earnings per share, and assuming that future growth will be the same as past growth.

13.3.2 Positions of the participants

Canadian Pacific Railway Company

[335] CP advocates use of the Morningstar/Ibbotson approach for implementing the multi-stage DCF Model. The methodology that CP recommends for estimating the earnings growth is as follows:

The first stage of the Morningstar/Ibbotson MSDCF model applies to a period that extends from one to five years in the future (the current year is considered to be year 0). In each year of the first stage, CP's annual earnings growth rate is assumed to be the average value of the "long term" (three- to five-year) earnings growth estimates made by railroad industry investment analysts during January, February, and March of each calendar year (which coincides with the release of the year-end financial statements). These analyst estimates are collected by the Institutional Brokers' Estimate System (I/B/E/S) and distributed by Thomson Financial through its Thomson ONE Investment Management service.

...

The second stage of the MSDCF model applies to a period that extends from six to ten years in the future. During this stage, cash flows are assumed to grow at the average of the investment analysts long term growth forecasts for the railroad industry, comprised of CP and CN.

...

The third stage of the MSDCF model begins 11 years in the future and continues in perpetuity. Starting in year 11, CP's growth rate is assumed to equal the long-run nominal growth rate of the Canadian economy. The long-run nominal growth rate used in the MSDCF estimate of the 2010 cost of equity is 5.74 percent, which is the sum of average historical real gross domestic product growth from 1961 to the present (3.23 percent) and the long-run inflation rate (2.51 percent).

The third stage growth rate is applied to a cash flow value that is based on two additional assumptions about the long run: (i) depreciation equals capital expenditures (i.e., zero net investment); and (ii) all tax expense is treated as a cash outflow (i.e., changes in deferred taxes are zero). That is, cash flow in the third stage of the model is based only on income before extraordinary items (IBEI), whereas in stages 1 and 2 it is based on the expression in Equation 2 above [CF = IBEI – CAPEX + DEP + DT].

Coalition of Rail Shippers

[336] CRS's comments relate mainly to the weaknesses it perceives in the predominant approach for estimating dividend growth rates. First CRS states that the DCF Model can be implemented using historical dividends and dividend growth rates as estimated from available data, but makes no further comment on this approach. CRS then states that the growth rates forecast by financial analysts are also used. CRS notes that the explicit assumption underlying the use of analysts' forecasts is that the forecasts do not exhibit known biases. CRS disagrees with what it considers to be an attempt by the Brattle Report to provide evidence that the effects of analysts' forecasts are less likely to be an issue for regulated utilities. CRS notes:

The literature documents that the earnings estimates of analysts exhibit substantial optimism and overconfidence biases, that revisions in analysts' forecasts cause variability in stock prices, and that the use of these upwardly biased estimates without removing the bias leads to an upwardly biased estimate of the required ERP [Earnings Risk Premium]and cost of equity.

13.3.3 Practice of STB

[337] The Morningstar/Ibbotson Model, as previously described by CP, was adopted by the STB in STB Ex Parte No. 664, Sub-No. 1 page 6:

Growth of earnings is also calculated in three stages. These three growth rate stages are what make the Morningstar/Ibbotson model a "multi-stage" model. In the first stage (years 1-5), the firm's annual earnings growth rate is assumed to be the median value of the qualifying railroad's 3- to 5-year growth estimates as determined by railroad industry analysts and published by Institutional Brokers Estimate System (IBES). In the second stage (years 6-10), the growth rate is the average of all growth rates in stage 1. In stage three (years 11 and onwards), the growth rate is the long-run nominal growth rate of the average U.S. economy. This long-run nominal growth rate is estimated by using the historical growth in real GDP and the long-run expected inflation rate.

13.4 Observed market price

13.4.1 Context and relevance

[338] The market price is an integral part of the DCF Model. The market price represents the present value of the firm's future earnings. The cost of capital is adjusted to equate future earnings to the present value of the firm ( i.e., its market price). Currently the Agency uses a randomly selected single daily closing price, usually selected on or near the submission date for the Western Grain cost of capital determination, of each railway company's common shares in the calculation of the DCF.

13.4.2 Position of CP

[339] CP is the only party that submitted a detailed methodology for implementing the DCF Model, including a discussion of the observed market price. CP proposes that the observed price should be defined as in the Morningstar/Ibbotson Model, the stock market value of CP's equity. CP uses the average of month end market capitalization for the first quarter of each year, which coincides with the release of its year-end financial statements.


  1. The Brattle Report, page 3.
  2. Brealey Myers, Principles of Corporate Finance, 7th edition, McGraw-Hill Irwin, 2003, page 534.
  3. Net investment in rail assets was originally created when the Agency, in 1997, ordered CN to convert from the "cash flow" method to the "balance sheet" method in determining the values required for the cost of capital calculations. It is a balancing figure representing all of the unidentified differences between the two methodologies, and is also used to reflect transactions between the company's rail division and its corporate head offices as per the Uniform Classification of Accounts (UCA).
  4. Damodaran, Aswath (2006). Damodaran on Valuation. New York: John Wiley & Sons. ISBN 0-471-75121-9.
  5. Surface Transportation Board, September 30, 2010. "Railroad Cost of Capital – 2009", Docket No. Ex Parte 558 (Sub-No. 13).
  6. This includes Burlington Northern-Santa Fe, CSX Transportation, Inc., Norfolk-Southern, and Union Pacific.
  7. Association of American Railroads, May 17, 2010. Submission for Railroad Cost of Capital – 2009. page 14.
  8. Uniform Classification of Accounts; Section 1701, page 95.
  9. Brattle Report, page 18.
  10. Surface Transportation Board, "Association of American Railroads – Petition Regarding Methodology for Determining Railroad Revenue Adequacy", October 23, 2008, Docket No. Ex Parte No. 679.
  11. CRTC Letter, October 22, 2004, "Use of an after-tax weighted average cost of capital (AT-WACC) in Phase II cost studies", as cited in the Brattle Report, page 74.
  12. Brattle Group, Responses to Submissions in the Canadian Transportation Agency's Cost of Capital Methodology Review.
  13. Surface Transportation Board, "Use of a multi-stage discounted cash flow model in determining the railroad industry's cost of capital", January 23, 2009 (Service Date - January 28, 2009), Docket No. ex parte no. 664 (sub-no.1).
  14. Ibid.
  15. Brattle Report, Table 3, page 103.
  16. In the 1984/1985 crop year, 70 percent of the rate to move grain was paid by the government, and only 30 percent by shippers. The subsidies dropped to about 50/50 by 1994/1995.
  17. Robert W. Faff and Usha Mittoo, "Capital Market Integration and Industrial Structure: The Case of Australia, Canada and the United States", Journal of Economic Integration, 2003, 18, 433-465.
  18. AUC 2009 Generic Decision: http://www.auc.ab.ca/applications/decisions/Decisions/2009/2009-216.pdf.
  19. Canadian Energy Law: http://www.canadianenergylaw.com/2009/05/articles/pipelines-and-storage/....
  20. CRTC Telecom Decision CRTC 98-2-1, Implementation of Price Cap Regulation and Related Issues.
  21. Revised Draft Decision ACCC - Sydney Airport Aeronautical Pricing Proposal, September 2000, cross referenced in Draft Decision ACCC - Australian Rail Track Corporation Access Undertaking, November 2001.
  22. STB, Ex Parte No. 664 Sub No. 0, January 17, 2008, page 7.
  23. STB Ex Parte No. 664 Sub 0, page 8
  24. Brattle Group, Responses to Submissions in the Canadian Transportation Agency's Cost of Capital Methodology Review, pages 7-8.
  25. Brattle Report, page 66.
  26. Brattle Report, page 62.
Date modified: