August 7, 2007 File no: R-3630-2007 IGUA’S response to Information Request #1 by the Régie to IGUA August 7, 2007 File #: R-3630-2007 Information Request #1 by the Régie to IGUA Page 1 of 3 INFORMATION REQUEST #1 FROM THE RÉGIE DE L’ÉNERGIE (“LA RÉGIE” ) ON SUBJECTS RELATING TO THE HEARING REGARDING A REQUEST TO MODIFY G AZ M ÉTRO, LIMITED PARTNERSHIP (“GAZ MÉTRO” ) RATES STARTING O CTOBER 1, 2007 1. Reference: Fair Return for Gaz Métro, Evidence of Laurence Booth, July 2007, Preamble: Fama-French model. Questions: 1.1 What is your opinion with respect to applying the Fama-French model in the context of a regulated company. Dr. Booth believes that the Fama-French model is not accepted widely enough to be used in a current regulatory hearing. The Fama-French model is one variation on a general class of factor models. The genesis of all these models is Ross’s arbitrage pricing theory (APT) that indicated that in an arbitrage free capital market under quite loose conditions expected returns would be linearly related to the risk premiums on externally specified risk factors. The Achilles heel to these models is the specification of the factors. Roll and Ross (JF 1980) specified the APT in terms of macro-economic variables such as industrial production, changes in expected and unexpected inflation and default and yield premia, whereas others have used factor analysis to see what factors returns “load on.”. However, William Sharpe, inventor of the CAPM, for which he received a Nobel prize, responded that such APT models could always be given a multi-beta interpretation consistent with the CAPM. This literature then died down as the focus moved to explaining anomalies that could not be explained by the CAPM. Many of these anomalies revolved around the fact that “high” priced firms subsequently earned lower rates of return. Anomalies such as the PE effect, the dividend yield effect, the over-reaction effect all involve the firm’s existing price. Another anomaly was the size effect, that institutions tend to be restricted to investing in large cap stocks and thus their expected rates of return tend to be lower since their prices have been bid up. In contrast, small cap stocks tend to be neglected by large institutions and there is less public information on them, so they tend to earn higher rates of return. The Fama-French model codifies these anomalies as risk factors, since in addition to the market return they add a size factor and a factor that includes the market price (the inverse of the market to book ratio). Professor Fama is the “high prince” of market efficiency and many have interpreted the Fama-French model as a way of interpreting anomalies to market efficiency as “risk factors” consistent with market efficiency. The Fama-French model is thus very controversial amongst finance academics. Those who do not accept the (University of) Chicago view of market efficiency simply see the Fama-French model as an example of “data snooping,” that is, they have spun the security return tapes enough times to find factors that work in explaining returns. Others who do not accept the extreme Chicago view on market efficiency see these same factors as evidence of market inefficiency. The problem is that there are only limited economic foundations for two of the three risk factors in the Fama-French model. The size variable very much depends on very small firms during the sample period, while the impact of the market to book ratio is the opposite to what intuition indicates. The normal intuition is that firms with high market to book ratios, like Google, are riskier because the high valuation reflects the existence of future excess profits, which by their nature are risky due to competition. In contrast firms with low market to book ratios, like utilities, are less risky since there are no excess profits to be competed away. Instead the Fama-French model implies that firms with high market to book ratios are lower risk, while utilities with low market to books are high risk, which doesn’t make any sense. Until there is a well-accepted economic foundation for the Fama-French factors, Dr. Booth believes it is a classic case of fitting an ad-hoc model to the data without any justification. Further, since the Fama-French model was developed there has emerged a growth industry in factor models as factors such as momentum, skewness and liquidity have all been put forward as “risk” factors. In Dr. Booth’s judgment these factor models are at too primitive a stage to be used in regulatory hearing and to the best of his knowledge none has ever been accepted or even presented before in a Canadian regulatory environment. However, the Regie should be aware of some of the implications of Dr. Chretien’s work using the Fama-French multi-factor model. In his three Canadian indexes, both for the Dow Jones Canada Gas Distribution index and the Datastream index, Enbridge is the most important component since these indexes are weighted by their market value. This is why the correlation with Enbridge is so high. Apart from this high correlation, the other Canadian utilities clearly have the best correlation with his sample of Canadian utilities. In contrast, the correlation with the US indexes is so low that for some of them they could in reality be zero. This confirms the importance of risk measures relative to a Canadian market index and that the only sample that is relevant is the sample of Canadian utilities. In Dr. Chretien’s Fama-French estimates in Table 2 (page 39) the beta coefficient on the market for his Canadian utility sample is 0.484, which approximately confirms the lower end of Dr. Booth’s beta range of 0.45. In his adjusted CAPM estimates in Table 3 on page 48 the beta estimates for the Canadian samples are 0.560, which confirms the high end of Dr. Booth’s beta range of 0.55. Without accepting his models, Dr. Chretien’s statistical work approximately confirms the range of Dr. booth’s own estimates for Canadian utilities of 0.45-0.55. The risk premiums from Dr. Chretien’s Fama-French estimates for his Canadian utility sample is 4.232%, which is HALF his estimates for the US sample confirming again the low risk nature of Canadian utility holding companies relative to those in the US and the inadvisability of using evidence from the US for Canadian companies. If this risk premium is added to the 4.3% forecast long Canada rate and 0.30% added for issues costs then the resulting estimate of 8.83% is only 0.28% higher than the Regie’s formula allowed ROE. In Dr. Chretien’s adjusted CAPM the risk premium for the Canadian utilities in Table 3 is 4.884%, which with a 4.3% long Canada rate and 0.30% for issue costs gives an ROE of 9.48%. However, this only comes about due to a 1.39% “error” correction, so without this the estimate is 8.09% lower than the Regie’s formula allows. Of note is that the “error” correction is essentially an adjustment for the fact that realised returns have been higher than expected due to interest rate declines in Canada. Whether this has any relevance for future returns is doubtful. To be consistent with his empirical testing and estimates, both Dr. Chretien’s Fama-French and Adjusted CAPM models should be used with the treasury bill or short term interest rate as the risk-free rate. Instead Dr. Chretien estimates the parameters using the short-term rate and then applies the models using his modified risk-free rate, based on long Canada yields. This is clearly a contradiction. Since long Canada yields are normally higher than the short term rate Dr. Chretien essentially double counts the adjustment to the CAPM and artificially inflates his estimates. These comments on Dr. Chretien’s implementation of the Fama-French and Adjusted CAPM models simply highlight the problems in using these novel models. The fact is that the practise of finance is to use a CAPM with the long Canada bond yield as the riskfree rate and an adjusted beta as the risk coefficient. These adjustments along with a market risk premium of 5.0% are used daily by the large investment banks in Canada in valuing firms in research reports, since they have stood the test of time. In contrast, the Fama-French and adjusted CAPM models so far have not. 2. References: i) Exhibit Gaz Métro-7, Document 8, pages 24 et 25. ii) Exhibit Gaz Métro-7, Document 8.20, page 1. Preamble: Modified risk-free rate proposed by Gaz Métro Questions: 2.1 What is your opinion about Dr. Chrétien’s proposal which recommends establishing the mean level of long term rates at 6.41% instead of 5.76% in order to better reflect the average conditions of 30 Canadian year rates in the formula for setting the risk-free rate. 2.2 Please make the link between the risk-free rate anchorage point to a given year and the effect on the return on shareholders’ equity of a regulated company. 2.1 Dr. Booth has no problem with a recalibration of the adjustment mechanism. This can be based on any interest rate as long as the resulting estimate of the ROE is fair. Since the Regie considered the results of the adjustment model fair in 1999, any new model has to be consistent with this decision. Otherwise, explicitly the Regie’s 1999 decision is thought to be unfair. For example, in document Gaz Metro 7 10.1 the fair ROE for 1998, and what it would be now, was described as follows: Risk free rate: Risk premium (0.55 beta*market risk premium of 6.5%) Issue costs Fair ROE in 1999 Change in risk free rate (5.76-4.30) Change in fair ROE 75% of change in risk-free rate Fair ROE 5.76% 3.58% 0.30% 9.64% -1.46% -1.09% 8.55% If Dr. Chretien’s recommendation for the risk-free rate is accepted then the risk premium for 1999 has to be altered accordingly, so that the fair ROE in 1999 is still 9.64%: 1999 Modified Risk free rate: 75% of 1999 rate of 5.76%: 25% of long run rate of 6.41%: Risk premium Issue costs Fair ROE in 1999 Current modified risk-free rate: 75% of current rate of 4.30%: 25% of long run rate of 6.41%: Risk premium Issue costs Fair ROE 5.92% 4.32% 1.60% 3.42% 0.30% 9.64% 4.83 3.23% 1.60% 3.42% 0.30% 8.55% In this way the smaller risk premium for 1999 of 3.42% is consistent with the motivation of the adjustment mechanism that the size of the market risk premium varies inversely with the level of interest rates. As a result, the 0.65% higher riskfree rate of 6.41% is offset by a 0.65% smaller risk premium of 3.42%, but the fair ROE remains that set by the Regie in 1999 of 9.64% If this adjustment to Dr. Chretien’s formula is made, the recommendation for the fair ROE for 2008 would be identical between the Regie’s existing formula and that proposed by Dr. Chretien. However, what Dr. Chretien seems to have in mind is the simple proposition that the Regie’s fair ROE in 1999 was too low and he seems to envisage using the same risk premium of 3.58% (+ issue costs) the Regie found to be fair in 1999 combined with his higher (modified) risk free rate. My understanding is that Dr. Chretien envisages the following:1 1999 Modified Risk free rate: 75% of 1999 rate of 5.76%: 25% of long run rate of 6.41%: Risk premium Issue costs Fair ROE in 1999 Current modified risk-free rate: 75% of current rate of 4.30%: 25% of long run rate of 6.41%: Risk premium Issue costs Fair ROE 5.92% 4.32% 1.60% 3.58% 0.30% 9.80% 4.83 3.23% 1.60% 3.58% 0.30% 8.71% Note that the fair ROE for both 1999 and now is higher by 0.16% as a result of Dr. Chretien’s modified risk-free rate. The fact is that his change to the Regie’s formula has nothing to do with interest rate modelling, instead it is simply a way of increasing the formula ROE by 0.16%. 2.2 1 In any risk-based model the “anchor” is the current risk-free rate. This holds whether the risk-based model is the CAPM, a multi-factor model or the arbitrage pricing model. This is because we are explicitly breaking the fair rate of return into time and risk value of money components. Consequently the use of anything other than the current risk-free rate is extremely difficult to understand. As is explained in answer to 2.1 it is always possible to substitute a “modified” risk-free rate, for the current risk-free rate, in a formula adjustment for the fair ROE, as long as the risk premium is adjusted accordingly. If the risk premium is not adjusted, as it is not in Dr. Chretien’s model, the resulting ROE estimate will not be fair. However, why anyone would want to use such a modified risk-free rate model is difficult to understand. This is based on his explicit examples in footnotes 18& 19 on page 25 of his testimony. In Dr. Booth’s judgment the use of this modified risk-free rate by Dr. Chretien results from a misunderstanding on his part as to why adjustment mechanisms have been adopted. In his testimony Dr Chretien refers to “interest rate modeling” and the fact that “The Regie has (indirectly) selected an average towards which the expected long-term risk-free rate must converge that equals the expected risk-free rate for 1999.” This interpretation is totally incorrect. Even if we accept his interpretation of the adjustment mechanism, there is nothing that causes the long-term risk-free rate to converge to anything. However, of more importance is that regulatory boards have not based the adoption of ROE adjustment mechanisms on any interest rate modelling. Dr. Booth participated in the original BCUC and NEB hearings that resulted in the adoption of their adjustment mechanisms, as well the initial applications in Ontario and Manitoba. In all cases the focus was on how the risk premium varied with the level of interest rates. The specific empirical regularity that justified this was the observation that risk premiums varied inversely with the level of interest rates. In the period of high interest rates in the late 1970’s into the mid 1990s utility risk premiums were very “low” and they subsequently increased as interest rates came down. As Dr. Booth discusses in his Appendix F bond market risk was very high at this time, so that utility risk premiums over long bond yields should have been very low. Consequently Dr. Booth judges both the economic and regulatory foundations for the adjustment mechanism to be the inverse relationship between interest rates and risk premiums. The use of ROE adjustment models doesn’t have anything to do with modified risk-free rates or interest rate modeling as envisaged by Dr. Chretien. This is a strange interpretation. 3. Reference: Fair return for Gaz Métro, Evidence of Laurence Booth, July 2007, page 43. Preamble: “In this case the BCUC felt that Terasen’s RSAM deferral 2 account was worth 0-3% on its common equity ratio. However, Gaz Metro is also deemed a 7.5% preferred equity ratio for a total equity ratio of 46% one of the largest of any Canadian utility since almost all Canadian utilities have been retiring their preferred shares. This is because of recent accounting changes that cause debt like preferred shares with hard retractions to be treated like debt for reporting purposes regardless of their legal characteristics. With a 46% equity ratio Gaz Metro has a very large offset to any remaining risk differences between it and the two large Ontario Gas LDCs.” Questions: 3.1 Please elaborate further on what treatment the Régie should consider for preferred shares. =========================================================== This depends entirely on what sort of preferred shares the Regie deems for Gaz Metro. Regulated utilities have used preferred shares since they were equity for reporting purposes but acted like debt. In this way they increased coverage ratios and acted as intermediate financing to help firms gain access when their coverage ratios may otherwise have restricted access. However, if the preferred shares are structured like debt where they are paid off in a short time period (5 years or so) then for reporting purposes they are now treated like debt. These accounting changes and the repeal of the PUITTA, which rebated the income tax paid ahead of preferred shares, reduced their attractiveness so that most utilities have been redeeming their preferred shares and replacing them mainly with debt. Since Gaz Metro finances with first mortgage bonds and faces no access problems in terms of having to maintain a particular interest coverage ratio to issue debt, there is no reason for the preferred shares. Consequently Dr. Booth would recommend that they be treated as debt. 4. Reference: Fair return for Gaz Métro, Evidence of Laurence Booth, July 2007, pages 3, 6, 7 and 70. Preamble: page 3 “The generosity of the current allowed ROE to Gaz Metro is also indicated by the fact that on October 10, 2006 GMi sold 2,9133,753 units of Gaz Metro for $17.16 a unit and recognized a gain of $16.70 million on the $50 million in proceeds. Again, this sale was at a significant premium to book value indicating that the allowed ROE is above the investor’s required rate of return or fair ROE.” pages 6 and 7 “Further as of September 2006 the book value of GMLP was $7.87 so at $16.84 GMLP was selling for over 2X book value. I will discuss the importance of market to book ratios as signals to regulators of the fair ROE later, but the fact that GMLP is selling at such a premium to book value indicates that the return earned by GMLP is in excess of what 1 investors require. From this market data on GMLP it is clear that the fair ROE is significantly less than the 10.19% estimate of Dr. Chretien on behalf of Gaz Metro.” page 70 “The objective of regulation is to treat investors fairly. This is accomplished by awarding a fair return such that the share price should only increase by the amount of earnings retained within the firm and not paid out as a dividend. If a utility paid out 100% of its earnings as a dividend, the share price should approximate its book value, as long as it continues to be awarded its fair return. » Questions: 4.1 Could you please be more precise as to your opinion, in the reference on page 70, with respect to the link you establish between the company’s distribution policy, the share or unit price and the book value. 4.2 Please elaborate as to the conclusions you present in the references on pages 3, 6 and 7. 4.1 If investors capitalize a utility with $11, but $1 is taken in investment banking fees and issue costs etc so that $10 is in the rate base, and the opportunity cost is 10% then they expect to earn $1.1 on the $11 investment or 11% on the $10 net to the company. If this fair treatment is expected to continue forever, then the stock would sell at a market to book of 1.1X. In this way a market to book of 1.1X reflects the dilution or issue costs that are incurred to raise capital. Absent any such costs the market to book would be 1.0. If the fair ROE then drops to 8%, but the regulator does not respond, assuming again that this is expected to continue forever, the stock price would increase to $13.75 and the market to book ratio becomes 1.375 indicating that the allowed ROE is excessive and should fall. Consequently effective regulation implies that the market to book ratio should be a small premium over 1.0. Of course the regulator cannot control market prices, but a persistent pattern of a utility’s stock price being significantly in excess of 1.0 indicates that allowed ROEs are generous and should be cut. As the Alberta EUB said (Decision U99099, Vol 1, page 303) “The Board would be derelict in its statutory responsibilities to recognize market capitalization ratios that are derived from a market value capitalization that deviates from the intrinsic long-run value of the regulated firm.” The EUB recognized that market values should be close to book values. This proposition is clear when 100% of earnings are paid out as dividends, since the stock can be simply valued as a perpetuity, since there is no growth. Slightly more assumptions have to be made to get the same result when a significant amount of earnings are reinvested within the firm. However, the general result still holds that market to book ratios for regulated firms indicate how satisfied investors are with the allowed ROE. In this sense it parallels results for the bond market where bonds selling at a premium to par value (market to book above 1.0) indicate that interest rates (fair returns) have fallen. These general ideas were expounded by Dr. Booth in an article in the NRRI Bulletin, “The Importance of Market to Book Ratios in Regulation, which is attached as Booth NRRI paper 1997.pdf 4.2 Dr. Booth believes that the fact that GMLP units sell significantly above book value indicates that the ROE earned by GMIP is higher than a fair ROE. Similarly the sale of utility assets at well above book value to other utilities indicates that their allowed ROEs are generous. Dr. Booth also notes that GMIP’s ROE seems to contain an income tax component which inflates the ROE above the regulated level. 4.3 5. Reference: Fair Return for Gaz Métro, Evidence of Laurence Booth, July 2007, Appendix E, Schedule 1, page 14 and Appendix F, Schedules 1 and 2. Preamble: Data series and sources Questions: 5.1 Please file in Excel format the data series as well as their sources. This is contained as Booth-Regie 1.xls