January Cost of Capital for PR09 A Final Report for Water UK

January 2009 Cost of Capital for PR09 A Final Report for Water UK

Project Team Dr Richard Hern Tomas Haug Anthony Legg Mark Robinson Contact Dr Richard Hern Ph: +44 (0)20 7659 8582 Fax: +44 (0)20 7659 8583 15 Stratford Place London W1C 1BE United Kingdom Tel: +44 20 7659 8500 Fax: +44 20 7659 8501 www.nera.com

Executive Summary In November 2009 Ofwat will set limits on the prices that water and sewerage companies in England and Wales can charge their customers during the five year period 2010-15. An important component of this review is the cost of capital. At PR04 Ofwat s approach to estimating the cost of capital relied on a range of evidence including the traditional CAPM, supplemented by dividend growth models, market to asset ratios and evidence from transactions involving water companies. In March 2008 Ofwat released Setting Price Limits for 2010-15: Framework and Approach ( SPL ), which indicated that its methodology for PR09 was likely to be quite similar to the approach at PR04. NERA prepared a cost of capital report for Water UK in June 2008 (henceforth referred to as the June report) 1 that was based on data up to the end of March 2008. In that report we calculated a range for the post-tax WACC of 4.4-4.9%. Since March 2008, conditions in the world s financial markets have deteriorated significantly and there is evidence that this has led to a re-pricing of risk in both the debt and equity markets. The financial turmoil has resulted in significant changes to cost of capital data, some of which may be short-term and some of which may be more structural. In this report we update our June report taking into account data up to the end of November 2008. Our revised post-tax cost of capital range is 4.6-5.1%. The increase in our cost of capital range is mainly due to a higher forward-looking cost of debt estimate than in the June report. We also revise upwards our ERP range to take account of the latest forward-looking data. Due to the recent volatile nature of international capital markets we recommend that this cost of capital range will need to be reconsidered closer to PR09 Price Determinations. Updated data during 2009 will assist in determining the likely long run impacts of the current financial crisis. We have not concluded on a point estimate of the cost of capital at this stage. In this report we have not considered the impact of the potential introduction of competition into the E & W water industry upon the cost of capital. We note, however, that investors perception of the potential opening of the market to competition will increase the cost of capital and will likely reduce the availability of funds. Cost of Equity We have relied primarily on evidence from the CAPM, taking into account that this is the model that Ofwat will likely rely on most heavily in its assessment of the cost of capital at PR09. We note that the Competition Commission has also recently concluded in the BAA airports quinquennial reviews that the CAPM remains the preferred tool for estimating the cost of equity. 2 1 2 NERA (2008) Cost of Capital for PR09, Final Report for Water UK. See Competition Commission (2008) Stansted airport Ltd: Q5 Price Control Review, 23 October, p92 and Competition Commission (2007) BAA Ltd - A report on the economic regulation of the London airports companies (Heathrow airport Ltd and Gatwick airport Ltd), 28 September 2007, p53. i

Although we place primary consideration on the CAPM, we recognise that standard approaches for estimating CAPM parameters use historical data, and therefore may not capture the full impact of changes in capital market conditions on the forward-looking equity returns that investors require. This is a particular concern with respect to the equity risk premium (ERP) given the current volatility in world capital markets. We deal with this issue in a number of ways in the report: First, we have cross-checked historical data on the ERP with forward-looking data on expected market returns for the stock market as a whole. The latest market evidence leads us to increase our central estimate of the ERP by comparison to our June 2008 report. Second, we have cross-checked the results from the CAPM with other market evidence. In particular, we have used evidence from forward-looking Dividend Growth Models (DGM) for the water sector. We find that cost of equity estimates based on the DGM are broadly consistent with our CAPM results. Taking into account all the available evidence, our analysis shows that the real post tax cost of equity for the E & W water sector is in a range of 7.2-8.6% using the CAPM based on a 60% gearing assumption. Estimating a DGM using 2008 data produces an average real cost of equity range for the industry of 7.4-8.2%. The lower end of this range is based on the assumption that dividend growth rates are equal to historic dividend growth and slightly less than long-term projected GDP growth. The upper-end of the range is based on a projected long-term growth trend equal to analysts forecast of near term growth. We note that DGM results for the individual companies lie within a wider range of 6.4-9.6%. However, since this report is concerned with the determination of an average industry wide cost of capital, we consider that this is best estimated using a central range reflecting the average DGM results. Averages across the four listed companies will also be less affected by market volatility or statistical error which might have affected any one company s estimates. To conclude on the overall real cost of equity, we use both CAPM and DGM. This provides additional comfort that our final recommendation on the industry wide cost of equity is not outside a reasonable range, which has particular merit during the current period of heightened market volatility. We conclude on the overlap of CAPM and DGM-derived cost of equity, which we consider an objective method to narrow down the possible range. This method produces a central range for the cost of equity of 7.4-8.2%. The results are shown in Table 1 below. ii

Table 1 Cost of Equity for PR09 (%) CAPM DGM Real Risk-free Rate 2.5 Equity Risk Premium 5.4 Gearing 60 60 Asset Beta 0.35 0.45 Equity Beta 0.88 1.13 Cost of Equity (real, post-tax) 7.2 8.6 7.4 8.2 Overlap Range 7.4 8.2 Source: NERA analysis The CAPM parameters shown in Table 1 above are derived as follows: Risk Free Rate In this report we show that all recent data on the UK risk-free rate is volatile and there is no perfect measure of the risk-free rate. Volatility in risk free rate measures has been especially marked since September 2008 following the collapse of major financial institutions, concern about deflation and uncertainty about the size of planned increased government borrowings. All of these factors have led to large swings in yields on securities such as gilts, government bonds and swaps that are used to measure the risk free rate. Due to the volatility in risk free rate measures post September 2008, we have attached little weight to risk free rate data during this period in deriving a cost of capital estimate for use at PR09. However, longer term trends in the risk free rate may become more apparent during the course of 2009 if financial markets stabilise. We have considered evidence from four different approaches to estimating the real risk-free rate: inflation-protected government bonds, deflated UK nominal government bonds, deflated international government bonds and deflated swap rates adjusted for inter-bank risk. All four approaches are theoretically appropriate methods for estimating the real risk-free rate. In summary, our analysis shows the following: Inflation Linked Gilt (ILG) yields are a biased measure of the risk-free rate, especially for long-dated maturities, and have been for around the last 10 years. A major cause of this is inelastic demand by pension funds for inflation-protected bonds, due to changes in the regulatory and accounting framework which encourage the holding of these types of assets (such as FRS19 and IAS17). UK nominal gilts are more liquid than the IL market, and demand for these assets is less affected by financial regulations. However, UK nominal bond yields are highly volatile and there is some evidence that there has recently been a bubble in UK nominal bond prices due to large numbers of investors chasing limited supply. This may change during 2009 if more nominal debt is issued to finance the fiscal stimulus. Evidence that UK nominal bonds yields are not a perfect proxy for the UK risk free rate is confirmed by a comparison between UK nominal yields and international iii

government bond yields. When all yields are deflated for expected inflation, the UK yields lie markedly below international bond yields. Moreover the UK yield curve is highly inverted whereas other bond yield curves such as the US and Germany are upward sloping (as theory predicts). International evidence on deflated nominal bond yields shows a risk free rate of 2.2-2.8%. Many recent papers in the finance literature show that the risk-free rate is better proxied by the swap rate than government bond rates for all maturities. This is because the swap market is a much more liquid market than the government bond market and swap rates are not distorted by financial regulations that impact on demand for government bonds. Swap rates, however, do include an inter-bank lending risk premium that should not be included in a measure of the risk-free rate. After adjusting for this premium estimates of the risk-free rate using a swap based approach average around 2.5% over the past ten years, and the 3-month estimate up to September 2008 is also 2.5%. Taking into account the evidence from all four approaches, our estimate of the real risk free rate to be used in the CAPM is 2.5%. This estimate places most weight on time series data on UK swap rates up to September 2008. However, we have cross checked this estimate with international data on government bond yields which shows a range of 2.2%-2.8%. We attach little weight to UK ILG yields and UK nominal bond yields in our conclusions due to liquidity concerns with these securities expressed above. We note that the Competition Commission still prefer to rely on risk-free rate measures using Index Linked Gilts (ILG), despite noting that yields on longer-dated yields are biased downwards due to pension fund demands driven by accounting regulations such as FRS17 and IAS19. 3 Using this methodology, the Competition Commission recently estimated a risk-free rate of 2.0% in the Stansted airport review based on data up to 12 September. However, if the CC s methodology was repeated using the very latest data up to November 2008, it would lead to a higher conclusion. Yields on UK ILGs have increased dramatically since the CC s Stansted recommendation. The 3-month historical average of 5-year yields was 2.6% at the end of November, compared with 1.6% as cited in Stansted (as at 12 September). The spot rate on a 5-year ILG was as high as 4.2% at the end of November (its highest level since 1992). This update shows that ILG data are highly volatile and strong conclusions should not be drawn on the regulatory risk-free rate based on this data alone. Notwithstanding this, the recent rise in ILG yields is supportive of our overall conclusion on a central risk-free rate of 2.5%. 3 The effect of these distortions is to depress observed yields on the affected range of bonds below the true risk free rate by the amount that pension funds are willing to pay to meet their legal obligations. For this reason, UK ILGs do not provide an accurate measure of the real risk free rate for estimating the cost of capital. A number of academics have noted these problems. For example, Stephen Schaefer (2008), Finance Professor at the LBS, stated in a recent submission to the BAA airports inquiry: If the yields on Treasury bonds are lower than swap rates primarily because Treasuries provide liquidity benefits then, on the (reasonable) assumption that equities do not themselves provide the same liquidity benefits as Treasuries, the swap rate may be a better measure of the risk-free rate than the Treasury rate when applying the CAPM to estimate the required return on equity. iv

In reaching our conclusion on a risk-free rate of 2.5%, we also note that Ofwat s range for the risk-free rate at PR04 was 2.5-3.0%, and that previous UK regulatory estimates of the risk-free rate generally lie in the range of 2.25%-2.75%. Regulators should be mindful of the need for regulatory consistency. Given the volatility in international capital markets relying too heavily on very short term data on imperfect proxies for a riskless security, as the Competition Commission has done in the Stansted review, injects an unnecessary degree of regulatory risk into the price control process. There is no substantial body of evidence to justify the Competition Commission s significant departure in the Stansted review from previous regulatory decisions on the risk-free rate. Equity Risk Premium In our June report we presented the ERP as a range of 4.2-5.4%, based on geometric and arithmetic averages of long run historical equity returns. As noted in our June report and confirmed by further analysis subsequently the geometric average has very little support amongst the academic community as the correct measure of the ERP for UK regulatory purposes. 4 The single statistic that is favoured by the vast majority of finance academics is the arithmetic average of historical returns which estimates the ERP at 5.4% for the UK. The reason why the arithmetic mean is the preferred measure of the ERP is because it represents the mean of all the returns that may possibly occur over the investment holding period. The arithmetic mean, therefore, explicitly accounts for the volatility of equity market growth. In contrast, the geometric average assumes that growth is constant and perfectly predictable. This is clearly an unrealistic assumption. One concern with the use of long-run historical data on the ERP for regulation is if prospective capital market conditions over the regulatory period are unrepresentative of normal economic conditions. If this is the case, then some adjustment to the historical data may be warranted. We have therefore cross-checked historical data on the ERP against the latest forward-looking data on expected market returns for the stock market as a whole. Real cost of equity estimates are obtained by applying a DGM to the FTSE 100. Over the course of 2008, the results show ERP estimates of 5-7% for 2008 based on a range of plausible long-term growth assumptions. These results are, therefore, broadly consistent with long run arithmetic (but not geometric) averages of historical returns. However, the analysis also shows that there is a strong increasing trend in the ERP estimate in recent months, which reflects the increased volatility and financial uncertainty that have led investors to demand greater compensation for risk. 5 We note that our estimate of the ERP of 5.4% lies outside the CC s ERP range for Stansted of 3-5%. In this report we describe a number of reasons why the CC s range is too low that include selective interpretation of evidence and flaws in the application of the 4 5 For instance, even the most prestigious former advocates of the geometric mean (Copeland et al (1990) p196) now state that the arithmetic average is the best estimate of future expected returns (Copeland et al (2000) p219). This effect has recently been noted by (among others) the Bank of England, which suggested that volatility was the reason for a rise in their DGM based ERP estimates. Bank of England (2008), Quarterly Bulletin, Q1, p8. v

DGM. The CC s range is also inconsistent with current forward-looking evidence on the ERP. In light of recent increased stock market volatility and uncertainty, which may prevail throughout the next price review, we believe it would be highly inappropriate for Ofwat to adopt the CC s ERP range, which is an outlier even in times of more normal market conditions. Asset Betas We show that asset betas for listed E & W water companies lie in the range of 0.35-0.45 over both the short and long-term. This range includes the 0.39 figure recommended in our June report. In this updated report we have decided to present our beta estimates as a range reflecting the fluctuation in this parameter over time and across companies, and in recognition that different companies have different risk profiles. DGM Results Our DGM results are based on 2008 market data (excluding December). Dividend growth forecasts reflect analysts dividend forecasts over the near term and our assumption of different plausible projected long-term trend growth rates. We base our DGM-derived cost of equity estimates on a range of long-term dividend growth assumptions, given the inherent uncertainty in determining this estimate. The average DGM-based cost of equity range at a notional gearing ratio of 60% is 7.4-8.2%. The lower-end of the range is based on a long-term trend growth of dividends which is equal to historic dividend growth and slightly less than long-term projected GDP growth. The upper-end of the range is based on a projected long-term growth trend equal to analysts forecast of near term growth. The conclusions from the DGM largely coincide with the latest CAPM range based on our revised ERP of 5.4%. This coincidence confirms the plausibility of our ERP value in preference to the lower ERP range presented by the Competition Commission. Cost of Debt and Gearing Our analysis presents data on both historic time series data on debt costs and current debt costs for different ratings. Time series evidence shows that average real debt costs for A- rated debt have been in the range of 2.6-3.8% across a range of debt instruments over a ten year historic period. However, recent evidence based on new debt issues of A- debt shows average real costs in the range of 2.5-5.4% over the 3-month period to the end of November 2008. In addition, our analysis suggests that the transaction and pre-funding costs associated with this new debt are currently around 60 bps. In concluding on the cost of debt we recommend that the allowed cost of debt is the weighted average of the following two components: Long-term time series evidence on the cost of debt for the proportion of debt the sector will not refinance over AMP5; and Current evidence on the cost of debt for the proportion of new debt (i.e. refinancing of existing debt and new debt to finance AMP5). vi

The weights to attach to the historic and forward-looking debt costs should be determined by a detailed analysis of company financing and re-financing requirements over AMP5. This approach takes account of the fact that the industry may have raised finance efficiently at different parts of the interest rate cycle, but that the forward-looking cost of debt may be very different. 6 Overall, based on evidence up to November 2008, after allowing for transaction and pre-funding costs, we recommend that the allowed real cost of debt at PR09 should be in the range of 3.8-4.3% for A- rated debt. This range is a significant increase on our recommended range of 3.4-4.0% in our June report and reflects the fact that access to new debt finance has subsequently become markedly more expensive. Cost of Capital Taking into account all the available evidence, our best estimate of the real (fully) post-tax cost of capital for the E & W water sector at the current time is a range of 4.6-5.1%. Table 2 Cost of Capital for E & W Water Sector for PR09 (%) Estimate Gearing 60 Real Pre-Tax Cost of Debt 3.8 4.3 Real Post-Tax Cost of Equity 7.4 8.2 Tax Rate 28 Pre-tax WACC 6.4 7.1 Vanilla WACC (Pre-tax debt, Post-tax equity) 5.3 5.8 Post-tax WACC 4.6 5.1 Source: NERA analysis We have not concluded on a point estimate of the cost of capital at this stage. This is due to the recent volatile nature of international capital markets which we recommend means that this cost of capital range needs to be reconsidered closer to PR09 Price Determinations. All aspects of the cost of capital will need to be reconsidered in light of developments in 2009, but in particular, the forward-looking cost of debt, ERP, DGM and risk-free rate estimates should be updated to reflect additional market data. Our cost of capital range of 4.6-5.1% is broadly in line with Ofwat s allowed cost of capital at PR04 of 5.1% before small company premium and financeability adjustments. Ofwat s 4.3% cost of debt allowance at PR04 lies at the top end of our 3.8-4.3% range, which reflects evidence on both the cost of raising new debt and the benchmark cost of existing water company debt. However, we note that the cost of new debt achievable by water 6 We note that the Competition Commission for Stansted (November 2008) has adopted a very similar approach where the cost of debt was weighted by the cost of existing debt cost and forward looking debt costs, taking into account the maturity profile of existing debt and the funding requirement of new capex. vii

companies has increased substantially in recent months and the debt markets will need to be monitored closely in the lead-up to PR09 in order to ensure that the allowed cost of capital does provide the right investment incentives at the margin. Ofwat s PR04 7.7% allowed return on equity is close to the middle of our 7.4-8.2% range, though this comparison is at least partially influenced by our slightly higher gearing assumption of 60% (compared to 55% by Ofwat at PR04). We note that there is some evidence from the most recent Water UK Investor Survey (March 2008) that certain risks (political, regulatory, force majeure and management) are perceived to have increased since PR04. Against this, there are some elements of the proposed PR09 methodology (revenue cap, symmetric treatment of capex overspends) that might be expected to reduce risk. Overall, it is important for Ofwat to fully justify the allowed rate of return at PR09 based on objective evidence. A simple benchmarking analysis drawing upon recent regulatory decisions to inform the cost of capital parameters would be an unsound approach to setting the allowed rate of return. In particular, Ofwat should not simply read-across from recent decisions such as those by the Competition Commission for Stansted or the Office of the Rail Regulator s decision for Network Rail as these decisions have not been market tested, differences in other aspects of the regulatory settlement may bias a direct comparison of the cost of capital and market conditions have changed dramatically since the analyses upon which these decisions were based were undertaken. Financeability In this report we have retained our assumption that the overall regulatory package (including the cost of capital) must allow companies to maintain an A- credit rating (or better), and we note that access to debt at ratings below A- has become markedly more expensive in recent months. This means that it is critical that Ofwat ensure that companies have sufficient headroom in their projected financial ratios to be able to raise finance at minimum A- ratings in both central case and plausible downside case scenarios. Rather than focussing solely on the K in the central case, financeability analysis should also aim to establish the range and likelihood of possible outcomes by financial modelling of scenarios, sensitivity and stress testing, and risk analysis designed to derive key financial measure statistics (best, worst, expected outcome, probability that the outcome is outside investment grade positions, risk of bankruptcy). This is so that the full implications of particular K s can be seen. In the current economic circumstances it is critical that this stress testing consider the potential impact of deflation upon companies financial ratios and the implications for credit ratings. Further, as part of this analysis the availability of market-based solutions to financeability should also be considered: our analysis shows that index-linked debt and new equity issuance are only likely to be available on a limited basis over AMP5. Evidence on MARs for listed WaSCs During the first half of AMP4, market values of regulated assets of listed WaSCs traded at premiums to their RCVs. At the time, Ofwat was concerned that observed MARs of significantly greater than 1.0 could be a sign of regulatory generosity with respect to the allowed rate of return. viii

In this report, we have updated our MAR analysis taking into account data up to the beginning of December 2008. Our analysis shows that the industry aggregate MAR has decreased since March 2008 from around 1.2 to around 1.0 in November 2008. The average aggregate MAR over AMP4 is around 1.1. However, we believe that we cannot attach significant weight to the informational value of MARs for the true cost of capital for the following reasons: First, the observed MAR range of 1.0-1.2 is imprecise and requires an estimate to be made of the valuation of the non-regulated business which is surrounded by considerable uncertainty; Second, the portion of the MAR that is explained by other factors, such as expected out-performance is also imprecise. We have investigated what investors assume over AMP4 and beyond, but there is little objective data available to support this; Third, in trying to back out what a particular value of MAR means for the market cost of capital, a further assumption needs to be made about what cost of capital investors assume Ofwat will set at future price reviews. We believe the most plausible assumption that investors make is that any wedge between the market cost of capital and allowed rate of returns will only be temporary and that Ofwat will correct for any difference at future price reviews. If this is the case, MARs observed over AMP4 could imply a market cost of capital (real post-tax WACC) in the wide range of 2.1-8.6%. This range is too wide to infer any meaningful estimate for the cost of capital at PR09. In summary, the MAR evidence over AMP4 implies a very wide range for the cost of capital. After adjusting for out-performance, our estimated range for the (adjusted) MAR averaged over AMP4 is almost symmetric around 1 which shows that Ofwat s allowed rate of return at PR04 at 5.1% is within the plausible range for the cost of capital. However, the range is too wide to make a judgement with sufficient confidence on the implied cost of capital required by investors. December Update Although the WACC estimates in this report are based on data up to November 2008, we also include a chapter in this report that discusses the very latest debt market data in December 2008 that has become available to us during the finalisation of the report. We note that: Yields on A rated bonds traded in the secondary market were around 6.75% at the end of December (down from a peak of 7.5% in mid-november). The corresponding BBB yield was 8.7% (from a peak of 9.2%). December saw a small number of new bond issues in the UK sterling market with a wide difference in nominal yields from 6.1% (United Utilities) to 9.4% across A rated bonds of medium to long term maturities. Short term inflation expectations fell significantly during December. Overall, our analysis suggests that the real cost of new bond debt in the month of December is in a range of 4.7-6.3% for A- rated debt. This is similar to the real cost of new bond debt during the 3-months previously. We have therefore not adjusted our regulatory WACC estimate for this latest data. ix

Introduction We do emphasise, however, that real costs of debt in the range of 4.7-6.3% are high relative to historical levels. As with other cost of capital parameters, further evidence on the cost of new bond market debt will need to be reconsidered closer to PR09 Price Determinations. x