Professional Documents
Culture Documents
by Mike Lockett
he cost of capital is one of the factors Debt capital: investors lend money to the
equity capital; debt can be secured against the In addition to being cheaper than equity,
firm’s assets-similar to a household debt capital offers the advantage of being tax
mortgage-thus reducing risk for the lender efficient. Current UK tax law allows companies
and keeping interest rates down. Equity capital to make debt interest payments before calcu-
providers can make unlimited returns if the lating taxable profits. With a current UK
firm prospers, but risk losing their entire corporate tax rate, t , of :30%,every pound of
investment if the firm folds, and consequently interest paid out reduces the company’s tax bill
they demand a higher interest rate to by 30 pence. This reduction is known as the
compensate for this higher risk. A company’s ‘debt tax shield, and reduces the debt interest
cost of capital can be derived by taking the rate by a factor of (1 - t ) .
weighted average of the cost of debt and the The weighted average cost of capital can be
cost of equity to arrive at the weighted average expressed as:
cost of capital (WACC).
‘Gearing’ describes the ratio of debt to total
capital employed: g = D/(D + E), where D and E
are, respectively, the total value of debt and where:
equity capital used by the company. Gearing
kd = cost of debt-the debt interest rate
describes the ‘capital structure’ of a company.
A company with 50% gearing funds its k, = cost of equity-the equity interest rate
operations half with debt capital and half with t = corporate tax rate
equity capital. g = gearing, D/(D + E)
1 Effect of gearing on
returns to equity capital
providers
(iii) From the capital market line to the expected return, = risk-jree rate +
security market line-beta (beta,,,,x ERP) (4)
William Sharpe3 was the first to develop and
use the concept of equity beta: the relative Sharpe effectively redefined the risk-return
responsiveness of the return of an individual relationship: beta is used to measure risk, and
share compared to the market portfolio’s expected return is a linear function of beta. The
return. Beta can be calculated by regressing the result is known as the security market line
historical returns of an individual share against (SML) and can be represented graphically (see
the market portfolio’s historical returns (five Fig. 7).
years of monthly data is usually used, see The return on a high beta share (beta > 1)
Fig. 6). demonstrates high volatility-a 1%change in
Beta can be expressed mathematically as: the market portfolio return results, on average,
in a greater than 1% change in the individ-
ual share’s return-consequently investors
demand a higher expected return than the
market portfolio. Vice versa for a low beta share
where p is covariance, (s’ is variance, subscript (beta < 1).
i denotes the individual share, and m denotes The regression process used to determine
7 Security market line beta eliminates movements in the individual Dividend growth model
share’s return caused by unique events (the (i) Theory
scatter of the data points in Fig. 6). Beta is The dividend growth model (DGM) is the most
therefore a measure of the non-diversifiaiable or widely used method in US regulated industries,
systematic risk of the individual share relative and is used in support of the CAPM by many
to the market portfolio. The expected return UK regulators. It is simple to understand and
figure generated by eqn. 3 only compensates apply. However, its bold assumptions about
the investor for the systematic risk associated constant growth, and the numerous methods
with the individual share. This is acceptable of estimating growth rates produce cost of
provided that the investor combines the share equity estimates that are no more than a rule of
in a market portfolio, thus eliminating the thumb.
share’s unsystematic risk. In 1962 Gordon (see Reference 2) showed
that, if the company’s dividends, D, grow at a
(iv) Problems with CAPM constant rate to infinity, gd,”,then the current
CAPM and the theories on which it is based share price, PO,can be presented as:
(portfolio theory and the separation theorem)
require a number of assumptions to be made:
(5)
no taxes
risk-averse investors Numerical subscripts denote time in years.
no transaction costs when buying and selling Rearranging gives:
shares
perfect information-all investors know
everything there is to know about a company
when deciding how much to pay for a share
risk can be measured by standard deviation Or, alternatively:
all investors are able to borrow and lend at
the risk-free rate. k, = dividend yield + growth in dividends (7)
These theoretical assumptions cannot be fully The company’s cost of equity is equal to its
satisfied; for example, investors clearly didn’t dividend yield plus annual growth in dividends
have ‘perfect information’ about Enron and (both expressed as a percentage). Eqn. 7 is
WorldCom before they collapsed. In addition, known as the dividend growth model (DGM),
there are practical problems when applying the or sometimes the Gordon growth model.
model; the true market portfolio consists of a
share in every asset throughout the world: (ii) Problems with DGM
equities, metals, real estate, cash, human For the DGM to hold, dividends must grow
capital, art etc. In practice, investors have to at a constant rate to infinity. This is clearly
use a market index as a proxy for the true unrealistic for most firms; only stable firms will
market portfolio; for example the FTSE All offer an approximate fit with the model. From
eqn. 5 it is clear that the model only applies altering the company’s gearing level so as to
where gdtv is less than he. As gd,v approaches he arrive at an optimum, minimum WACC.
the share price tends to infinity, which is clearly It was stated above that debt is cheaper than
nonsense. equity, and therefore one might assume that the
How should we forecast growth in divi- company should use as much debt as possible
dends? Historical dividend data can be an to minimise its WACC; however, it was also
unreliable guide, especially if the company has shown that debt introduces financial risks
changed over recent years; expert city analysts’ which are borne by investors.
predictions are rarely reliable to more than two As gearing increases from zero, equity
years into the future, not to infinity as required investors are exposed to additional financial
by the model; and accounting data on profit risk and demand a higher equity interest rate.
growth can be manipulated (within accounting Initially, this is outweighed by the benefits of
rules, of course). There is no simple answer. cheaper, tax-efficient debt and the WACC
The DGM makes a very bold assumption reduces. However, above a certain gearing level
that simply does not hold in practice: the marginal benefits of debt are more than
dividends cannot be expected to grow at a offset by marginal costs from financial risk
constant rate to infinity. In addition the DGM factors such as the risk of cashflow problems,
struggles to cope with high-growth companies, risk of bankruptcy etc. and the WACC begins
and there is no reliable measure of dividend to rise. If the gearing level gets too high, the
growth. The DGM should be used with costs of both debt and equity begin to rise
caution, but is useful as a comparison for other sharply, causing a steep rise in the WACC.
methods. In practice there is an optimal capital
structure and minimum WACC, but un-
Additional models certainty about the costs and benefits of debt
There are other financial models, but none is as means that the exact shape of the WACC curve
simple to use as CAPM and DGM. The multi- is almost impossible to calculate. It would
factor arbitrage pricing theory (APT) is more appear that the best we can do is make an
flexible and more accurate than CAPM, but it educated guess at the company’s optimal
is difficult to understand, apply and interpret. gearing level (see Fig. 8 ) .
APT links equity returns to a number of macro-
economic factors, rather than the single factor Applying the theory
used in CAPM (the return on the market In addition to the problems applying the
portfolio); for example, oil prices, interest models listed above, there are particular issues
rates, inflation, GNP growth etc. A mass of data for the DNOs. No longer are any of the DNOs
is analysed statistically to determine which of quoted on a stock exchange, which means
these factors were historically correlated with that key components of CAPM (beta), DGM
equity returns, and a model is constructed to (dividend yield) and capital structure models
predict future returns, and hence the cost of (market gearing level) cannot be calculated.
equity capital for the company in question. In Given these difficulties, it is more appropriate 8 Optimal capital
contrast multi-factor regression models use to calculate a range for the cost of capital, structure
company-specific data such as profits,
accounting ratios, firm size etc. to predict the
cost of equity capital.
Given the large amounts of data involved in
applying these models, they are much more
cumbersome to use and consequently prove to
be less popular. None of the UK regulators uses
them.
Capital structure
The discussion and examples above show that
the mixture of debt and equity used to fund a
company can be altered. Different gearing
levels result in different company WACCs, and
different levels of risk and return for investors.
The capital structure debate is concerned with
rather than a single figure. The approach gearing levels debt interest rates tend to vary
described below follows finance theory as with gearing-the comp,any may have a fixed
closely as possible. Where practical compro- interest rate (bond coupon), but the price at
mises prove necessary, these too are made with which its bonds trade in the market determines
regard for underlying theory. the current yield (and hence DRP): a measure
The DNOs are regulated on an RPI-X basis- of the market’s current perception of the risk
revenues and asset values are inflated each year associated with lending money to the company.
by the retail price index, less an efficiency The majority of the DNOs maintain their
factor X. Consequently, as inflation is already debt at levels above the lowest investment
built into the regulatory process, the cost of grade; at the last distribution price control
capital is determined on a real basis, i.e. net of review in 1999, the PES!? had an average DRP
inflation. of approximately 1.40X5If we assume that the
DNOs may have to gear up to fund future
Cost of debt capital expenditure programmes, but remain within
Real risk-free rate their ‘investment grade’ licence condition, then
Reputable government bonds are essentially an appropriate DRP is 2.5%, this being the
free from default risk (e.g. UK or US bonds, average currently attracted by BBB and BBB-
not Argentinean bonds). The UK Government (SQP) corporate bonds (data from Barclays6).
issues both conventional and index-linked A DRP range of 1.4% to 2.5% is used in this
bonds or ‘gilts’. The face value of an index- analysis.
linked gilt (ILG) is inflated each year by the
retail price index and so it attracts a lower Inflation risk premium (IRP)
interest rate than its ‘conventionap counter- The majority of DNO bonds are the ‘con-
part. ILG yields can be used to estimate the real ventional’ type, and consequently investors
risk-free rate. are exposed to inflation risk: the risk that
ILG prices (and therefore current yields) are inflation will rise unexpectedly, and signifi-
determined by highly liquid and well-analysed cantly reduce the real value of future bond
financial markets: hundreds of analysts and coupons they receive. To compensate for this
traders are involved in setting market prices investors require an inflation risk premium.
through their buy and sell transactions. Conse- An estimate of the IRI’ can be made by sub-
quently, forward-looking market rates are tracting actual inflation from past indicators of
appropriate for use in estimating the real risk- inflation. The IRP has received relatively little
free rate. However, market rates demonstrate attention in terms of studies to determine its
significant volatility in the short term, and value; the few current available estimates place
therefore rolling averages over previous years it in the range 0.6% to 0.9% (see, for example,
should be used. Using up to five year rolling O X E W discussion7).
average yields on UK ILGs with more than
five years to redemption results in a real risk- Real cost of debt
free rate in the range 2.5% to 3.0% (see, for Combining these constituent figures leads to a
example, OFGEMs discussion4). range of 4.5% to 6.4% for the real cost of debt
(see Table 2).
Debt risk premium (DRP)
The cost of debt can be considered as a generic Cost of equity capital
risk-free rate plus a company-specific debt risk As for debt we will determine a real interest
premium that reflects the market’s perception rate.
of the risk associated with lending to the
company. The DNOs have a licence condition Capital asset pricing model
that obliges them to maintain ‘investment (i) Equity risk premium (ERP)
grade’ status on their debt, as assigned by Referring to eqn. 3, the equity risk premium is
credit rating agencies Standard Q Poor’s (S&P) the expected return on the market portfolio in
and Moody’s Investors Service (Moody’s). The excess of the risk-free ra1.e. Historical estimates
lowest investment grades are BBB- and Baa3, of the ERP are derived by subtracting historical
respectively. As a company gears up, more and government bond yields from stock market
more cash is required to fund debt interest returns. Forward-looking estimates can be
payments; credit ratings give an indication of derived by surveying investor opinion.
the company’s ability to pay up. At higher Unlike for the risk-free rate, forward-looking
~ low high !
real risk-free rate (%I IAl 2.5 3.0 ,
~ debt debt risk premium (%) [B] 1.4 2.5 1
I inflation risk premium (%) [C] 0.6 0.9 I
amroach
real post-tax WACC gross of debt shield (%) [GD + (1-G)H] 5.1 0 7.24 ,
textbook real post-tax WACC net of debt shield (%) [GD(l-T)+ (1-G)H] 4.43 6.09 I
approach I
estimates of ERP are not derived directly from but on the other that this stable environment
the market, and consequently contain an enables them to maintain high gearing levels
element of subjective judgment and bias. with associated financial risks.
The Competition Commission8 found that
forward-looking estimates of the ERP tend (iii) Real CAPM cost of equity
‘not to be derived from rigorously-structured In summary, this leads to a real CAPM cost of
surveys’. ERP estimates should be based equity in the range 5.7% to 8.5% (see Table 2).
primarily on historical data, but with regard for
investor opinion. Real DGM cost of equity
Surveying a number of historical and market (i) Dividend yield
estimates of ERP leads to a range of 4.0% Dividend yield presents this year’s or next year’s
to 5.0% (see, for example, Competition anticipated dividends as a percentage of the
Commission’ and Barclays’). current share price. Again, as the DNOs are not
listed on a stock exchange, we have to use
(ii) Beta comparable companies and industry sector
Determining beta for a company that is not averages as a guide. Figures for UK and US
listed on a stock exchange proves problematic: utilities show that a range of 4.0% to 5.0% is
industry sector betas or comparable company appropriate (figures from Datastream).
betas can be used to provide a reasonable
estimate. However, for the DNOs there are no (ii) Dividend growth rate
directly comparable companies. We could use As described above, there are a number of
the other regulated companies: water, Transco, methods of estimating growth rate, each with
NGC etc., or we could use US electricity its associated problems. A good rule of thumb
utilities; but none of these shares the exact is to use real growth in gross national product.
same regulatory and commercial environment. UK economic data shows that over time
These betas only give a guide. aggregate company earnings tend not to
A great deal of subjective judgment has to go change significantly as a percentage of GNP-
into setting a beta value for the DNOs. The best i.e. the ‘average’company will show dividend
we can propose is a ‘reasonable’range of 0.8 to growth in line with GNP growth. A suitable
1.1.Remember that, by definition, the average range for real growth in GNP is 2.0% to 3.0%
company on the stock exchange has a beta risk (see, for example, OFGEM”)
level of 1.0.The proposed range reflects, on the
one hand, that regulated electricity companies (iii) Real DGM cost of equity
may be considered lower than average risk due This results in a range for the real DGM cost of
to their stable demand and regulated income; equity of 6.0% to 8.0%, which is broadly
Regulatory approach
Textbook application of cost of capital theory is
designed to calculate a suitable discount rate.
This differs fundamentally from the regulatory
process, which is designed to agree an
appropriate amount of allowable revenue. The
regulators want the lowest acceptable figure as
this results in low customer bills, whereas the
companies want the highest acceptable figure
as this increases returns to their shareholders.
The Competition Commission plays an
important role in the regulatory process. If the
companies and their regulator cannot reach
an agreement, the Competition Commission
Table 3 Example of inconsistent UK regulatory approaches to determining the real risk-free rate and
equity risk premium
I Railtrack, initial proposal13 spot yields on lLGs and market surveys and regulatory
I
conventional gilts of mixed precedent ,
1 maturity (as per OFWAT method) 1
’ Scottish HYdrol6’ reference to ‘the relevant literature’ reference to ‘the relevant literature’ 1
, MMC reports I
i Competition Commission 3.0% 4.0%
j (2000)-Sutton and East ILG spot yields adjusted for ‘best estimate’ based on historical ~
1 Surrey WateP anticipated increase in rates averages and a range of utility and I
investor surveys of future expectations ~
makes the final determination; consequently Regulators (and companies) make selective
the companies and regulators pay particular use of regulatory precedent to support and
regard to the Competition Commission’s justify their negotiating positions.
approach to calculating the cost of capital. Regulators often cite seemingly authoritative
A review of recent regulatory WACC deter- reports (including their own) in order to
minations by OFGEM, the Office of Water support their proposals; however, examina-
Services (OFWAT), the Office of Rail Regula- tion of these references often leads to a trail
tion (ORR), the Monopolies and Mergers of citations with little substance.
Commission (MMC) and the Competition
Commission shows evidence of unnecessary For example, see the range of figures and
risk in the regulatory process: methods used to determine the real risk-free
rate and equity risk premium in Table 3.
Regulators frequently use arbitrary, subject- These regulators calculate the cost of capital
ive judgment to set cost of capital figures. primarily using CAPM with DGM in support,
Despite a stated desire for consistency over but because of the issues described above,
time and between regulators, the evidence regulatory cost of capital determinations are
shows that this does not occur in practice. essentially subjective judgments based on data
that links a share's return to its covariance with the market portfolio, as
1
risk-returncombinations available by lending or borrowing at the risk-free
investing in the market portfolio
of debt and equity capital employed in the firm. Measured by the gearing ,
which meets the requirements of the two the author, and not necessarily of any person,
theoretical models to only a partial extent. company or organisation mentioned herein.
Conclusions References
The cost of capital is a significant figure for 1 WILLIAMS, E , and STRBAC, G.: ‘Costing and
pricing of electricity distribution services’, Power
the regulated UK utilities (not just electricity
EngineeringJournal, June 2001,15, (3), pp. 125-136
distribution); in part it determines billions of 2 ARNOLD, G. C.: ‘Corporate Financial Management’
pounds of market value, customer bills and (Financial TimesRrentice Hall, London, 2002, 2nd
returns to capital providers. edn.)
The current DNO real cost of capital of 6.5% 3 SHARPE, W. E: ‘Capital asset prices: a theory of
market equilibrium under conditions of risk,
is towards the low end of the range produced Journal ofFinance, 1964, 19, pp. 425-442
in this article, which indicates that the 4 OFGEM: ‘The transmission price control review
‘
companies may not be making sufficient of the National Grid Company from 2001-draft
returns to compensate capital providers for the proposals’ June 2000, www.ofgem.gov.uk/public/
risks associated with a regulated electricity pubOl.htm, pp. 42-59
5 OFGEM: ‘Review of public electricity suppliers
distribution business. 1998 to 2000: Distribution price control review draft
Textbook finance theory suggests that proposals’, August 1999, www.ofgem.gov.uWpublic/
calculating the cost of capital for a company is adownloads.htm, pp. 65-74
a mechanical process. However, theoretical 6 Barclays: Weekly Bond Price List, 9th September
2002, Barclays Capital, United Kingdom
problems with financial models and practical
7 OXERA: ’Inflation risk premium’, The Utilities
problems applying them turns the process into Journal, March 2000, pp. 16-17
an estimation exercise. The DNOs face 8 Competition Commission: ‘Sutton and East Surrey
additional problems because they are not Water plc. A report on the references under sections
quoted on a stock exchange and they face an 12 and 14 of the Water Industry Act 1991’ (The
Stationery Office, London, 2000)
inconsistent UK regulatory regime (so far as
9 Barclays: ‘Equity-gilt study’ (Barclays Capital, UK,
cost of capital determinations are concerned). 2002)
Investors, companies and customers would 10 OFGEM: ‘Review of public electricity suppliers
all benefit from the reduced risks associated 1998 to 2000: Distribution price control review
with a common, less subjective approach to consultation paper’, May 1999, www.ofgem.gov.uk/