Linked Bonds Model Used By The Uk

Published Date: 02 Nov 2017

Student Name – Chitra Naraindas Relwani

Student ID – A4042738

MSc Finance Intake 10

Table of Content

Abstract 2

Introduction 3

Literature Review 6

Research Philosophy 20

Question, Hypotheses and Aim 21

Methodology 23

Data collection 29

Result 32

Discussion 42

Limitations 43

Conclusion 44

Reference 45

Abstract

This paper looks at the possibility of implementing inflation-linked bond model used by UK linkers to Indian utility companies. After establishing the correlation between utility company revenues and inflation in India, a possible model for implementing linkers is studied.

Keywords – inflation, inflation-linked bonds, utility companies, break-even inflation, deflation floors, Canadian model

Introduction

The corporate bond market in India is underdeveloped (Khanna and Varottil, 2012). The main source of finance for the rapid Indian economic growth has been retained earnings and equity capital. Regulators have been trying to boost this market, however there is not much growth. Indian companies have been looking to the international markets for debt offerings to raise capital (Khanna and Varottil, 2012).

It would be expected for the less risky debt market to develop before the more risky equity market, however, with a strong and growing equity market, the time seems ripe to develop the corporate bonds market (Khanna and Varottil, 2012). Although debt financing can be obtained as a bank loan, it would be cheaper to obtain financing by issuing debt in an active bond market (Khanna and Varottil, 2012).

Currently, the corporate bond market forms 2% of India’s Gross Domestic Product (GDP) in contrast to the equity market capitalisation which forms 56% of GDP (Khanna and Varottil, 2012). This is in sharp contrast to developed countries where the corporate bonds market is larger, or at least as large as the equity market.

Utilities companies are important for the development, growth and wealth-creation for any country – developed or developing (Sutton, 2007). The utilities sector includes electricity, water and sanitation, oil and gas transmission and distribution and even telecommunication (Sutton, 2007), however, this paper focuses on electricity and oil and gas sub-sectors.

The importance of utilities companies lie in the following benefits they provide (Sutton, 2007):

Water, sanitation and electricity form the core of human development. Hygiene, required for healthy living conditions is still something that needs to be achieved in various parts of the world. Electricity is critical for education and information dispersal.

Electricity, water, oil and gas and telecommunication are considered the most basic of infrastructures needed for economic development.

In India, the electricity sector was ravaged by capital crisis (Reddy, 2001). There was an inability to self-finance any improvement or growth projects during the early part of this century. The problem was aggravated by heavy borrowing and whatever aid government treasuries offered in the past, could no longer be done due to lack of government finances. To solve this issue, private capital injections were taken and this included foreign capital investments (Reddy, 2001).

The World Bank diagnosed the sector problem to be due to its size, monopoly, government interference and centralisation causing inflexibility and inefficiency. With infusion of private capital, privatisation took place thus making the sector more competitive and in some aspects more efficient (Reddy, 2001). However, it is still unable to meet its required expansion and suffers from inefficiencies in transmission and distribution (Shah, 2011). The need for financing, along with political reforms is needed within this sector to bring it up to the mark for better infrastructure that needed for a growing economy.

Taking the cue from this, the role of inflation is seen to be critical in earning revenues. The electricity sector, under the Electricity Act 2003 can increase its tariff based on inflation and related measures (Ministry of Power, 2011).

Inflation is considered a part of growth, however, there are cases where growth has been seen without inflation, (Salian and Gopakumar 2012). Over the years India’s policies have created economic growth, 3.5% in the 1970s to 5.5% in the 1980s , before privatisation started in the early 1990s fuelling growth to a high of 9.8% in 2007 (World Bank) and inflation growing from 1.7% in the 1950s to 8% in the 1980s and in recent years to very high levels of 12% in 2010 (Salian and Gopakumar, 2012).

The use of inflation linked bonds has been known to be an indicator of future inflation expectations and real interest rates which are commonly used around the world by central banks (Hurd and Relleen, unknown). The Bank of England regularly uses the prices of inflation linked and conventional bonds to derive real and nominal yield curves from which it infers the market-based inflation measure.

Using this information background, inflation linked bonds can offer a solution to the finance issues faced by utility industry and offer an attractive investment opportunity to the inflation-wary population of India.

Literature Review

Inflation Linked Bonds – A bond is a security which recognises the issuer’s debt obligation (Fabozzi, 2007). Inflation linked bonds or linkers have been around for centuries. The first issue of such an instrument was in 1780 (Waisberg et al, 2004) in the state of Massachusetts. More recent issues have included governments and corporates across the developed and emerging markets. These bonds are different from nominal bonds because their cash flow is linked to a price index. This indexed cash flow provides a hedge against the eroding properties of inflation (Deacon et al, 2004). This cash flow is divided into two parts, the nominal rate and the compensatory part which protects against inflation (Deacon et al, 2004). With emerging markets facing high inflation rates (Harding and Giles, 2011), it makes sense to index these debt issues to the inflation figures. By doing this there is a greater likelihood of attracting more retail and institutional investors such as pension funds and insurance companies who would like to hedge their assets against inflation. Recently the Reserve Bank of India announced the re-introduction of inflation linked bonds (linkers) into the Indian bond market (Sikdar, 2011). This announcement comes in the wake of high inflation rates which erode the purchasing power of the public. The next step if such an issue by the government is successful would be for companies to issue such linkers instead of nominal bonds. The companies that can successfully issue linkers are those whose revenues are linked to inflation such as utility and retail companies (James and Sooben, 2010). Being able to issue debt instruments which hedge the investor against inflation can provide companies with a more ready source of capital and can tap into a different investor base (Deacon et al, 2004). Diversifying the debt portfolio can decrease the company’s risk (Deacon et al, 2004) and finally, with inflation linked bonds, the initial coupon payments will be low compared to nominal payments (Deacon et al, 2004) since the inflation indexing is lagged. A study by Briere and Signori (2009), argues that the inflation linked bonds have reached a high level of volatility and inflation expectations have stabilised thereby making them substitutes for nominal bonds. This study was conducted on data collected during the time frame 1997-2007. However, after the recent financial crisis, inflation targeting by governments has come under fire (Woodford, 2012) causing an instability in inflation expectations. Thus contrary to Briere and Signori, inflation linked bonds do offer a diversification in the portfolio of both issuer and investor. With structural inflation caused due to high growth phases (de Silva et al, 2011) in India, it seems like a good time for companies to think about issuing inflation linked bonds rather than conventional bonds.

Using the UK as the model for implementing corporate linkers in India was based on multiple factors. UK has a well established corporate linkers market and to be able to use a tried and tested model would be in India’s favour. Also, UK shifted from its old UK model to the new style Canadian model with much success, so obviously it would be a good market to study. The only point of contention is whether or not to add deflation floors which are not present in the current model used in the UK.

The UK bond market has an active government issued inflation linked bond section, much like the remainder of the G7 countries. The first government issued inflation linked bond was introduced in 1981, maturing in 1996 (Bettiss, 2010). The older model followed by the UK was different from the new style Canadian model. The old-style inflation linked bonds do not trade in real terms, instead they trade in clean-price cash terms. Hence, with inflation rising, the price at which the bond will be traded will be uplifted and no index ratio is used to adjust the price (Bettiss, 2010). The old-style linkers trade with an 8 month inflation lag (Bettiss, 2010). Currently the UK follows the Canadian model which was introduced in 1990 by Canada and is currently the standard followed by many countries (James and Sooben, 2010). The Canadian model was modified to include a deflation floor which made it all the more popular (James and Sooben, 2010). However, the UK decided not to include this deflation floor in its Canadian model implementation (James and Sooben, 2010). This is a point of difference between the current discussions of linkers in India versus the new bonds issued in the UK.

The inflation index used by the UK linkers is the Retail Price Index (RPI). Although the Consumer Price Index (CPI) is the government’s target index, the RPI is used for indexation because most inflation linked liabilities are linked to the RPI (Bettiss, 2010).

Price indices should reflect only the price changes and not changes in quality. The inflation indices around the world are subject to hedonic adjustments. This adjustment is done to justify the changes in the quality of goods and services, especially with high technology goods which evolve rapidly creating the need for adjustments. A hedonic adjustment is done by relating price of the item to its measurable characteristics. With these characteristics, a regression is run and the co-efficient values represent the change in price for every unit of change in that characteristic (Brereton, 2005). In the UK, the Producer Price Index (PPI) is adjusted using hedonic regression. However RPI is not adjusted using hedonic regression (Ball and Allen, 2003). The process of hedonic adjustments is described by Silver and Heravi (2002) in a European Central Bank working paper. The matched model method is currently used wherein the prices and quantities are checked every month for changes which are then reflected in the CPI. One drawback of the matched model method is that it might continue matching models for samples which become unrepresentative of the population thereby creating a distortion. Another disadvantage of this method is that the price collectors do not change to the new models until they are forced to do so. This causes unusual price changes by shifting from an already-obsolete model to a newly launched model. Several other issues of hedonic adjustments such as weighting of matched models versus unmatched models and imputation of currently unavailable model price from available matched surveys distort inflation figures which can be detrimental to investors.

In the UK although nominal bonds dominate the corporate bond market, there are those sectors whose revenue are linked to inflation and hence they issue linkers. Such sectors include utility and retail sectors. Utility companies in the UK have been the main issuers of corporate linkers. This is attributed to the regulations which link their revenue stream to the Retail Price Index (RPI) (Bettiss, 2010). National Grid Plc is the largest issuer of corporate linkers in the UK. In 2011, it issued a RPI linked 10 year sterling bonds which were targeted for the retail investors raising a record GBP 260 million (London Stock Exchange web site).

One of the important considerations for corporate issuers of linkers is accounting standards and their acceptance of linkers for hedge accounting (James and Sooben, 2010). Historically UK allowed bonds to be accounted on an amortised basis while the uplifts were financial charges. However, recent changes allow inflation to be hedge accounted only if changes in inflation form a part of cash flows in the contractual instrument (James and Sooben, 2010).

A Bank of International Settlements (BIS) paper (February 2006) discusses the development of the corporate debt market in Asia. The authors, quite accurately identifies that as the emerging markets develop economically, they require more funds to run their businesses and that bank loans, which are generally information-intensive do not suffice. The debt market forms an integral part of corporate borrowings around the world. In this paper, Sharma and Sinha of the Reserve Bank of India talks about the requirement of a well-developed corporate bond market in India.

Although several academic papers have been written about deriving inflation expectations from inflation linked bonds and their use in the western markets, no study has been conducted regarding the possibility of introducing inflation linked corporate bonds in the Indian debt market. In 1997 Indian government introduced inflation linked bonds to the market but the response to this was no more than lackadaisical which was attributed to the lack of a deflation floor which is put in place to protect the invested capital (de Silva et al, 2011). Although this is a very good reason for institutional investors to stay away from the bond, a lay-person would not invest in a new product where the cash-flow calculations are wrought with complexities as discussed by the Reserve Bank of India paper on Capital Indexed Bonds (2004) and the secondary market is illiquid and underdeveloped (Patil, 2001).

In 2004 the Reserve Bank of India (RBI) published a discussion paper which dealt with the technical issues of the 1997 issue and possible modifications for them. One change was to link both the principal and coupon with the inflation index rather than the principal alone as was the case in 1997. A deflation floor was also added to the principal repayment at maturity. The Canadian model for indexation was to be followed since the index lag time was considerably less than the UK model. This, according to the paper would be preferred since fluctuations in inflation can be captured quicker with the Canadian model. Such considerations and changes will of course make a linker more attractive as a hedge against inflation.

Another technical paper published by the RBI in 2010 defines the benefits and risks of investing in linkers and defines a similar cash flow calculation as in the 2004 paper. This paper also explains why India uses the Wholesale Price Index (WPI) and not the Consumer Price index (CPI) to link the bonds to. Since the social sections of India vary vastly, there is no single CPI number that can justify each section, thus the WPI numbers are used as the inflation index for bonds. Although using WPI solves the social section issues, it is not a true measure of the inflation faced by the end-user since it measures the wholesale price rise rather than the consumer or retail price rise. If this issue could be solved, a linker would be more representative of inflation hedging that it would be now with the WPI. The WPI in India is adjusted for seasonality and for quality changes using hedonic estimates (Nadhanael and Pattanaik, 2010). These hedonic adjustments can distort the indices thereby putting an investor invested in index linked savings at a disadvantage.

A short explanation of the WPI, its composition and period revision along with changing the base index is warranted. The WPI is considered the most important index in India for inflation measurement, more so than the CPI since it captures price movements in the most complete way (Anon, unknown http://eaindustry.nic.in/WPI_Manual.pdfhttp://eaindustry.nic.in/WPI_Manual.pdf). It is not only used as an index for inflation linked financial products, but it is also used by the government for fiscal and economic policies. Due to changes in economic situations and changes in people’s tastes, some economically significant items may become obsolete over time and the method used for compiling the WPI, Laspeyre’s formula, has the disadvantage of using fixed weighting of components thus ignoring their dynamic nature. These reasons require period revision in weights and items that compose the WPI. Currently, the 2004-2005 financial year WPI is taken as the base (where base = 100). Prior to that 1993-1994 WPI was considered the base. As the base changes, the WPI increase or decrease will also be distorted. For example, when changing from 1993-1994 base to 2004-2005 base, the annual inflation figure for calendar year 2009 was changed from 237.1 to 127.86. Due to change in base year, the base effect causes the inflation figure to decrease. The Office of Economic Advisor published a manual which shows the details of how the composition is developed and changed. An interesting thing to note is the increase in the number of items included in the series from base year to base year, with 1970-1971 having a total of 360 items and 2004-2005 having 676 items, showing a change in economic conditions. With this background in the Indian inflation bond market, a look at the UK market will lead us into the research of the proposed question.

The Indian Accounting Standards have been moving towards internationally accepted standards and recognises inflation linked bonds for hedge accounting. The IAS 39 is used wherein the bond with a fixed payment at maturity and a fixed maturity date with the bond’s interest payments linked to the inflation index will take the principal to be separate from the interest payments. The interest payments are considered to be an embedded option and are accounted for at fair value as discussed by the Ministry of Corporate Affairs Indian Accounting Standard.

Certain issues of inflation linked bonds which usurp the full power of hedging are (Deacon et al, 2004):

The consumption basket that is used for calculating the inflation indexes may not match every individual’s consumption basket.

There is a lag of 3 to 8 months in reporting the index values. The technical paper Inflation Indexed Bonds by the RBI (2010) proposes to use the 4 month lagged WPI numbers if the government were to issue linkers.

From a technical perspective linkers have the issue of identifying breakeven inflation and figuring out the nominal duration to be assigned to them (James, 2010).

Break-even inflation is described as the inflation rate that will make the returns on inflation linked bonds equal to the returns on comparable nominal bonds (James, 2010). The breakeven inflation rate is achieved when a linker is held to maturity and breaks even with a comparative nominal bond also held to maturity (Phoa, 1999). When the yield rates are relatively low, the breakeven inflation rate is low (James, 2010), however in high yield markets like India, the breakeven rate may be distorted. Fisher’s equation is used to calculate break-even inflation.

n = r + f + p

where, n = yield on nominal bonds

r = real yield on index linked bonds

f = inflationary expectation

p = risk premium

bei = break-even inflation

The inflationary expectation and the risk premium together form the break-even inflation which then equates nominal yield to real yield.

f + p = bei

This convention is a simplified version and is currently used in the UK as a market convention. However, this formula gives accurate results if the yield is low. In India, the bond yields are much higher at around 7.87% as of 14th January 2013 for 10 year government bonds (Bloomberg website, 2013 http://www.bloomberg.com/quote/GIND10YR:IND). Corporate bond yields will be higher that will account for credit spread to compensate for additional risk. Due to this, the more complex version of the formula will give a more accurate break-even. Below the equations given are for annual and semi-annual break-even inflations respectively (James, 2010).

(1 + bei) = (1 + n) / (1 + r)

(1 + bei) = [1 + (n/2)]2 / [1 + (r/2)]2

As the equation shows, breakeven inflation has two components, risk premium and inflation expectation. Once the break-even inflation rate is obtained, it is a difficult task to assign separate values to these two components (James, 2010). It is logical to assume that as the maturity of the linker increases, the inflation risk premium will also increase, however, in practice it has been seen that the investors do not necessarily demand higher premium for inflation linked bonds for longer maturities (James, 2010). This concept is applicable universally and thus can be an issue even in the Indian scenario. However, since inflation swaps, which are considered a more accurate reference for inflation expectation (James, 2010) (and hence calculating the risk premium) are not currently traded in India, break-even inflation can be used as a measure of inflation expectation.

Deriving the modified duration for a linker is the same procedure as that for a nominal bond. A study conducted by Dryden and Hancock (1992) show that a parallel move in inflation and nominal yields occur, there will be an overstatement of prices for the indexed security. This implication adversely affects investors like pension funds and insurance companies who are long term investors. Another measure of sensitivity is the effective duration which takes into account the effect of sensitivity on cash flow (Fabozzi, 2007) . For inflation linked bonds, the beta value can be used to measure the sensitivity of real yields to changes in nominal yields. The beta value however is unstable and depends on the frequency of data, that is, whether it is daily, weekly or monthly data. To overcome this, the volatility of relative returns can be used. This shows inflation carry and is important for longer maturity inflation bonds and can also show trends in inflation data (James, 2010).

Despite the drawbacks of linkers, four main reasons can be cited in favour of indexing debt as discussed by a paper by Scholtes (2002):

An ex ante benefit for both the issuer and the investor is seen by removing the uncertainty of real cost of borrowing and return on investing

The issuer is benefitted by having a cheaper source of ex ante debt finance

Investors are protected against inflation, albeit not cent per cent

In high inflation countries, it improves monetary control and helps in the development of the debt capital market

Looking at the obvious advantages of issuing and investing in linkers justifies this study. The paper studies the possible implementation of inflation linked bonds model used in the UK for Indian utility companies.

Other ways of Hedging for Inflation: Derivatives have been used historically for the purpose of hedging against risk. Risk is innate to the financial markets and with a rapid globalisation and liberalisation, it has become all the more predominant. Inflation is a risk faced by all investors as it erodes purchasing power. The recent financial crisis, heated growth and currency depreciation have fuelled India's inflation. Many investors use various asset classes like commodities, equity and even real estate to hedge against inflation (Bruno and Chincarini, 2010).

Infrastructure is believed to be a new asset class to hedge against inflation (Inderst, 2010). Indian infrastructure companies including infrastructure financing companies have issued bonds which are also tax deductible (Mahesh, 2012). The financial and economic features of an infrastructure investement is such that they offer low risk and long term returns which are a-cyclical and also hedged against inflation (Inderst, 2010). A study by Bruno and Chincarini (2010) suggests an optimal hedge protection portfolio of 57% government bills, 6% oil, 12% gold, 21% world bond and 3% emerging equity. Another interesting point discovered in this study is that inflation linked bonds owned in isolation rather than in combination with other asset classes may provide effective inflation hedging.

Inflation derivatives are also available in the market mainly in the form of inflation swaps (Deacon, Derry and Mirfendereski, 2004). Historically, interest rate swaps and inflation swaps share many common features.

Utility Sector in the UK - The utility sector forms the basis on which the economy can grow. It impacts all other industries and it contributes by increasing employment opportunities (Anon, 2011). In 2010 it was noted that the power and gas sector contributed GBP 28billion towards the Gross Value Added (GVA) of the UK economy (Anon, 2011). GVA of each sector of a country adds up to its Gross Domestic Product (GDP). This sector was among the few that created jobs during the early years of recession since more capital investment was required for infrastructure development (Anon, 2011). Since the initial capital investment required is very high creating a natural entry barrier, monopolies are possible in the utility sector. To prevent this it is a closely regulated sector (Scarsella, 2008). Regulation also ensures a high quality of supply and service (Scarsella, 2008). The regulators issue licenses which set out rules about the price, quality of service and prohibition of cross-subsidy (Scarsella, 2008). The pricing model followed allows the companies to increase their tariffs each year in line with the RPI-X. RPI-X is based on the regulatory bodies’ estimated of operating expenditure, capital expenditure and asset replacement along with estimated depreciation and an allowed rate of return on capital employed (National Grid Annual Report, 2010-11).

Utility Sector in India – India was the fifth largest power generator as on 30th September 2009 generating 152 GW, about 4% of global power (Anon, 2010). The power sector remains dominated by the central and state government, whether it is generation or distribution (Anon, 2010). The power sector, like the water utility sector is heavily regulated. The Electricity Act in 2003 expanded this sector to invite private as well as public participation, installed licensing requirements to produce energy, encouraged competition (Anon, 2010) and placed pricing regulations which were adjustable based on indicators of inflation and purchasing power (Ministry of Power, 2011). A Planning Commission report in 2006 analysis the Indian utility sector regulation as an evolving idea and mentions that the sector regulation at the time were similar to those in the UK at the turn of the century. Thus, following the UK’s regulatory system, Indian utility regulation will need some time to consolidate and regularise and standardise the systems it has put in place (Planning Commission, 2006). As David Hull mentions in his paper in 2008, inflation linked bonds issued by utility companies are generally long term issues and investors will invest in them only if they are sure of the company’s business and revenue. With the increasing demand of energy in India and economic reforms being implemented, the time for issuing such bonds has drawn closer.

Research Philosophy

In order to conduct research which is valid and reliable, a research philosophy needs to be followed. It describes the assumptions that underpin the study being conducted (Saunders, Lewis and Thornhill, 2009). The research carried out here is based on the observable knowledge that UK utility companies issue inflation linked bonds because their revenue stream is linked to inflation. This observable knowledge is being applied to another market where utility companies have their revenues linked to inflation. The application of inflation linked bonds in the Indian scenario is driven by data observed and collected from government websites and company reports. Based on this, the research philosophy followed here is empiricism in general and positivism in particular (Ryan, Scapens and Theobold, 2002 http://www.cengagebrain.com.au/content/ryan28817_1861528817_02.01_chapter01.pdf).

The deductive research approach has its base in scientific research. It requires thorough testing before a theory can be accepted. It involves hypothesising, researching and testing the hypothesis for the specific outcome and then modifying the theory in light of the results (Saunders, Lewis and Thornhill, 2009). Based on this description, the approach followed in this research is a deductive approach. Part of the research involves explanatory studies (Saunders, Lewis and Thornhill, 2009) of the relationship between WPI and revenues in Indian utility companies by use of correlation and regression analysis.

The data collected for this study is mainly secondary data which is described by Saunders et al (2009) as data which was originally intended for another purpose than for this research. Although the revenue information and inflation information are directly from the source, that is, from the company reports and government website respectively, the purpose of data emission was for public information and government policy-making. Due to the nature of this data, this study is strategically archival in nature. The data collected is analysed in multiple techniques such as correlation, regression, Granger Causality to name a few. This invokes the multi-method analysis procedure (Saunders, Lewis and Thornhill, 2009). In this research, time series data is collected, that is, it is the same type of data collected over time. This classifies it as longitudinal data according to Saunders et al (2009).

Question, Hypotheses and Aim

Research question – Implementing UK’s Inflation Linked Bonds model to Indian utility companies.

Hypotheses – The current research ventures into a new area in the Indian fixed income market. On the basis of previous research, current news flow and developments in the Indian debt market, this research has been undertaken. The following hypotheses are to be proved or disproved–

There will be a positive correlation between WPI and revenues of the sample Indian utility companies. This assumption is based on the principles of tariff fixing in the Electricity Act of 2003 where the tariff adjustments will be limited to indicators of purchasing power and inflation indexes (Ministry of Power, 2011).

The linker model used in the UK will be implementable to Indian utility companies.

Aims and Objectives - The aim of the paper is twofold – by conducting correlation and regression studies of revenues of listed Indian utility companies and the Consumer Price Index, the feasibility of issuing linkers can be found out and second, the possibility of implementing the model used in the UK to these companies. These can be broken down to multiple objectives to define the research further:

To understand the model used for inflation linked bonds in the UK. As explained in the literature review, the model followed is the Canadian model with the exception of deflation floors.

To establish a relationship, whether positive or negative, and the strength of this relationship between WPI and revenue stream of a sample of listed Indian utility companies. This will include correlation and regression analysis

Financial ratio analysis, mainly liquidity and coverage ratios will be conducted to determine whether the companies are financially sound to issue linkers

After the eligibility and ability is established, the implementation of the model will be studied by looking at possible pricing, deflation floors and caps that can be added.

The likely findings will be that there is a positive correlation between inflation and company revenues and that the model can be successfully implemented.

Methodology

Methodology – The research will be carried out mainly using observations and quantitative techniques. The observable aspects will be used to define the current inflation linked market in the UK, the application of these bonds in the utilities sector, the current economic conditions of India and its impact on the Indian bond market. The quantitative aspects will be the correlation and regression studies, including studies of normality of the variables involved in the issue of corporate inflation linked bonds, such as revenue stream, inflation index, interest rates and time to maturity. The first step of this paper is to identify the linker model used by UK utility companies. From a study conducted by Alexander and Shi-Chien in 2002 a basis for data collection is used. To study the UK utility linkers, the data collected will include:

Total value of the issue

Date of redemption

The coupon of the bond including the breakeven inflation

The price of the bond

The yield to maturity

The margin of yield over the government linkers

The RPI figures

This information is available in the issue prospectuses and on company websites.

Once the UK bond model is understood, a study will be conducted to see the correlation between Wholesale Price Index (WPI) figures and the revenue streams of the sample Indian utility companies. Also, a regression study between WPI, which is the independent variable and the revenue which is the dependent variable will be conducted to understand clearly the relation between WPI and revenue stream of the companies. A paper by Aggarwal et al (2009) talks about energy price forecasting based on regression which is a method useful in this case to forecast the relationship between inflation and revenue since a regression study is going to be conducted. The companies chosen are among the two utility sectors – oil and gas sector and power sector. Within the water utility sector, the companies are government held and not listed on the stock exchanges. The financial information required is not available and they are managed locally by either each city’s or each state’s governing bodies for example, the Bangalore Water Supply and Sanitation Board is run by the Bangalore Municipality while the Karnataka Urban Water Supply and Drainage Board is managed by the government of Karnataka state. Among the power and gas utility sectors several publically listed companies are available. The sampling method used is stratified sampling based on the sub-sectors within power and gas utility sector.

Power Sector

Power Generation and Distribution Companies

National Thermal Power Corporation (NTPC), Tata Power Limited, Power Grid

Power Financing Companies

Rural Electrification Corporation (REC), Power Trading Corporation Limited (PTC), Power Finance Corporation Limited (PFC)

Oil and Gas Sector

Upstream Companies

Oil and Natural Gas Corporation (ONGC), Cairn Energy

Refining Companies

Mangalore Refinery Petrochemicals Limited (MRPL), Essar Oil

Marketing Companies

GAIL India Limited, Indraprastha Gas Limited (IGL)

Table 1 – List of Indian utility companies to be studied

Since power financing companies have the onus of raising capital to finance power projects, they are a valid sample for this study. Although in 2002 the government dismantled the Administrated Pricing Mechanism to help oil and gas be priced by the market (Ministry of Petroleum and Natural Gas, 2002), some control is still exercised in pricing fuel (petroleum and diesel) and cooking gas (Liquefied Petroleum Gas – LPG) with prices being hiked to cover rising oil prices.

From the financial statements of the sample companies ratio analysis will be conducted to analyse the liquidity condition and the coverage ratios. Comparisons between the sample companies and with industry averages will help determine whether the company has financial feasibility to issue the bond.

The final step will be to see whether or not the linker models used in the UK are implementable in the Indian context by building a pricing model. Concepts of adding deflation floors and caps will be discussed and a foundation for further research in these floors and caps will be laid down. Beyond this point, future research can be in the possibility of adding additional options such as call and put options, conversion into equity etc.

The various tests used to confirm a relationship between the revenue stream of the selected companies and the WPI include correlation, Granger Causality, Dickey Fuller test for stationarity and finally regression analysis with a test for serial correlation, the Breusch Godfrey test. These tests can help confirm that WPI can be used to predict revenues of the companies. However, due to various reasons discussed later, there are limitations to this certainty.

Breusch Godfrey Test: Serial correlation causes incorrect standard error estimates. No correlation between the error terms is one of the assumptions on which regression is based (DeFusco et al, 2004). This problem may be exacerbated by autoregression (DeFusco et al, 2004). A diagnostic test called the Breusch Godfrey test is used. The test consists of hypothesis testing, with the null hypothesis being that there is no serial correlation. If the probability value is greater than 0.05 (at 5% significance level), there is not enough evidence to reject the null hypothesis.

Augmented Dickey Fuller Test for Non-Stationarity: When a time series has constant statistical attributes, such as mean, variance and auto-covariance, it is said to be stationary as time progresses, if not it is considered to be a non-stationary series (Brooks, 2008). This characteristic of a time series is important since a non-stationary time series will cause the error term in regression analysis to have a non-declining effect on the current value of the dependent variable which is a spurious regression (Brooks, 2008).

Augmented Dickey Fuller test is used in this case to test for non-stationarity in the time series. This is a one-sided left-tailed test and the null hypothesis is that the dependent variable contains a unit root (that is, it is non-stationary). If the test statistic is more negative compared to the critical value, then there is enough evidence to reject the null hypothesis since it is a one-sided test (Brooks, 2008). A limitation of using this method in EViews software is that the process is calculated for twenty observations. Thus, a sample of seven may not give accurate results.

Granger Causality: Although a simple correlation analysis shows the existence of a relationship between two sets of data, it does not necessarily mean that one causes the other. The direction of influence, if any cannot be established (Gujarati, 2004). To determine whether one time series influence the other, Granger causality test is conducted. It must be kept in mind that this test does not show causation in the generic sense, rather it shows which series precedes the other or one series can be used to predict or forecast the other series (Jacobs, Leamer and Ward, 1979).

To perform this test, a suitable lag is required, however in this case, since the data available is limited to seven observations, lags of beyond one period were not possible. This limitation can be overcome by increasing the number of years of data included.

Regression analysis: Working with time series possess challenges which make simple linear regression ineffective. Various issues need to be dealt with when using time series data. Autoregression is a key feature of time series data which needs to be addressed before applying linear regression models (DeFusco et al, 2004). Autoregression occurs when the current period value of the series is related to its previous period value. Another issue faced in time series regression is heteroskedasticity which occurs when the variance of the errors differs across observations (DeFusco et al, 2004). When this phenomenon is present, although the consistency of regression parameters is not affected, it can lead to mistakes in inference due to unreliable F-tests (DeFusco et al, 2004).

A number of models were used and rejected before the ARMA model was accepted. First, since the time series are seen non-stationary, they will be tested for co-integration so that co-integrated regression could be applied. According to Engle and Granger (1987), a linear combination of two non-stationary series may be stationary. Due to a lack of enough observations, the Johansen co-integration test could not be performed; instead the Engle-Granger test was performed. The null hypothesis is that the series are not co-integrated.

The Autoregressive Conditional Heteroskedastic model will be looked into since heteroskedasticity is common in time series data. However, since this will involve calculating the volatility, thereby reducing the number of observations further, the results obtained will be further compromised due to small-size bias.

The penultimate model that will be tried is the distributed lag model. This model takes into consideration that the dependent variable may be influenced by a time-lagged value of the independent variable. For this, a lag of one period will be taken since inflation figures that are announced have a lag period but going beyond one period (here one year) is not likely.

The final model will be the ARMA model. This takes into consideration the autoregression factor. We introduce the first order autoregressive term, AR(1) into the equation which takes into consideration the immediately prior period value of the time series. F-statistic confirms the significance of regression parameters and if this is unreliable due to heteroskedasticity, the entire analysis falls apart (Agung, 2009). To prevent this, robust standard errors need to be calculated. In EViews, this is done by using White heteroskedasticity-consistent standard errors and covariance. Once a regression model is accepted, the Theil Inequality Coefficient gives a measure of the suitability of the model to predict the dependent variable by looking at the difference between the estimated value and the actual value. A value of zero shows that the model is a perfect fit while a value of one means the model is no better than a guess.

Data collection

Before starting with the analysis of WPI and revenues of Indian companies, the data of the selected UK companies is collected. The revenue data consistently available for all companies is from 2006 to 2012 from the annual reports. RPI and data is downloaded from the Office of National Statistics website. This being a government website is a reliable source of information and hence no cross examination is required with any other source of information.

National Grid – National Grid is an investor-owned energy company and is an international player providing electricity and gas across Britain and the northeast United States. The company mentions in its website that it focuses on the medium term to help realise its vision.

In 2011 the company introduced the RPI linked 10 year sterling bonds offering an interest rate of 1.25% per annum (Anon, 2011). The interest is payable semi-annually on the 6th of April and 6th of October every year until the face value is repayable or if it is redeemed. The maturity date of the bond is 6th October 2021. The bonds were available with the brokers or financial advisors with the minimum initial amount purchasable of GBP 2000 after which they could be bought or sold at GBP 100 face value. The company National Grid Plc is rated Baa1/P2 by Moody’s, BBB+/A2 by S&P and BBB+/F2 by Fitch. The interest rate offered would include the credit spread necessary for this range of credit rating.

The RPI index is used for the calculation of interest. The fixed component of the interest rate is 1.25% which is the figure before any inflation adjustments. Since the coupon is paid semi-annually, the figure is divided by 2. This figure is adjusted for inflation using the available inflation figure (which is lagged by 8 months) using a base RPI figure. This base RPI is the one available on the date of issue, that is on 6th October 2011 (thus, due to the lag, the figure will correspond to February 2011 RPI figure).

(100*coupon rate/2)*(RPI relating to the month 8months prior to the coupon payment date/base RPI)

This is rounded to the nearest penny and multiplied by (invested capital/100)

(Anon, 2011)

The future payments of interest will depend on the changes in the RPI. Once the bond has reached maturity and no prior redemption has been made, the final payment will consist of the final coupon and the face value adjusted for RPI changes. A floor is introduced here in that the payment will consist of at least the face value if there is no inflation or deflation and an upward adjustment if there is inflation as shown below:

Face Value and Face Value*(RPI relating to February 2021/Base RPI).

United Utilities – United Utilities provides water and sewage services in the North West parts of England. To finance its activities, the company has issued several bonds in GBP, Euros and US dollars and is also publically traded on the London Stock Exchange. Its latest bond issuance was in 2006 when it issued GBP 50million 40 year, GBP 50million 50 year, GBP 100million 36 year, GBP 50million 37 year and GBP 100million 40 year linkers. Some of these issues were under the Euro Medium Term Notes Programme. Since the model used in these bonds is identical, only one of them is given as an example. Considering the GBP 100million 36 year linker maturing in 2042 giving a coupon of 1.5802% paid semi-annually (Anon, 2006). The denomination of these bonds was specified to be GBP 50,000 with the minimum tradable amount being GBP 1,000. The interest and redemption amount were linked to the index. As explained in the prospectus, the redemption principal will be calculated as the nominal amount of the bond multiplied by the index ratio applicable to the redemption month. The base index figure taken for this issue was the RPI figure for the month of issue (here March 2006) and the index ratio would be calculated as the index figure of the given month divided by the base index figure. The interest payable would be calculated as 1.5802% per annum multiplied by the index ratio in the month of interest payment. The interest rate of 1.5802% includes the credit spread for the issue which has a credit rating of A- with S&P and A2 with Moody’s. The prospectus also mentions the reference gilt used was the 2% linker treasury stock due in 2035. Another important point mentioned in the prospectus is about any changes in the index due to changes in the base year will lead to a change in the base index to be calculated as the product of the index applicable to the month of issue (I) and the index immediately following the change (Inew) divided by the index value before the change (Iold).

New base index = (I x Inew)/Iold

Although the prospectus does not clearly define a deflation floor, it mentions that if there is a change in the index which is a disadvantage to the issuer, it has the right to redeem the principal amount at its nominal value. This clause is as good as having a deflation floor.

Severn Trent – Severn Trent provides water, treats water and sewage across UK and internationally. It is listed on the London Stock Exchange and is a FTSE 100 company. The company recently issued a 10 year inflation linked bond offering an interest rate of 1.3% per annum adjusted at the time of payment for inflation payable semi-annually maturing in July 2022 (Anon, 2012). This issue also follows the same model as National Grid and United Utilities. The prospectus also mentions that the index is announced with an eight month lag and this value is used for calculating the index ratio as shown in previous calculations.

Pennon Group – Pennon Group consists of two core business subsidiaries, South West Water and Viridor. South West Water provides water and sewage services across Devon, Cornwall and parts of Dorset and Somerset. Viridor started as a traditional waste management company but has evolved into a recycling, renewable energy and resource management company. South West Water finances show that 24% of their debt is index-linked (Anon, 2011 http://www.southwestwater.co.uk/media/pdf/i/a/2011_Regs_FINAL_inc_Cover.pdf). Their current index linked bonds mature in 2057 and pay RPI+1.99% with a total principal amount of GBP 220.2million. The prospectus of this bond issue was not available hence no information about the principal repayment and deflation floor is available.

Once the correlation between the UK companies’ revenue and RPI is established, the next part involves collecting data for the Indian companies and the WPI. The WPI data available from the Ministry of Commerce is limited to financial year 2005-2006 to 2011-2012. Following this, the annual data for companies are collected for the same time period from their annual reports.

Result

A preliminary step toward carrying out this study is to first establish concrete evidence that revenue and RPI in the UK are positively correlated. A sample of four companies is taken – National Grid, United Utilities, Severn Trent and Pennon Group. These companies have been chosen since the former three have inflation linked bonds in issue while the last company does not, providing a contrast to the others. From the data available from 2006 to 2012, the correlation study shows a high positive correlation between RPI and the revenues of three of the four companies. United Utilities shows a negative correlation although it is also regulated as are the remaining companies.

Correlation with RPI

 

RPI

National Grid

0.541

Pennon Group

0.967

United Utilities

-0.755

Severn Trent

0.904

Table 2 – Correlation between UK utility companies and RPI

Establishing the correlation between inflation and revenues of the selected utility companies in India is the next step. The period from 2005-2006 to 2011-2012 was taken for this analysis. Each company tested showed a very high positive correlation of more than with the WPI index. As compared to the UK utility companies, Indian companies have a much higher correlation to WPI.

Correlation with WPI

 

WPI

Cairn Energy

0.923

Essar Oil

0.942

GAIL

0.994

Indraprastha Gas

0.962

MRPL

0.923

NTPC

0.989

ONGC

0.923

PFC

0.997

Power Grid

0.990

PTC

0.879

REC

0.996

Tata Power

0.994

Table 3 – Correlation between Indian utility companies and WPI

The Granger Causality test is performed between each revenue stream and WPI. The results show that at 1% and 5% significance levels only bidirectional causality is seen which indicates contemporaneous feedback, that is, no establishment of either series preceding the other.

With one period lag

 

F-Statistic

Probability

Interpretation

WPI does not Granger Cause Cairn Energy

2.092

0.244

Since probability is above 0.05 (for a 95% confidence interval), the result is insignificant. Since this result is true for both directions, it is bidirectional and indicates contemporaneous feedback between the two time series

Cairn Energy does not Granger Cause WPI

0.0066

0.813

 

 

WPI does not Granger Cause Essar Oil

4.063

0.137

Since probability is above 0.05 (for a 95% confidence interval), the result is insignificant. Since this result is true for both directions, it is bidirectional and indicates contemporaneous feedback between the two time series

Essar Oil does not Granger Cause WPI

0.8

0.437

 

 

WPI does not Granger Cause GAIL

2.16

0.238

Since probability is above 0.05 (for a 95% confidence interval), the result is insignificant. Since this result is true for both directions, it is bidirectional and indicates contemporaneous feedback between the two time series

GAIL does not Granger Cause WPI

0.422

0.562

 

 

WPI does not Granger Cause Indraprastha Gas

0.289

0.628

Since probability is above 0.05 (for a 95% confidence interval), the result is insignificant. Since this result is true for both directions, it is bidirectional and indicates contemporaneous feedback between the two time series

Indraprastha Gas does not Granger Cause WPI

0.604

0.493

 

 

WPI does not Granger Cause MRPL

4.127

0.135

Since probability is above 0.05 (for a 95% confidence interval), the result is insignificant. Since this result is true for both directions, it is bidirectional and indicates contemporaneous feedback between the two time series

MRPL does not Granger Cause WPI

3.754

0.148

 

 

WPI does not Granger Cause NTPC

8.464

0.062

Since probability is above 0.05 (for a 95% confidence interval), the result is insignificant. Since this result is true for both directions, it is bidirectional and indicates contemporaneous feedback between the two time series. However, at 90% confidence interval, the result becomes significant for WPI Granger cause NTPC. Here WPI Granger Cause NTPC since the F-statistic is large.

NTPC does not Granger Cause WPI

0.003

0.956

 

 

WPI does not Granger Cause ONGC

6.147

0.089

Since probability is above 0.05 (for a 95% confidence interval), the result is insignificant. Since this result is true for both directions, it is bidirectional and indicates contemporaneous feedback between the two time series. However, at 90% confidence interval, the result becomes significant for WPI Granger cause ONGC. Here WPI Granger Cause ONGC since the F-statistic is large.

ONGC does not Granger Cause WPI

0.3

0.622

 

 

WPI does not Granger Cause PFC

0.035

0.864

Since probability is above 0.05 (for a 95% confidence interval), the result is insignificant. Since this result is true for both directions, it is bidirectional and indicates contemporaneous feedback between the two time series. However, at 90% confidence interval, the result becomes significant for PFC Granger cause WPI. Here, PFC Granger Cause WPI since the F-statistic is large.

PFC does not Granger Cause WPI

5.987

0.092

 

 

WPI does not Granger Cause Power Grid

4.673

0.119

Since probability is above 0.05 (for a 95% confidence interval), the result is insignificant. Since this result is true for both directions, it is bidirectional and indicates contemporaneous feedback between the two time series

Power Grid does not Granger Cause WPI

0.054

0.83

 

 

WPI does not Granger Cause PTC

0.023

0.888

Since probability is above 0.05 (for a 95% confidence interval), the result is insignificant. Since this result is true for both directions, it is bidirectional and indicates contemporaneous feedback between the two time series

PTC does not Granger Cause WPI

0.013

0.916

 

 

WPI does not Granger Cause REC

3.526

0.157

Since probability is above 0.05 (for a 95% confidence interval), the result is insignificant. Since this result is true for both directions, it is bidirectional and indicates contemporaneous feedback between the two time series

REC does not Granger Cause WPI

2.631

0.203

 

 

WPI does not Granger Cause Tata Power

43.236

0.007

Since probability is above 0.05 (for a 95% confidence interval), the result is insignificant. Since this result is true for both directions, it is bidirectional and indicates contemporaneous feedback between the two time series. However, at 90% confidence interval, the result becomes significant for WPI Granger cause Tata Power. Here, WPI Granger cause Tata Power since the F-statistic is large.

Tata Power does not Granger Cause WPI

0.6

0.495

Table 4 – Granger Causality test

The next test called the Augmented Dickey-Fuller test shows that the raw data (level) is non-stationary at all levels (1% and 5%) of significance.

 

 

t-critical at 1%

t-critical at 5%

t-value

p-value

Interpretation

Cairn Energy

 

 

Level

-5.119

-3.519

0.182

0.94

Non-stationary

Essar Oil

 

 

Level

-5.119

-3.519

-0.352

0.857

Non-stationary

GAIL

 

 

Level

-5.605

-3.695

1.893

0.996

Non-stationary

Indraprastha Gas

 

 

Level

-5.605

-3.695

2.415

0.998

Non-stationary

MRPL

 

 

Level

-5.119

-3.519

0.151

0.937

Non-stationary

NTPC

 

 

Level

-5.119

-3.519

-0.163

0.893

Non-stationary

ONGC

 

 

Level

-5.605

-3.695

1.3

0.989

Non-stationary

PFC

 

 

Level

-5.119

-3.519

8.449

1

Non-stationary

Power Grid

 

 

Level

-5.119

-3.519

0.823

0.981

Non-stationary

PTC

 

 

Level

-5.119

-3.519

-1.06

0.652

Non-stationary

REC

 

 

Level

-5.119

-3.519

4.392

0.999

Non-stationary

Tata Power

 

 

Level

-5.119

-3.519

1.204

0.991

Non-stationary

WPI

 

 

Level

-5.605

-3.695

2.462

0.998

Non-stationary

Table 5 – Augmented Dickey Fuller Test for Non-Stationarity

The final step is to conduct the regression analysis of the dependent term (revenues) and the independent term (WPI). Several regression methods were attempted but the ARMA method was finally chosen as the best fit.

The first regression analysis considered was the co-integrated regressive model. For this the Engle-Granger test was conducted. The probability values for Engle-Granger test showed that the null hypothesis could not be rejected at 5% significance level. The exception to this was REC where the t-statistic showed significant evidence that there is co-integration at 5% significance level, however, the normalised autocorrelation value (z-statistic) does not agree. Due to this discrepancy, the result is inconclusive. Since none of the tests showed conclusive significant evidence that there is co-integration, the co-integrated regression model was not applicable.

 

WPI

 

tau-statistic

Probability

z-statistic

Probability

Cairn Energy

-2.02

0.551

-4.626

0.722

Essar Oil

-2.374

0.414

-6.681

0.363

GAIL

-3.214

0.203

-6.618

0.352

Indraprastha Gas

-1.763

0.663

-3.349

0.838

MRPL

-1.751

0.669

-4.552

0.739

NTPC

-3.352

0.162

-7.053

0.231

ONGC

-1.748

0.67

-4.076

0.763

PFC

-4.363

0.059

-8.448

0.122

Power Grid

-2.787

0.281

-6.777

0.354

PTC

-1.386

0.805

-6.629

0.353

REC

-5.519

0.02

-8.898

0.098

Tata Power

-1.804

0.645

-4.784

0.022

Table 6 – Engle-Granger test for co-integration

Next, the possibility of using ARCH/GARCH models was looked into. For this purpose, the volatility is used by calculating the log of ratio of the values, that is, log [value at time t/value at time (t-1)]. This step compromises the number of observations further. Also, the results achieved were not significant at 5% significance level and many gave negative R^2 and Adjusted R^2 values showing that the model was not applicable.

WPI

Residual(-1)^2

GARCH(-1)

 

z-statistic

Probability

z-statistic

Probability

z-statistic

Probability

Adjusted R^2

Cairn Energy

0.021

0.984

-0.143

0.886

1.018

0.308

-0.459

Essar Oil

0.259

0.796

-0.119

0.905

0.082

0.935

-0.154

GAIL

1.526

0.127

-0.129

0.897

0.249

0.8

0.738

Indraprastha Gas

1.134

0.257

-0.062

0.95

0.051

0.959

0.35

MRPL

0.726

0.468

-0.04

0.968

0.036

0.971

0.532

NTPC

0.17

0.865

-0.206

0.837

1.606

0.108

-0.351

ONGC

0.797

0.425

-0.423

0.672

0.818

0.413

-0.495

PFC

0.803

0.422

-0.049

0.961

0.147

0.883

0.009

Power Grid

0.784

0.433

0.052

0.958

1.259

0.208

-0.143

PTC

-0.063

0.95

-0.238

0.811

0.184

0.853

-0.253

REC

-0.045

0.964

0.044

0.964

0.061

0.951

-0.238

Tata Power

0.363

0.717

-0.049

0.961

0.04

0.968

-0.371

Table 7 – GARCH model

Another approach was the distributed lag model was used. The values do not show significance at 5% levels with the exception of GAIL, NTPC, PFC and REC, with the former three showing significance only for WPI while the last showing significance for WPI and WPI(-1).

 

WPI

WPI(-1)

 

t-statistic

Probability

t-statistic

Probability

Cairn Energy

2.254

0.11

-1.434

0.247

Essar Oil

0.679

0.546

-0.002

0.999

GAIL

6.859

0.006

-1.855

0.161

Indraprastha Gas

1.664

0.195

-0.502

0.65

MRPL

1.234

0.305

-0.709

0.529

NTPC

4.151

0.025

1.702

0.187

ONGC

0.966

0.405

-0.155

0.886

PFC

5.702

0.011

2.594

0.08

Power Grid

1.538

0.221

0.38

0.729

PTC

0.158

0.885

0.242

0.824

REC

350878.8

0

270640.9

0

Tata Power

0.138

0.899

0.407

0.711

Table 8 – Distributed Lag model

Finally, the Autoregressive Moving Average model was chosen wherein WPI has no lag while an autoregressive term (AR(1)) is added. This model is optimum as confirmed by the Theil Inequality Coefficient and also with having no negative R^2 values and the coefficients are not extremely high.<