Project Finance and Political Risk - AU Purepure.au.dk/portal/files/3002/Sara_Toft_THESIS_aug_2008.pdf · finance, in which more risk leads to more syndicated loans being given as

Project Finance and Political Risk

- An empirical study of the relationship between project finance and

political risk

Author: Sara Møller Toft MSc. Finance and International Business

Academic adviser: Christian Bjørnskov Department of Economics Århus School of Business Århus University August 2008

Abstract:

Christa Hainz and Stephanie Kleimeier (2006) have developed a model of double moral hazard that

helps explain the use of project finance in countries with high political risk. A phenomenon that

seems odd given that project finance is typically given as non-recourse loans. However, their model

lacks a proper clarity of political risk as 6 components have been aggregated into one measure.

Thus, the study presented here utilizes the 12 components of the ICRG’s political risk index

grouped by a principal component analysis into three components. It can then be shown that

different aspects of political risk have different effects in the model of double moral hazard. More

specifically, only the risk of poor quality of institutions have the claimed positive effect on project

finance, in which more risk leads to more syndicated loans being given as project finance deals.

Policy inferiority on the other hand will lead to fewer project finance loans given and war, tensions

and other conflict have no effect on upon the fraction of loans given as project finance loans. The

model of double moral hazard thus lacks a more complex structure to explain the use of project

finance loans especially as it is shown that the effect of the political risk components also differs

between income levels.

Århus School of Business Sara Møller Toft MSc. Finance and International Business Political risk and the use of Project Finance

Table of Content 1. Introduction .............................................................................................2

2. Problem statement...................................................................................4

3. Delimitations ............................................................................................4

4. Methodology of the paper.......................................................................5

5. Theoretical framework ...........................................................................6

5.1 Project finance....................................................................................................... 6 5.1.1 Project finance defined.................................................................................... 7 5.1.2 Risks of project failure................................................................................... 12 5.1.3 Summary ........................................................................................................ 16

5.2 Political risk ......................................................................................................... 16 5.2.1 Political risk defined...................................................................................... 16 5.2.2 Previous literature on political risk variables............................................... 18

5.3 The model of double moral hazard ..................................................................... 23 5.3.1 Moral hazard defined .................................................................................... 23 5.3.2 The model of double moral hazard................................................................24 5.3.3 Measurement of political risk in Hainz and Kleimeier (2006)...................... 27 5.3.4 Measuring political risk in the model of double moral hazard ..................... 29

6. Political risk analysis.............................................................................34

6.1 Methodology......................................................................................................... 34

6.2 Data ...................................................................................................................... 35

6.3 Empirical evidence .............................................................................................. 35 6.3.1 Assumptions................................................................................................... 35 6.3.2 Results............................................................................................................ 38

7. Project finance and political risk analysis ..........................................46

7.1 Methodology......................................................................................................... 46 7.1.1 The model ...................................................................................................... 46 7.1.2 Method........................................................................................................... 48

7.2 Data ...................................................................................................................... 50

7.3 Empirical evidence .............................................................................................. 53

8. Conclusion..............................................................................................65

9. Critical perspective ...............................................................................67

Bibliography ..............................................................................................69 Appendix 1 ................................................................................................................. 72

Appendix 2 ................................................................................................................. 73

Appendix 3 ................................................................................................................. 74

Appendix 4 ................................................................................................................. 78

Appendix 5 ................................................................................................................. 79


1

Appendix 6 ................................................................................................................. 80

Appendix 7 ................................................................................................................. 82

Appendix 8 ................................................................................................................. 85

Appendix 9 ................................................................................................................. 86

Appendix 10 ............................................................................................................... 89

Appendix 11 ............................................................................................................... 90

Appendix 12 ............................................................................................................... 91


2

1. Introduction The use of project finance has been growing through out the past 20 years. It is a

financing method in which the project is set up as a legal independent entity that is

mainly financed through non-recourse syndicated loans. This means that the sponsoring

firm behind the project has isolated their balance sheet from the risk of a possible failure

of the new project. One would thus think that if a syndicate were to grant such loans, it

would be for projects in stable and safe environments. However, the fact is that project

finance has become more used in countries with higher risk, as can be seen from the

graph below.

Figure 1.1: Risky countries has received more project finance over the years

Source: Data set collected on project finance deals from Reuters DealScan web database Note: Light green is high risk and grey is very high risk It can be seen that since 2004 more than 50% of the volume of project finance deals is

given to projects in countries with moderate to high risk. However, the question remains

as to why it is efficient to use project finance in such countries, when such loans

encompass limited or non-recourse debt? Would this not make the loan even more risky

remembering that the project is already in a risky country?

First of all, it is important to note that project finance encompasses many contracts with

the different parties involved, which helps reduce the level of uncertainty in the project

and hence the level of risk. On the other hand, high risk countries are often lacking legal

systems that protect property rights properly.

0 %

1 0 %

2 0 %

3 0 %

4 0 %

5 0 %

6 0 %

7 0 %

8 0 %

9 0 %

1 0 0 %

1 9 8 8 1 9 9 0 1 9 9 2 1 9 9 4 1 9 9 6 1 9 9 8 2 0 0 0

V e ry H ig h

H ig h

2 0 0 4 2 0 0 6

V e ry L o w

L o w

M o d e ra te

2 0 0 2

R is k

D is tr ib u t io n o f v o lu m e o f p ro je c t f in a n c e a c c o rd in g to R IS K


3

Hainz and Kleimier (2006) has suggested a model of double moral hazard that explains

the trade of between the moral hazard of the firm and that of the bank and how this trade

of can lead to the use of non-recourse loans. Large banks or syndicates of banks can

influence governments and hence reduce the level of political risk. But they need an

incentive to do so and non-recourse loans will provide this whereas ordinary full-

recourse loans will give the firm an incentive to act and reduce their moral hazard.

Project finance is about assigning risk to those parties best able to bear them, which

may explain the reasoning behind having non-recourse debt as an incentive for banks to

help deter political risk.

However, the question is what political risk factors are determining the use of project

finance and are some more important than others? A large amount of research and

literature have looked into how different political risk variables may affect economic

growth of a country or the level of investment in that country. It has been found that

democracy or variables of political violence does not capture the real political risk to the

investors and that researchers should instead concentrate on the variables that look into

the properties of the political system that are relevant to private sector-driven growth.

Hence, when looking at the political variables that could help determine if project

finance is given (as non-recourse loan), it must be variables that are important to the

investment decision, the success of a project and that are under the influence of a bank.

Hainz and Kleimeier (2006) have used the World Governance Indicators as proxy

variables for political risk. However, these have been criticized by Iqbal and Shah

(2008) and at the same time Hainz and Kleimeier (2006) have aggregated all measures

of political risk into one variable. As political risk covers several aspects, such an

aggregation means that it cannot be seen which aspects that does indeed affect the use

of project finance.

The main purpose of this paper is therefore not only, to find appropriate measures of

political risk, but also to combine these measures into different aspects of political risk

and subsequently use these in the model of double moral hazard for explaining the use

of project finance.


4

2. Problem statement In order to properly use the model of double moral hazard for explaining the use of

project finance loans in high risk countries several questions must be answered:

• How could project finance loans be affected by risk and especially political risk?

• How do we measure political risk?

• Can we group the political risk variables into components that are theoretically

sound and have an influence on the use of project finance?

The answers to the above mentioned questions will assist in answering the main

problem of the paper:

• Using these new measures of political risk along with the bank influence and

firm moral hazard variables, what is the influence of political risk on the use of

project finance and is there still an empirical ground for the double moral hazard

model proposed by Hainz and Kleimeier (2006).

3. Delimitations Due to time and size constraints this paper will mainly be focusing on the political risk

aspects of the model of double moral hazard. Hence, the firm moral hazard part is

explained but the measurement of it is not discussed as to whether it is sound. That

same applies to bank influence.

One of the advantages of a firm taking on debt is tax benefits. These may vary from

country to country and could be a partial explanation to why some countries succeed in

attracting more finance than others. This will however not be considered in this paper.


5

4. Methodology of the paper This paper is divided into three parts which will answer the problems set forth in the

problem statement. The first part is a theoretical part. This part of the paper will be

looking into the three theoretical areas that the paper covers namely project finance,

political risk and moral hazard. In order to test the model of double moral hazard that

explains the level of project finance given to a country it must be defined what project

finance is and what differentiates it from ordinary financing methods. That is what

section one will discuss. The second section will be looking at how one should measure

the political risk of a country in order to find a measure that can be used in the model of

double moral hazard. The last section of the theoretical part will look into the

definitions of moral hazard in order to give a theoretical explanation of the model of

double moral hazard. The short comings of the used proxies for political risk in the

paper by Hainz and Kleimeier (2006) will be discussed and a section will follow

describing the proxies of political risk to be used in this paper and the relevance of these

in relation to what was found on project risks.

The second part of the paper will by the use of a principal component analysis combine

several measures of political risk into components that are uncorrelated and covers

different aspects of the political risk relevant to an investor for use in the model of

double moral hazard.

The last section will then use these political components derived in the previous part to

test the model of double moral hazard on the use of project finance in order to establish

whether political risk is a determinant of the use of project finance.


6

5. Theoretical framework

This part of the paper will review literature on project finance, political risk and briefly

on moral hazard in order to properly identify what political risk is and how it relates to

project finance and thus how to best measure it in the model of double moral hazard.

The section will start out be explaining the concept of project finance as it is important

to understand its basic features in order to understand the model of double moral hazard

and how political risk might relate to project finance. Later a section on previous

measurement of political risk will follow in order to determine which measures are most

adequate in measuring the different aspects of political risk and at the end the model of

double moral hazard will be explained along with the main problems of the Hainz and

Kleimeier (2006) test of it and the corrections this paper will make.

5.1 Project finance Not only has more project finance loans been given over the years but there is also an

increasing usage of these in lower income countries and countries with high political

risk. This can be seen from figure 5.1 below.

Figure 5.1: Average share of loans made as project finance loans over the 4 periods1.

1 These figures are based on the volume of loans. However, they do look the same for the number of loans given. See appendix 1 for these.

1625

46

11

19

15

26

28

17

42

35

87-05

53

32

96

87-90

41

54

129

91-95

29

40

110

96-00

45

36

01-05

Very low

Low

Moderate

High

46 49

72

11

21

14

29

53

19

37

29

87-05

32

53

87-90

43

38

91-95

25

39

96-00

36

34

01-05

High

Upper middle

Lower middle

Low

Source: Reuters DealScan web database Note: The figure on the left is across different risk levels and the figure on the right is for income levels. The last column of both figures is an average for the entire period. Both figures have been made for the volume of loans.

Risk Income


7

Both for the risk groups and the income groups the share of loans given to high risk and

low income countries has increase from 1990 to 2005. The last column is the average

for the entire period of which it can be seen that the average share of project finance

loans out of total syndicated loans for the low income group is higher than for the other

income groups. It might be that low income countries receive less loans overall due to

less efficient financial markets and more risk, however, a larger share of those loans are

given at project finance loans.

But what is project finance? That is what this section will look into. In order to

investigate the risks faced by borrowers and lenders it must firstly be described what is

understood by project finance loan deals, the participants and the possible failures of it.

5.1.1 Project finance defined Project finance is defined slightly different by different authors. Esty (2004) defines it

as: “Project finance involves the creation of a legally independent project company

financed with equity from one or more sponsor firms and non-recourse debt for the

purpose of investing in a capital asset” (Esty, 2004 pp.25), while Finnery (2007) says:

“Project finance is raising funds on a limited-recourse or non-recourse basis to finance

an economically separable capital investment project in which the providers of the

funds look primarily to the cash flow from the project as the source of funds to service

their loans and provide the return of and the return on their equity invested in the

project” (Finnery, 2007 pp.1). Finally, Yescombe (2002) defines it in the following

manner: “Project finance is a method of raising long-term debt financing for major

projects through “financial engineering”, based on lending against the cash flows

generated by the project alone; it depends on a detailed evaluation of a project’s

construction, operating and revenue risks and their allocation between investors,

lenders and other parties through contractual and other arrangements” (Yescombe,

2002 pp.1).

There are hence several aspects of project finance. Most importantly is the fact that

when using project finance the project is set-up as a legally separate entity2 and the non-

recourse debt can only be paid from the cash flows generated by the project itself. Using

non-recourse or limited recourse financing is at the heart of project finance. The reason

2 Also referred to as a separate purpose vehicle - SPV


8

is that in many cases the size of the project may be larger than the size of the

participating companies’ balance sheet (Fight, 2006). Therefore, project finance differs

from corporate finance, also called on-balance-sheet financing, in the way that in

corporate financing the repayment/servicing of the loan is backed by the sponsoring

firms’ entire balance sheet and thus all their projects and internally generated cash flows

(Ahmed, 1999). However, when using project finance the debt given becomes non-

recourse (or limited-recourse), as the repayment cannot be paid by cash flows generated

within the sponsoring company3. If the project fails it will therefore not affect the

sponsoring firm’s balance-sheet. Project finance is therefore a way of protecting the

corporate balance sheet from the incremental distress costs of a failing project (Esty,

2004)4.

It should also be noted that off-balance sheet does not mean completely self-supporting

entity with no guarantees or credit support from financially credible third parties.

Contracts granting guarantees and credit support from third parties makes the project

financing possible by adding up to a satisfactory credit risk for the lenders (Nevitt,

2000).

Participants The legal separation of a project requires that the facility or assets can function

profitably as an independent economic entity. To do so, the operation of the project is

supported by several contractual arrangements in order for the project to have the

incontestable ability to generate adequate cash flows to service its debt. This is why

project finance is also at times referred to as “contract finance”. These contracts will

help in forecasting the future cash flows on which the repayment of the debt is

contingent, and with all of these contracts a major part of project finance also becomes

the allocation of risks to the parties best able to bear them. (Finnerty, 2007)

Figure 5.2 shows a simplified picture of project finance including some of the many

contracts and participants involved in the project.

3 Or only under certain agreed circumstances – e.g. pre-completion guarantee (Ahmed, 1999) 4 It should however, be noted that for many financial accounting standards a project must be included in the footnotes of the sponsoring firm’s balance sheet which means that rating agencies and investors can include this information in their judgment of the sponsoring firms securities (Finnerty, 2007). This means that if the project fails, it will to some degree affect the company – although not directly by drawing on the sponsoring firms resources to repay the loan, then indirectly by possibly affecting the credit rating of the firm.


9

Note: The SPV is the special purpose vehicle – the project. These are the main participants and their relation to the project. The sponsor is the firm initiating the project and will manage the project. The borrower

is, however the separate entity – the special purpose vehicle, which helps the separation

of the balance sheets.

The lenders are often a syndicate of banks. This happens because the loan amount is too

large for one bank alone to take on. It is here important to note that although project

finance does bear certain similarities to syndicated lending, there are some issues that

make it different; these being the non-recourse/limited-recourse debt and the separate

legal entity (Fight, 2006). The syndicated can be of several classes of lenders from

international banks to local banks, export credit agencies and international agencies

lending or guaranteeing development credit (Fight, 2006). Projects also tend to be

highly leveraged. The average book value debt-to-total capitalization ratio is 70% (Esty,

2004), which means that lenders supply a large amount of the project financing and are

thus highly concerned with the risks of the project. However, this depends on the

strength of the contracts (especially the purchase contract) and the future prospect of the

project.

Figure 5.2: Project Finance participants

Participants in a project finance deal

Source: Finnerty 2007

SPVSupply

Contract(s)

Raw material

Output

Purchase Constract(s)

Equity

funds

Return to

investors

Debt

repaym

ent

Loan funds

Host

Government

Suppliers

Lenders

Buyers

Sponsors

Credit

support


10

Contracts are also made with suppliers as well as purchasers. The contracts with the

suppliers are very important to the success of the project as they supply the raw material

to be input in the project and thus needed in order to produce the outcome that will

generate the cash flows that services the debt. This means that the lenders and sponsors

are concerned with the economic feasibility of the supplier and whether the supplier will

be able to live up to the agreement. Of the same importance is the contract with the

purchaser. Many times the project sponsors will try to get a long term arrangement with

a buyer in order to ensure more steady cash in flows. These contracts can also be used

as a security to the lenders (Fight, 2006).

There might also be different contracts for the development of the project and the

operation and management of the project. These can be of different forms but are

mainly to secure the constructions and development phase of the project and the

operational phase following. Besides these contracts there are contracts made with the

host government in order to obtain the licenses, permits and authorizations needed to

operate the project. Thus all in all an extensive amount of contracts are drawn for each

deal.

Contracts and legal environment According to Nevitt (2000) “The key to a successful project is structuring the financing

of the project with as little recourse as possible to the sponsor, while at the same time

providing sufficient credit support through guarantees or undertakings of the sponsor

or third party, so that lenders will be satisfied with the credit risk” (Nevitt, 2000 pp.2).

This means that the many contracts undertaken are indeed done to allocate the risk so

that the lenders will be satisfied with the risk faced by giving the loan to the project.

However, the many contracts made stresses the importance of a supportive legal and

regulatory framework in which the contracts can be enforced. Esty (2004) states the

following: “Typically risk allocation is done through long-term contracts. Invariable,

however, these contracts ends up being “incomplete”: it is either impossible or too

costly to describe all possible contingencies, and it may be costly to enforce them as

well. When the contracts are incomplete, there is a potential for incentive conflicts

between key decision makers. Some of the most important, yet most incomplete,

contracts are those with host nations. In this context, contractual incompleteness is due

to an inability to enforce contracts rather than unforeseen or unspecified situations.

When contracts are unenforceable, property rights are uncertain, and if project cash


11

flows are high, there is a serious risk of expropriation. Because so many projects are

located in countries with non-existent, untested, or unpredictable legal systems,

managing sovereign risk becomes one of the most critical tasks for project managers”

(Esty, 2004 p.10). Thus even though there are contract to govern most of the operation

of the project in order to ensure the cash flow, there is the risk of unenforceable

contracts, which may render the project to fail partly by changing the cash flows or

entirely by making the project unable to operate. This risk must be mitigated by means

of an insurance coverage or by involving third parties that can help mitigate this risk by

their mere presents. These could be international development banks or agencies or

countries that the host government does not want to jeopardize their economic relations

with (Fight, 2006).

An advantage of project finance is therefore the organisational approach to risk

management by the sponsoring firm, by separating the project from their balance sheet

and taking limited or non-recourse loans (Esty, 2004). This is an important advantage as

it helps the underinvestment problem that might arise based on risk aversion in high risk

countries and risk aversion due to the size of the project. If the project could yield

positive NPV’s but it is placed in a country that due to uncertainties creates high risk to

the project, managers may not want to take on the project out of fear of failure due to

the high political/sovereign/country risk. If the project is a large scale project, taking it

on by itself may also pose to great risk for the firm and hence project finance not only

sheds the sponsoring firm from some of the risk it also provides risk sharing through

allocating risk to the parties best able to bear them and thus no participating party

assumes all the risk on its own.

Tinsley (2000) argues that an important advantage is the risk transfer in relation to

political risk. He states “Even large companies facing political risk will use a project

financier as a way to get political risk cover on the debt side of the project. Roughly

half of all project financings are to secure political risk coverage” (Tinsley, 2000 p.7).

This can be either by isolating the risk of the political uncertainty inside the project and

hence away from the corporate balance sheet of the sponsoring firm, or by including

many large and international lenders and participants in the project. The last option will

help abstain the host government from interfering with the project and jeopardizing the

probability of success of the project.


12

Combining this with the last advantage mentioned by Fight (2006); the advantage that

“Lenders are more likely to participate in a workout than foreclose” (Fight 2006, pp.6),

shows that the use of project finance with the non-recourse or limited recourse loan

gives the lenders an incentive to rather workout problems than see the project fail. This

also means that if the lenders have any influence over the host government they will be

able to shed some political risk by participating in the syndicated loan and by exercising

this influence to help the probability of success of the project if necessary. This is the

issue that the double moral hazard model by Hainz and Kleimeier (2006) addresses and

that this paper will be looking further into.

5.1.2 Risks of project failure The importance of risk to project finance has been stated several times. There are no

complete definitions of risk, but it is often considered in relation to uncertainty and

incomplete information. An uncertain or unknown future event can have a potential

negative impact on some characteristic of value/the business and that is what is most

often understood by risk. In relation to project finance it is as mentioned very important

to identify, assess and allocate the risks. As the lenders place a large degree of reliance

on the performance of the project, they will naturally be concerned with the feasibility

of the project and its sensitivity to possible events affecting it negatively. In order for

this paper to later on discuss the political risk factors that can affect a project and how

they should be measured, it is important to discuss some of the main reasons why a

project may fail to produce the expected cash flows. Hence, this section will merely be

looking into the negative impact that may occur.

As each project is different, there are no complete lists of the risks involved or a clear

division of those or the causes for failure in a project. Yescombe (2002) divides risk

into tree main categories:

• Commercial risk – those inherent in the project itself or the market it operates in

• Macro-economic risks – financial risks or economic effects not directly related

to the project


13

• Political risk – related to the effects of the government actions or political force

majeure events5. (Yescombe 2002)

These three areas will be able to capture all the different risk associated with a project.

In itself, these three classifications are not enough according to Tinsley (2000) who

argues that the focus should always be on the relation of risk to cash flow changes. And

as such these three classifications are to diffuse and do not provide any guidance on the

structuring of the risk or approaching the risk identification (Tinsley, 2000). However,

Yescombe (2002) does actually further elaborates on the understanding of the three

categories by subdividing it into smaller risk components which can be seen in figure

5.3.6

“Commercial risk” captures the risks associated with the project itself and covers the

risks that are associated with the different phases of the project: construction and

operation. During the construction phase there is the risk of financial failure of the

contractor (Nevitt, 2000) meaning that if he goes bankrupt then the project is at risk of

not being complete or being delayed, which will increase the costs and postpone the

revenue stream. The other risks in this phase are the risk of not obtaining adequate

5 Force majeure is considered any event outside the control of the parties. These events are acts of man, nature, government/regulators, or impersonal events. Contract performance is forgiven or extended be the period of force majeure (Tinsley 2000 p.276) 6 A further break down of the risk components into issues under the risks can be found in appendix 2

Risk components Sub-componentsCommercial viability

Completion risksEnvironmental risks

Operating risksRevenue risks

Input supply risksForce majeure risks

InflationInterest rate risks

Exchange rate risksInvestment risks

Change of law risks Quasi-political risks

Commercial risks

Macro-economic risks

Political risks

Figure 5.3: Yescombe’s risk matrix with sub-components

Source: Yescombe (2002)


14

permits and sites, which is granted by the host government and the risk of costs overrun,

which increase the overall cost of the project and hence the viability of the project. The

operation phase contains risk such as management of the operation, technological

failure or obsolescence, input risks and revenue risks. The input risks relates in

particular to the price and supply of the raw materials which could be hampered either

by market forces or actions taken by the host government or governments of countries

in which the raw material is extracted/produces or processed. These actions may cause

the production to either slow down or become more expensive, which in both cases

changes the expected cash flows and thus affect the debt repayment. Revenues can be

divided into quantities and price and both factors can be affected by several other risk

factors and participants. The quantities can decline based on demand decline, lack of

raw materials, new competition, Decline in the quality of the product, delay of the

production site, force majeure and legal disputes (Tinsley, 2000). In the same way price

can be affected by many different risks and parties. Competitors may lower prices or the

government may impose price controls, tariffs or royalties. Input and revenue risks are

tried secured trough contracts with suppliers and buyers but again such contracts

requires good legal systems and no government interference.

Yescombe (2002) has also classified financial risks as being those associated with

economic or financial effect not directly related to the project. These are inflation,

interest rates and exchange rate risks. Such risks that have not been accounted for

properly in the cash flow projections can always change the expected cash flows if they

turn out to be higher and thus hamper the viability of the project.

The political risks discussed here are those that writers of project finance literature

views as being considered political risk. Yescombe’s (2002) political risk component

contains three risk factors. The first is investment risks, which relates to currency

convertibility and transfer a component, which has to do with government restricting the

outflow of money from the country and converting the cash flows into other currencies.

This is often necessary as the loan taken on the project is denominated in other

currencies than the one of the host country. Developing markets often have poor

financial markets that are not capable of providing the funds needed for the project, in

which case the loan is denominated in foreign currency. Investment risk also includes

the risk of expropriation of the investment by the government and war or internal and


15

external conflicts, which makes the project unable to function properly or entirely. The

second component of Yescombe’s (2002) political risk component is “change of law”

which include factors like: price controls, withdrawal of permits, licences or

concessions, deregulation of the market introducing new competitors, increases in tax,

tariffs, import duties or controls. All of which could create some of the input or revenue

risks discussed previously. It also includes new rules and requirements on

environmental issues, safety, health and employment. All of these can interfere with the

operation of the project and the primary political risk is in government interference by

changing the current setting in which the project operates. Under this is also creeping

expropriation, which is found in Yescombe’s (2002) last political risk component:

Quasi-political risk. This also includes “sub-sovereign” risk, the risk that lower levels of

officials interfering with the projects viability, and breach of contract which

incorporates the risk of the host government not honouring their obligations or the legal

system not being objective. This risk of the legal system not providing objective ruling

is also mentioned earlier as the many contracts in a project finance deal depends on the

legal system of the host country.

As can be seen from the examination of the different risk that may cause a project to fail

in servicing its debt based on cash flow changes, many of the risks are related to

governmental interference. On an overall level these can also be described under the

following three head lines:

• Risks from changes of legislation

o Price controls, permits, profit transfer, duties, taxes, deregulation, rules

• Risks from bad legal system and protection of property rights

o Expropriation (outright and creeping), breach of contract

• War and conflicts in the country.

o War and civil disturbance

It should be noted that governmental interferences may not only happen in developing

countries but are also a risk to be considered in developed countries where governments

may also impose new legislation, new regulations or new interpretations of existing

rules.


16

5.1.3 Summary Project finance is thus a financing method in which the project becomes a separate legal

entity that obtains limited- or non-recourse debt from a syndicate of lenders. The

payback can then only happen through the cash flows generated within the project

company. Project finance must thus be build upon many contracts with the different

participants of the project in order to share and allocate the risk. The risk that can make

a project fail can be divided into 3 categories: Commercial, Economic/Financial and

Political. The political risk is of particular interest as the interference of government can

influence many of the other risks that may cause the project to fail. At the same time the

influence of a particular bank being part of the non-recourse loan may help deter some

political risk by causing the bank to object if the host government jeopardise the

probability of success of the project. This along with the influence of political risk can

be tested through the model of double moral hazard proposed by Hainz and Kleimeier

(2006). In order to do so one must find ways to measure the risks just described. The

following section of the paper will be looking into political risk and previous literature

on how to measure it properly, in order to test the influence of political risk upon the

use of project finance.

5.2 Political risk There seems to be some correlation between the use of project finance and the level of

political risk in a country according to the previous discussions. However, in order to

test this relationship one must find appropriate ways of measuring political risk. This is

not an easy task. Many academics have discussed what measures to use when testing the

importance of political risk upon economic growth or investment. This section will

therefore try to establish which measures to use based upon previous findings in the

literature on which measures to use when reflecting the political risk that an investor

faces when investing domestic or foreign.

5.2.1 Political risk defined To measure political risk it must first be identified what is understood by political risk.

Finnerty (2007) writes: “Political risk involves the possibility that political authorities

in the host political jurisdiction might interfere with the timely development and/or

long-term economic viability of the project” (Finnerty, 2007 pp.82). This means, that

political risk is considered any action by the host government that can affect the project


17

negatively either by postponing the capital inflows or by directly affecting the

profitability of the project. This definition is in line with Howell & Chaddick (1994)

who states that “"Political risk" is the possibility that political decisions, events, or

conditions in a country, including those that might be referred to as social, will affect

the business environment such that investors will lose money or have a reduced profit

margin” (Howell & Chaddick, 1994 pp.71). They not only include the actual action but

also the possibility of an action taking place, which is the uncertainty that is usually

included in the notion of risk. The definition also includes both direct government

actions as equated by their decisions, indirect government actions as stated by events

and also the conditions in the country, which could affect all investors, making the

definition of political risk very broad. Conditions in a country could be civil unrest that

may lead to political instability or violence that could affect the operation of a

project/company in which the investor has invested.

Isham et al (1997) defines three interrelated dimensions of government action:

• What public decisions are taken,

• How public decisions and authority are exercised,

• How well public decision and authority are exercised (efficacy of government).

This definition states the importance of looking at government actions not only in

relation to what they do but how they do it and how well they do it. That means,

political risk is not only related to the actions government may take (eg. expropriation

of a foreign investment) but also to underlying structures of the political institutions that

defines the way the decisions are made and whether interfering actions is likely to be

made by the government (Democracy or autocracy). This aspect also includes the social

structures in a country that could be of determination to government actions. The last

notion of “efficacy of government” will also affect the level of political risk in a

country, as inefficient bureaucracy may interfere with the timely development of an

investment. It might also be that inefficient law enforcement may interfere with the

management of the investment.

Hence, political risk to a foreign investor includes many aspects. But how do we

measure the political risk of a country? The literature review in this section covers two

categories; the research done within identifying the political variables that can be used

when looking at the relation between political environment and economic growth and


18

another part of the empirical literate that tries to find a connection between foreign

direct investment and political risk, and the variables used in such a study. These are

somewhat related as classic growth theory states that economic growth is dependent on

savings and investment through accumulating capital (Ray, 1998).

5.2.2 Previous literature on political risk variabl es Some of the literature is summarised by Brunetti (1997) in Political variables in cross-

country growth analysis, where previous literature on political variables used in cross-

country growth analysis’ have been divided into 5 categories according to the

development over time. They developed from a very general measure of political

institutions to more specific properties of the political system:

1. Democracy

2. Government stability

3. Political violence

4. Policy stability

5. Subjective perception of politics

The first category is democracy, which was used in the earliest literature as a measure of

political risk. The main reason given for the use of democracy as a determinant of

economic growth is that most developed countries have democracy and that many of the

least developed countries does not have democracy and have been unsuccessful in

supporting growth. Democracy is often measured using the Gastil index for civil

liberties and political rights where the latter mainly concentrates on the election process

and whether these are subject to fair, competitive and clear procedure (Brunetti and

Weder 1995). However, no unambiguous and clear relationship between democracy and

growth has been found in the literature. Brunetti (1997) finds that the explanatory power

of democracy is very low in cross country growth studies.

Feng (2001), on the other hand, finds a positive relationship between democracy

measured as political freedom (civil liberty and political right from the Gastil index) and

investment. The positive relation is argued to be due to the fact that democracy requires

a broad support and consensus in the population to make the political process efficient


19

and secure. This based on the thought that an autocracy lacks the support and hence the

stability that an investor seeks when investing.

The second category, Government stability, refers to the number of or probability of

changes in government, keeping in mind that the changes must be fundamental and thus

lead to possible change or reversal of the rules and decisions set in place by the former

government. This is theoretically thought to affect growth and investment through the

fact that less stability will lead investors to invest in short-term assets. These are more

liquid and thus easier to divest in case the government should change and make changes

that affect the investment. This will lead to less accumulation of physical capital and

hence less growth (Feng, 2001).

In Brunetti and Weder’s article from 1995, it is concluded that “the political variables

used in these studies i.e. democracy and political instability7, do not adequately reflect

the “investors problem”” (Brunetti and Weder, 1995 pp.125). The article refers to a

survey made that found that entrepreneurs agree that uncertainty in the discretionary

state interventions was the most crucial obstacle to private investment. As it become

important to the investors that rules are created and enforced in a credible and

predictable way, it is also clear why democracy and political stability may not capture

these aspects of quality of the political system. A democracy may have unpredictable

rules as can a state with political stability (Brunetti and Weder, 1995). Hence, what is

viewed by an investor as an important aspect, of whether the risk of investing is too big,

is the aspect of how easy it will be to forecast what the government will do in the future

and how well the investor can rely on the rules already set in place. If the government

constantly makes new rules or decisions on how resources are allocated, it becomes

difficult for the investor to predict the possible future cash flow of an investment. But

whether the rules are changed often or not may not only be reliant upon whether the

country is a democracy or not, or whether the political regime is stable. Therefore, these

two measures are not adequate in capturing all the political risk to an investor.

When comparing this again to Isham et al (1997)’s 3 dimensions of government actions,

it becomes clear that although the how, relating to the underlying political structure and

institutions that authors often measure as democracy and political stability, is important,

7 Political instability refers to regime stability which is the same as government stability.


20

the what actions are taken and How well those decisions are carried out are just as

important to the investor.

For the same reasons as above political violence, the third category of political

variables, although a risk to the investor, does not capture all the risks of predictability

of rules, enforcement of contracts and other risks to the investment. Brunetti (1997)

finds that evidence of the negative effect of political violence on growth is far from

clear. Hence, this variable on its own will not be enough for a measurement of political

risk.

As a summary, it can be concluded that the political risk, that should reflect all risk to

an investor from the government’s actions, is not adequately captured simply by

measures of democracy, government stability or political violence. This means, that in

order to look at the effect of political risk on the use of project finance, one cannot only

us these variables as a measure of political risk. Other aspects of the risks perceived by

an investor are also important and must be included in a measure of political risk.

A fourth option is to measure political risk as the uncertainty of policies. It captures

more of the issues relevant to an investor, as it is a proxy for the amount of uncertainty

that is created by government-controlled policies (Brunetti, 1997). This would be the

predictability and credibility of the rules created as it is measured as the volatility of

monetary and fiscal policies. In this instant, the results in the growth analysis’ seems

more clear towards a negative relationship meaning that higher volatility will lead to a

lower growth although the statistical significance frequently depends on the

specifications of the models when the specifications are richer (Brunetti, 1997).

Feng (2001) measure of policy uncertainty is measured as variability of government

capacity. As Feng (2001) states in his paper “The fluctuation of government capacity

indicates that the government lacks consistency in its power to get a job done.

Uncertainty about government effectiveness can be more adverse than the policy it self

by deterring investors from committing their assets. Given a bad policy with certainty

about its execution the investor can still find ways to make money” (Feng, 2001 pp.276)

It is also interesting in relation to the How well dimension by Isham et al (1997) as it

take into account how well the rules are executed and hence the efficacy of government.

When testing this variable’s relation to private investment, he finds an adversely effect.


21

This means that if a country is experiencing volatility in government capacity to make

decisions and hence in their efficiency in allocating resources or in the monetary and

fiscal policies made, a country will most likely experience lower investment and lower

growth, as investors will postpone their investments until it is more certain what will

happen. Thus, policy uncertainty becomes a political risk that foreign investors (as well

as domestic investors) will assess before making a commitment of their investment.

However, there is still inconsistency in the way that this aspect of political risk is

measured making it hard to judge which to use in a study of political risk. It does also

not capture other aspects of political risk that may be important, like the cause of

inefficiency and the ability of the legal systems.

The latest part of literature on cross-country growth analysis’ that Brunetti (1997) have

summarised, being category five, have focused on using more subjective perceptions of

politics. These are subjective evaluations of private agents divided into two categories;

primary data from surveys among entrepreneurs or commercially provided business

indicators like ICRG8, BERI9 and Business International10. Such measures try to capture

more aspects in one measure by the use of sub components. An example would be

ICRG political risk index that has 12 components of political risk covering areas such as

Internal Conflict in the country (an aspect of political violence), Government Stability,

Bureaucratic Quality (incl. aspects of policy uncertainty) and Investment profile

(including aspects of contract viability and risk of expropriation)11.

This literature is rather limited according to Brunetti (1997), but does show some

indication of a positive relationship between the country experts’ perception and growth

(or investment in one instance). These variables appear to be better at capturing more

accurately the properties of a political system that are significant for private sector-

driven growth.

Stephen Knack and Philip Keefer (1995) also find in their paper “Institutions and

economic performance: Cross-country tests using alternative institutional measures”

that when using subjective measures like ICRG and BERI, that are arguable more

relevant to the incentive for innovation, investment and growth, they find that property

8 International Country Risk Guide provided by Political Risk Service 9 Business Environment Risk Intelligence 10 Business International corporations was in 1986 acquired by The Economic Group and eventually merged with Economist Intelligence Unit. 11 The measure of ICRG will be explained in detail in the next section.


22

rights have a higher impact on investment and growth. This indicates that the Gastil

index covering frequencies of revolutions, coups and political assassinations is less

relevant to growth and investment, than a variable based on 5 measures of property

rights, contractual rights and efficiency of the government from ICRG12 or BERI13.

Hence, for a study on political risk, it seems that the more subjective measures provided

by private agents are better at capturing what is really important to investors and growth

as they cover more areas of all the risks that the investor faces.

Literature on political variables and foreign direct investment (FDI) is briefly

summaries by Harms (2002) in his Political risk and equity investment in developing

countries where he states which authors have tested what aspects. Many of the different

authors have tested the effect on FDI as a percentage of GDP of different measures of

political risk – either constructed by them self or obtained from one of the many

services giving ratings to countries (e.g. Institutional Investor, ICRG, BERI and

Business International). Slightly different results have been found from their studies but

mainly that such measures have a significant effect on the level of FDI. Harms (2002)

stats that it has been argued that one of the reasons for the inconclusive result may be

from the fact that the data sets analysed contains mainly high and middle income

countries and thus neglect the low income countries where political risk is most

prevalent. Thus, his paper includes a larger number of low-income countries and he

finds that political risk is an important determinant of the sum of foreign direct

investment to a country.

The measures he uses for political risk is the risk of “repudiation of contracts by

governments”, the risk of “losses from exchange controls” and then risk of an

“expropriation of private investment” all published by the ICRG14. These measures

more accurately capture aspects important to an investor, as they are aspects of

predictability and credibility of rules.

Again, the subjective measures have been shown to be good measures as they have a

significant effect on the level of FDI, and should thus also be good measures to use in

12 The five measures used from the ICRG political risk indicators are: Expropriation risk, Rule of law, Repudiation of contracts by governments, Corruption in government and Bureaucratic quality 13 The measures used from BERI is: Contract enforceability, Infrastructure quality, Nationalization potential and Bureaucratic Delays 14 These measures are all components of the financial risk component by ICRG but do capture some of the notions that we usually relate to political risk.


23

the study in this paper on the effect of political risk on the use of project finance, which

is also a form of FDI.

Therefore, in order to measure political risk and test its effect on the use of project

finance, previous literature states that one should use a measure of political risk that

captures all aspects of government actions, events and country conditions that are

relevant to an investor. It seems that the measures best able to do this are the more

subjective measures as commercially provided business indicators like BERI and ICRG.

5.3 The model of double moral hazard After having looked at both project finance and political risk, the model of double moral

hazard can be introduced, in order to use it to test the relationship between the level of

political risk in a country and the use of project finance. Before that, it must however be

established what moral hazard is.

5.3.1 Moral hazard defined “Moral Hazard can be defined as actions of economic agents in maximizing their utility

to the detriment of other, in situations where they do not bear the full consequence or

equivalently do not enjoy the full benefits of their actions due to uncertainty and

incomplete or restricted contracts which prevents the assignment of full damage

(benefits) to the agent responsible” (Eatwell et.al. 1987-94 p.549). That means moral

hazard is a term used to describe how the actions of an actor may become more reckless

when he is allowed to escape the consequence of these risky actions. It was first

described by George A. Akerlof in his paper “The market for “Lemons”: Quality,

Uncertainty and the market mechanism” from 1970 in which he describes how

uncertainty and asymmetric information leads to uncertain quality and thus the market

for lemons arises.

As stated, the moral hazard problem arises due to the uncertainty, which is often

expressed in terms of information asymmetry. The two parties participating in the

agreement may not have the same information and they are usually not able to monitor

or measure each other’s effort. When effort cannot be observed, incentives can be given

in order to insure that the type of effort desired is exercised. In this way the market for

lemons can be partly avoided by decreasing the reckless behaviour of the participants.


24

Double moral hazard is a term used to describe how both the parties to the agreement

have little information as to the other party’s effort exerted. “Not only is the effort level

of the agent not fully known to the principal due to informational asymmetries, but

additionally the action of the principal is also not fully transparent to the agent”.

(Murshed & Sen 1995 pp.500)

5.3.2 The model of double moral hazard 15 Hainz and Kleimeier (2006) have found in their paper “Project Finance as a Risk-

Management Tool in International Syndicated Lending” that political risk in the country

has an effect on the use of project finance when using their model of double moral

hazard. This model describes the trade off between the incentives to exert effort by the

two parties involved – the bank and the firm - by granting the project full-recourse or

limited/non-recourse debt.

Through proposition 1 and 2, Hainz and Kleimeier (2006) explain how moral hazard of

the bank can be solved by granting a non-recourse loan, and the moral hazard problem

of the firm can be solved through granting a full-recourse loan

The moral hazard on the part of the firm arises if the firm is not fully accountable for its

actions and therefore tends to involve itself in more reckless behaviour. Not being fully

accountable of its actions happens when the project is financed through non-recourse

debt. In such cases, the bank has no recourse to the sponsoring firm of the project and

should the project hence fail, it will not hurt the sponsoring firm as the project has been

isolated from the firm, as a legally separate entity. It thus allows the manager16 of the

project to take more risky chances as those actions will only affect the firm if they are

positive – leaving him with only upside chance and no downside risk. This may

jeopardize the cash flows of the project; the only source of repayment the banks/lenders

have. Thus, in order to ensure that the firm exerts and effort to help the project succeed,

a full-recourse loan should be granted. In that case, the firm will be accountable for its

risky actions and will thus abstain from taking too risky actions, limiting the downside

risk but also the upside chance for the firm.

15 The model described in this section is based on the descriptions of it given by Hainz and Kleimeier (2006) in their paper and all reference is to their paper if nothing else is stated. 16 Most often the manager of the project is hired by the sponsoring firm or assigned by the sponsoring firm and thus represents them in the firm moral hazard aspects.


25

With regards to the banks, one cannot really talk about reckless behaviour, as they are

usually not in a position in which they can take risky actions that can jeopardize the

project. They are however, capable of exercising some actions that can instead help the

project succeed. As mentioned under advantages of project finance, lenders of non-

recourse loans are more likely to take actions to help the project succeed. This can be

done by exercising their power over the host government, in order to ensure that they do

not interfere with the project in a way that may change the probability of success and

hence the cash flows of the project. The bank may not even have to take a direct action

for this to happen; due to their mere participation in the loan the host government might

refrain from intervening out of fear of repercussion or effects on present or future loans

from that bank. Therefore, by granting non-recourse loans, the bank will take more

actions to ensure the success of the project, as they cannot gain the repayment from

anything else than the cash flows generated by the project.

Hence, the two alternative loan types carry two different moral hazard and incentives. If

both incentive problems are severe, the double moral hazard problem cannot be solved,

as we need both full-recourse debt and non-recourse debt, in order to give both parties

the incentive to act in the best possible way to help the project succeed. The model to

describe the sever state, where both incentives cannot be solved, is stated as follows:

1 1

( ) ( )H H H L

X e bp p p p

≤ +− −

(1)17

where X is the payoff if the project succeed, ( )p p− is the increase in the probability of

success if the firm exerts effort, ( )H Lp p− is the increase in the probability of success if

the bank exerts effort, b is the influence of the bank, and e is the cost of the firm of

exerting effort. As this is not possible to solve for both incentives, incentive must be

given to the party who can solve the moral hazard problem most efficiently.

Thus project finance loans, as a financing method should be preferred when:

17 This is given in proposition 4 in Hainz and Kleimeier’s (2006) paper. Proposition 3 relates to the less server case that can be solved by given limited-recourse loans. However the analysis of Hainz and Kleimeier’s (2006) paper is restricted to the server case as they perceive the incentive problem in reality to be server. Thus this analysis will also take its departure from the server case


26

( )1

( )H L

L L

p p X b

p p X e

− − >− −

(2)

as this indicates that the effect of the banks effort minus it’s cost is greater than the

effect of the firms effort minus it cost and we thus solve for the moral hazard of the

bank. This means that the fraction of project finance loans granted is dependent upon

four factors; namely the increase in probability of success from the firm exerting effort:

( )L Lp p− , the cost to the firm of exerting effort: e, the increase in probability of success

from the bank exerting effort: ( )H Lp p− and the cost to the bank of exerting effort, b.

Political risk is usually connected with government interference and the government’s

actions have a high influence on the probability of success of the project as described in

the section on risk of project failure. If the bank can affect the actions of the

government, in such a way, that the probability of success of the project increases, then

( )H Lp p− can be interpreted as political risk. If political risk is high, then the bank

exerting effort and keeping government from interfering will lead to a high increase in

the probability of success of the project, which means the numerator will increase in

equation (2). The banks effort cost is dependent upon its influence on the host

government. If the influence is high, the cost of restraining the government from taking

actions that will hurt the project is less, which again increases the numerator of equation

(2).

The firm’s moral hazard also has to components. The effect of effort of the firm is

captured by ( )L Lp p− X, which will increase in countries with better economic health.

Hence, the better economic health of the country will make the increase in probability

of success greater. This will in turn decrease the fraction in equation (2), which means

that less project finance loans are given to countries with a better economic health, as

more of the incentive problems are solved for the firms.

The cost of the firm’s effort can also be interpreted as private benefits to the manager.

In countries with less developed corporate governance systems, this cost will be higher

as the manager has higher private benefits that must be given up if not keeping to the

best corporate strategy. If e increases, the fraction in equation (2) will increase and it

will be better to grant a project finance loan and thus solved for the bank’s moral


27

hazard. This also makes sense in relation to what was found in section one about project

finance. The many contracts written as part of a project finance loan restricts the

managements’ discretion, which is needed in countries with less developed laws and

legal systems to handle/encourage good corporate governance.

Thus, it is expected from the model that countries with high political risk, more bank

influence, bad corporate governance and less healthy economy will have more project

finance loans, as this will lead to solving the double moral hazard for the bank rather

than for the firm by granting project finance loans.

5.3.3 Measurement of political risk in Hainz and Kl eimeier (2006)

In order to properly test the model previously described, it is important to use measures

of the independent and dependent variables that does indeed measure what they should

measure. It was previously discussed how political risk is most appropriately measured

and it was found that the fifth category – the subjective perception of politics – was best

at capturing the risk to an investor. Hainz and Kleimeier (2006) does indeed use such a

subjective measure of political risk, namely the World Bank Governance Indicators

(WGI). However there are two main problems to address. First WGI may not be the best

provided political risk measure and the aggregation of all aspects of the political risk

into one variable seems inappropriate according to the review of how different aspects

of political risk may affect a project as discussed earlier.

The WGI have been criticised by Kazi Iqbal and Anwar Shah (2008) in their paper

“Truth in Advertisement: How do Worldwide Governance Indicators Stack Up?”. They

state that: “WGIs indeed characterize what Klitgaard et al (2005) call “an explosion of

measures, with little progress toward theoretical clarity or practical utility”( p.414) and

we agree with Thomas (2006) that “ reliance upon them for any purpose is premature”

(p.1)” (Iqbal & Shah 2008 p.48). They state two main problems of the indicators; lack

of a conceptual framework when the primary indicators are aggregated and

measurement problems due to problems with the primary indicators and the distribution

of weights. The aggregation of many primary indicators makes the measurements less

clear on what is measured and at the same time each question used in the indicators are

given equal weight regardless of the importance of that question to the concept of


28

governance(Iqbal & Shah 2008). This makes the WGI conceptually flawed and perhaps

even useless for use in further studies.

The other problem is also severe for the use in testing the model of double moral

hazard. The aggregation of sources that might be correlated due to use of same experts

or reliance on the same external events will lead to correlated errors across countries

and with in countries. At the same time different weights of sources over time and over

countries makes cross country and time comparison misleading (Iqbar & Shah 2008).

Such comparison is highly important for the model of double moral hazard measuring

the use of project finance loans across time and countries.

The second problem in the paper by Hainz and Kleimeier (2006) is the aggregation of

all notions of political risk into on variable. As discussed in the section on project

finance there are several aspects of political risk that can affect the project and if testing

the influence of political risk one cannot aggregate all the aspects, as they might be

different in the effect. Hainz and Kleimeier (2006) do acknowledge that there are

different types of political risk. Based on Smith (1997)18, they have divided the political

risk into three categories, which are very sound based on what was found under project

finance and the events that can make a project fail. These are Traditional political risks:

Expropriation, currency convertibility, transferability, political violence (war, sabotage,

terrorism), Regulatory risk: Risk from unanticipated regulatory changes (taxation,

foreign inv. laws (output price)), and Quasi-commercial risks: Risks when having state-

owned supplier or customer with questionable willingness to fulfilled obligations.

However they do not relate this to the six governance measures from the World Bank

and they aggregate all the six measures into one variable by use of equal weights. They

do test each of these six measures in simple linear regressions with project finance, as

the dependent variables, but without other relevant independent variables that conclude

that each is important for the use of project finance.

All in all the main problem of Hainz & Kleimeier (2006) in measurement of political

risk is the use of WGI measures, which are not well defined and misleading when

18 Smith, W., 1997 Covering political and regulatory risks: Issues and options for private infrastructure arrangements. In: Irwin, T (Ed.), Dealing with Public Risk in Private Infrastructure, The international Bank for Reconstruction and Development, 2nd Edition, Washington, pp. 45-88


29

compared across time and country, and at the same time the political risk variables is an

aggregation of different measures that allows the reader little information on the real

aspects of political risk that leads to the use of project finance.

The next section will thus discuss, which measure of political risk to use in the paper

when testing the model of double moral hazard regarding the use of project finance.

5.3.4 Measuring political risk in the model of doub le moral hazard As established in the section 5.2.2, the political risk measure used in this paper will be a

subjective measure namely the political risk index provided by ICRG. Compared to

WGI they have the advantage of not being an aggregation of other primary sources.

ICRG’s risk index is also an aggregation of different components but these are assign

points on a more consistent basis. Political information is collected along with financial

and economic data, which are all converted into risk points for each individual

component based on a pattern of evaluation that should be consistent. This should also

be ensured, both between countries and over time, by having the points assigned on a

series of pre-set questions for each risk component (PRS group). According to Iqbal and

Shah (2008) ICRG is however still biased in their interpretation of causes and effects.

They believe that if the same government has been in place for a long time this will lead

to more corruption. However the defaults of ICRG are far out weighted by those of

WGI and ICRG is thus deemed the best for use in this study.

ICRG political risk index consists of 12 components that together constitute political

risk in a country. The 12 components are described in Political Risk Service group’s19

methodology on International Country Risk Guide and are summarised below in table

5.1:

19 PRS group


30

Table 5.1: Variables comprising the ICRG’s political risk index20

Variable Points Assessment of:Government

stability 0-12 A government’s ability to carry out its declared program(s)

and its ability to stay in officeThe socioeconomic pressure at work in a society that couldconstrain government action or fuel social dissatisfaction. Sub-components are: Poverty, Unemployment and ConsumerConfidence.Factors affecting the risk to investment not covered by otherpolitical, economical and financial risk componentsSub components: contract viability/expropriation, profitsrepatriation, payment delays

Internal conflict 0-12 Political violence in the country and its actual or potentialimpact on governance

External conflict

0-12 Risk to the incumbent government from foreign actions,ranging from non-violence external pressure to violentexternal pressureCorruption within the political systemFinancial corruption and corruption in the from of excessivepatronage, nepotism, job reservation, favour-for favoursandsuspiciously close ties between politics and business

Military in politics

Religious tensions

0-6 Domination of society and/or government by a single religiousgroup that seeks to replace civil law by religious law and toexclude other religions from the political and/or socialprocess.

Law and Law: The strength an impartiality of the legal systemorder Order: popular observance of the law (people following the

law)Ethnic

tensionsHow responsive government is to its people on the basis thatthe less responsive it is the more likely is it that thegovernment will fall (peacefully or violently)Ranging from Alternating democracies to Autarchy The strength and expertise to govern without drastic changesin policy or interruptions in government services. Ability to absorb shocks to minimise revision of policy whengovernments change.

Bureaucracy quality

0-4

0-6

0-6 Degree of tension within a country attributable to racial,nationality or language divisions

Democratic accountability

0-6

Corruption 0-6

0-6 Military involvement in politics even at a peripheral level is adiminution of democratic accountability

Socioeconomic conditions

0-12

Investment profile

0-12

Source: Political Risk Service group – ICRG Methodology

These variables may not directly measure what actions/decisions/rules the government

may take, as described by Isham et al (1997), but as Howell and Chaddick (1994) states

in their paper “Models of political risk for foreign investment and trade”: “In the

development of forecasting techniques, the task is to link theoretically the act resulting

in loss (such as civil strife damage) to the causes of the act (such as an ethnic dispute

20 For a full statement of the different risks please refer to appendix 3


31

dissolving into open conflict) or predictors of the cause or the event (such as the

existence of ethnic tension)… A modeller would presumably start with a list of acts,

having political or social manifestations that result in losses to an investor and look for

predictors that can be observed in society” (Howell & Chaddick 1994 pp.73). Hence the

variables created by ICRG will also be looking at predictors of the cause or the event

leading to a loss for the investor. E.g. the variable External Conflict suggests that

foreign actors may act against the incumbent government based upon its actions. These

acts may be in the form of a trade sanction and may lead to lack of supply from other

countries to the project, as discussed in the section on “Risks of project failure”.

These variables cover all the 4 categories that literature has used as measures of political

variables before using the subjective measures like the ICRG political risk index. E.g.

Democracy is covered by Democratic accountability, Government stability is measured

by the variables of the same name, political violence is covered by Internal conflict and

policy uncertainty is covered by Bureaucratic quality while the rest of the variables

either help explain one of these notions or other more specific areas of risk that a

government could pose to an investor like the legal systems ability to protect property

rights.

What is much more important to look into, is whether these variables do in fact cover

the aspects of risk related to a possible project failure, as discussed earlier. The main

focus is on the actions of a government and the events and conditions of a country that

may lead to a project that is not able to generate the projected cash flow and hence not

able to pay back the debt. As the debt is given as non-recourse, the bank is thus the

investor to whom the political risk aspect is of specific interest in regards to getting their

money back. The sponsoring company is naturally also concerned with political risk, as

it may jeopardize the project and although it may not cost them more than they have

invested from the beginning, amounts that can still be substantial, it will be a risk they

will have to diversify and take into account when making the project, in order to get the

debt finance they need for it.

Yescombe’s (2002) 3 variables constituting risk to a project are all related to some

extent to the ICRG political risk index although Yescombe (2002) only classifies one of

the areas as political risk. It was however seen under the section “risk of project failure”


32

that some of the political risks constituting change of legislation could affect the

commercial risks of input and revenue.

The commercial risks, which relates to a project’s construction and operational phase,

are covered by variables such as Bureaucratic quality in relation to obtaining permits

and keeping them even though the government might change. The same risk is also

related to the level of corruption in government, as it will be an indication of whether

the government or sub-levels or the government might grant favourable terms to local

competitors to the project which may hurt the operation of the project. The input risk is

covered by external conflicts, as discussed earlier, which also covers war and conflict

risks found under Yescombe’s (2002) investment risk component. The investment risk

consisting of war and conflict along with currency convertibility and risk of

expropriation, is thus covered by many of the ICRG political risk index variables,

namely all those relating to conflicts and tensions while the Investment profile covers

the last two risks. The “change of law” component from Yescombe’s (2002) risk matrix,

is mainly covered by Bureaucratic quality, as it deals with policy uncertainty and the

ability of the government to make predictable policies that does not change when

government change. This could also be related to the government stability, as it covers

the government’s ability to carry out its declared program(s). This would suggest

whether the policies made, will be somewhat predictable, as they have declared in the

program and hence should not suddenly pose new laws that could alter the profitability

of the project. The last component of Yescombe’s (2002) political risk, is the quasi-

political risk which is taken into account through the two variables “Rule of Law” that

looks at the effectiveness of the legal system, and “Investment Profile”. The former is

highly important in project finance due to the many contracts made and the latter

includes viability of contracts and thus the likelihood of government breach of contract

and also the applicability of the contracts.

The other variables of ICRG political risk index that are not directly mentioned may,

however, also be relevant to the level of project finance given. The socio economic

conditions in a country affect the government’s actions by fuelling dissatisfaction or

social unrest based on poverty, unemployment or consumer dissatisfaction. Such unrest

could potentially lead to changes in policies, which may harm the project; e.g. the

government poses a new environmental law to increase consumer confidence, a new

labour law to lower unemployment or it may demand the project sponsors to build new


33

schools or roads to help the development of the country in exchange for giving permits.

The other end of the scale may be a civil unrest that eventually leads to more violent

measures that may lead to a total discontinuation of the project. Or it may be that an

unrest focusing negatively on the foreign operations, may lead to expropriation of such

projects/investments.

Military in politics and democratic accountability are related as military in politics will

lead to a decrease in democratic accountability. The latter is also important to the

project, as it relates to how responsive the government is to its citizens and thus how

likely it is that the incumbent government may stay in office. If a government is less

responsive to its citizens, it is also less accountable to the citizens, which to an investor

means that the government may not do what is best for the citizens and hence

development of the country, but may be more interested in diverting wealth towards

themselves or their allies. It is necessary for an investor in order to trust that the

government will support growth and investment and hence the project, and not engage

in expropriation or creeping expropriation.

It can thus be seen that all the variables in the ICRG political risk index, may be of

importance to an investor, and hence to the level of project finance given to a country.

When looking at these 12 variables, one will notice that they do overlap several places

and that there must thus be high correlation among the variables. It is thus suggested

that a principal component analysis is done on the 12 variables in order to create a fewer

amount of components for use in the model of double moral hazard. This will be done

in the following section.


34

6. Political risk analysis As has been discussed earlier in this paper, the way Hainz & Kleimeier (2006) measure

and combine the political risk measure in their paper “political risk in syndicated

lending: theory and empirical evidence regarding the use of project finance” makes

interpretation of the result difficult, as it cannot be said which dimension of political

risk it is that affects the use of project finance. This is problematic, as it was seen in the

theoretical section that there are many ways in which different aspects of political risk

can affect a project.

Hence, this part of the paper is about finding the right components to measure political

risk in the model of double moral hazard. It was found that the ICRG measures of

political risk appropriately captured the investor risk, which should be incorporated in

this measure. However, the 12 components of the ICRG’s political risk index, show a

high level of correlation and it is thus necessary to combine them into a smaller number

of components, remembering that these should still make sense theoretically.

6.1 Methodology As mentioned, this paper will make use of a principal component analysis. This will be

used to define the underlying structure among the ICRG political risk variables, by

analysing the correlation among the variables and then combine the variables that are

highly interrelated. This will be done without much loss of information by accounting

for most of the variance found in the ICRG political risk variables (Hair et al. 2005). In

other words, it is the pattern of correlations within a set of observed variables that factor

analysis will try to establish and then combine those variables mostly correlated into

components.

This paper will follow the approach set out by Hair et al (2005). in the book:

“Multivariate data analysis” chapter 3 on factor analysis. This means that the sequence

will be to firstly look at the assumptions, then the extraction method follow by rotation

of the solution, to find the best solution. In the end validity and further use will be

discussed. However, before all of this the data used in the analysis will be discussed.

The analysis is done in SPSS, as the data set is not too large for SPSS to handle and the

output given in this program are extensive.


35

6.2 Data The data used for this analysis is the International Country Risk Guide’s 12 variables

for political risk. These cover both political and social aspects of the countries that may

affect international business operations. These assessments are made upon a subjective

analysis by what is referred to as country experts and are based on the data available.

The 12 components were described in the previous section.

The variables are given points that are ranged from 0-12, 0-6 or 0-4 in the ICRG

political risk index, which means that they are weighted differently once summarized in

a total political risk measure. However, for the use in the analysis, all the variables have

been converted into 0-12 point scales making them approximately interval scaled,

which is best in order to calculate correlations among the variables. This does mean that

the weighting used by ICRG political risk index is removed and all have equal weight.

The data covers 139 countries21 from 1984 to 2005, which means that there are 3058

observations when using yearly data and 556 when using 5-year averages. However,

several countries are missing data for some years. So when running the PCA cases are

excluded list-wise when data is missing and we are thus down to 2450 and 484

observations, respectively22. According to Hair et al (2005), there should be no problem

as long as there is more than 5-10 observation per variable, which is by far reached for

both analyses’. Both analyses have been done in order to validate the result. The 5-year

average has the advantage of taking out the noise that may be present due to one year

deviations, which is why this analysis will be presented in this paper, and the other

analysis on the yearly data will be found under validation of the result.

6.3 Empirical evidence23 Before looking at the final result of the analysis, it must be established if the

assumptions of the analysis have been fulfilled.

6.3.1 Assumptions The assumptions are divided into 2 areas, being conceptual issues and statistical issues.

The first one relates to the relevance of the analysis and states that a conceptual

appropriateness must be underlying the analysis. This is fulfilled in this analysis, as the

21 See appendix 4 for a list of all the countries included in the analysis. 22 See appendix 5 for the descriptive of each sample – incl. the number of valid observations for each analysis 23 This section is referenced to Hair et al. 2005 unless otherwise stated.


36

12 components all measure some aspects of political risk. There should thus be some

underlying structure of these components, as they measure some of the same aspects;

e.g. conditions in the country and ways in which the government works.

The statistical assumptions can be tested through several outputs in the analysis. The

main assumption to fulfil, in order to use factor analysis, is that there is enough

correlation between the variables to make representative factors. Hence, the data matrix

should have sufficient inter-correlation to justify the use of factor analysis. There are 4

ways of judging the overall inter-correlation between variables justifying a factor

analysis: Correlations between variables, Partical correlations, Bartlett test of sphericity

and Measure of Sampling Adequacy. Each of these will be evaluated in the following

sub-sections.

Correlations24

Very few of the correlations in table 6.1, are below 0.3 and this should thus justify the

use of a factor analysis in relation to the ICRG political risk variables, as they correlate

enough to assume that there is an underlying structure of the variables so that they can

be combined into a smaller number of components.

Table 6.1 Correlation matrix

Note: The top part shows the correlations and the bottom part the significance of the correlation

Bartlett test of Sphericity and Kaiser-Meyer-Olkin Measure of sampling adequacy

The Bartlett test examines the entire correlation matrix and performs a statistical test for

the amount of correlation. The hypothesis and the alternative are stated below:

24 Please refer to Appendix 6 for this table

Bureaucratic quality 1 0,659 0,612 0,293 0,348 0,220 0,514 0,456 0,650 0,678 0,282 0,660

Corruption 0,659 1 0,549 0,319 0,298 -0,008 0,438 0,150 0,585 0,549 0,370 0,490

Democratic accountability 0,612 0,549 1 0,257 0,416 0,157 0,427 0,389 0,466 0,575 0,352 0,310

Ethnic tension 0,293 0,319 0,257 1 0,387 0,317 0,606 0,245 0,526 0,412 0,372 0,292

External conflicts 0,348 0,298 0,416 0,387 1 0,326 0,625 0,348 0,428 0,468 0,395 0,261

Government stability 0,220 -0,008 0,157 0,317 0,326 1 0,445 0,650 0,408 0,258 0,117 0,229

Internal conflict 0,514 0,438 0,427 0,606 0,625 0,445 1 0,457 0,736 0,668 0,448 0,463

Investment profil 0,456 0,150 0,389 0,245 0,348 0,650 0,457 1 0,426 0,471 0,207 0,563

Law & order 0,650 0,585 0,466 0,526 0,428 0,408 0,736 0,426 1 0,645 0,348 0,550

Military in politics 0,678 0,549 0,575 0,412 0,468 0,258 0,668 0,471 0,645 1 0,380 0,529

Religious tensions 0,282 0,370 0,352 0,372 0,395 0,117 0,448 0,207 0,348 0,380 1 0,196

Socioeconomic conditions 0,660 0,490 0,310 0,292 0,261 0,229 0,463 0,563 0,550 0,529 0,196 1

Bureaucratic quality 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000

Corruption 0,000 0,000 0,000 0,000 0,434 0,000 0,000 0,000 0,000 0,000 0,000

Democratic accountability 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000

Ethnic tension 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000

External conflicts 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000

Government stability 0,000 0,434 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,005 0,000

Internal conflict 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000

Investment profil 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000

Law & order 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000

Military in politics 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000

Religious tensions 0,000 0,000 0,000 0,000 0,000 0,005 0,000 0,000 0,000 0,000 0,000

Socioeconomic conditions 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000

Religious tensions

Socioeconomic conditions

Correlation Matrix

Internal conflict

Investment profil

Law & order

Military in politics

Democratic accountability

Ethnic tension

External conflicts

Government stability

Correlation

Sig. (1-tailed)

Bureaucratic quality Corruption


37

H0: No significant correlation among variables

H1: Significant correlation among at least some variables

The result can be found in table 6.2. As the p-value is less than 0.05 the null-hypothesis

is rejected and there are strong indications of enough correlation in the data matrix to

justify factor analysis. There is one main problem with the Bartlett test that should be

noted. When the sample size increases, even small correlations tend to become

statistical significant, which leads to a lower appropriateness of this test. That is why

this paper cannot solely rely on this test. However, the other tests mentioned so far,

have given the same result as to justifying the use of factor analysis.

Table 6.2: Evaluation of correlation

KMO and Bartlett's Test

,848

3414,226

66

,000

Kaiser-Meyer-Olkin Measure of SamplingAdequacy.

Approx. Chi-Square

df

Sig.

Bartlett's Test ofSphericity

The Kaiser-Meyer-Olkin Measure of sampling adequacy (KMO - MSA), often referred

to as the over all MSA, also measures the inter-correlation among the variables. The

measure ranges from 0-1 where 1 is obtained when each variable can be perfectly

predicted by all the other variables without error, thus the higher the number the better.

It should at least be 0.5. In table 6.1 it can be seen that it has a value of 0.848, which

according to Hair et al (2005), is meritorious for a factor analysis to be performed.

Partial correlation

Partial correlation is the part of the correlation that is left unexplained when all the other

variables have been taken into account. That means that if it is high, it is an indication

of little correlation among the variables and thus no underlying structure of the variables

justifying a factor analysis. The partial correlation should be less than 0.7 to justify

factor analysis and they can be found in the Anti-image correlation matrix given in table

6.3. It should be noted that the values given are the negative value of the partial

correlation.


38

Table 6.3: Partial correlation and individual MSA values

Note: the off-diagonal values are the partial correlations and the diagonals (indicated by an a) shows each variables individual measure of sampling adequacy As can be seen in table 6.3, all of the off-diagonal values, representing the negative

partial correlations, are below 0.7 (in absolute values) and hence this indicates that a

factor analysis is appropriate for the given data as found in the previous tests.

Variable specific measures of inter-correlation

In table 6.3, is also given the measure of sampling adequacy (MSA) for each individual

variable. It is the same measure as before but is now given for each individual variable

in order to judge the individual variables appropriateness in the factor analysis.

As can be seen, in the table all individual MSA are above the required 0.5, the lowest

being 0.668 and the highest being 0.936, indicating that each variables is correlated

enough with the others to be part of the factor analysis.

It has thus been established that all 12 variables of the ICRG political risk index can be

included in the factor analysis and there is enough correlation among the variables to

justify the use of factor analysis. The next step is to find the number of components to

extract from the analysis.

6.3.2 Results What is needed is a component solution, which account for as much variance as

possible. Thus each component will be the best linear combination of variables, based

on it accounting for as much of the variance that is still unexplained, as possible. Thus

Anti-image Correlation

Bureaucratic quality 0,885 a -0,245 -0,301 0,104 0,018 -0,037 0,050 -0,011 -0,187 -0,232 0,057 -0,351

Corruption -0,245 0,862 a -0,264 -0,032 0,001 0,101 0,041 0,260 -0,233 -0,061 -0,173 -0,220Democratic

accountability -0,301 -0,264 0,824 a -0,007 -0,170 0,146 0,051 -0,296 -0,016 -0,148 -0,066 0,328

Ethnic tension 0,104 -0,032 -0,007 0,912 a -0,006 -0,119 -0,271 0,104 -0,136 -0,022 -0,146 -0,066

External conflicts 0,018 0,001 -0,170 -0,006 0,895 a -0,081 -0,377 -0,005 0,105 -0,038 -0,129 0,024Government

stability -0,037 0,101 0,146 -0,119 -0,081 0,668 a -0,121 -0,606 -0,243 0,138 0,096 0,268

Internal conflict 0,050 0,041 0,051 -0,271 -0,377 -0,121 0,871 a 0,004 -0,375 -0,277 -0,138 -0,055

Investment profil -0,011 0,260 -0,296 0,104 -0,005 -0,606 0,004 0,702 a 0,098 -0,155 -0,090 -0,505

Law & order -0,187 -0,233 -0,016 -0,136 0,105 -0,243 -0,375 0,098 0,901 a -0,083 0,035 -0,110

Military in politics -0,232 -0,061 -0,148 -0,022 -0,038 0,138 -0,277 -0,155 -0,083 0,936 a -0,033 -0,008

Religious tensions 0,057 -0,173 -0,066 -0,146 -0,129 0,096 -0,138 -0,090 0,035 -0,033 0,908 a 0,085Socioeconomic

conditions -0,351 -0,220 0,328 -0,066 0,024 0,268 -0,055 -0,505 -0,110 -0,008 0,085 0,771 a

Law

& o

rder

Milit

ary

in

polit

ics

Rel

igio

us

tens

ions

Soci

oeco

nom

ic

cond

ition

s

Exte

rnal

co

nflic

ts

Gov

ernm

ent

stab

ility

Inte

rnal

con

flict

Inve

stm

ent

prof

il

Bure

aucr

atic

qu

ality

Cor

rupt

ion

Dem

ocra

tic

acco

unta

bilit

y

Ethn

ic te

nsio

n


39

SPSS will originally present as many components as there are variables in the first place

where the first principal component accounts for as much of the variability in the data as

possible, and each succeeding component accounts for as much of the remaining

variability as possible.

To obtain the numbers of components that is underlying the data, there are 3 extraction

methods to consider.

The first is Eigen-value method, which states that if the Eigen-value is above one for the

component, that component should be included. If Eigen-values are above one it

indicates that the specific component accounts for at least 1/12 % of the total variation,

which means that it has explanatory power and is significant as the component account

for the variance of at least one variable. The Eigen-value is closely linked to the amount

of variance the component accounts for, as it can be calculated as the number variables

included in the analysis multiplied by the percentage of the total variance that the

component account for25.

When extracting the number of components based on this rule, 3 components will be

chosen, as there are three components with Eigen-values above one, as can be seen in

table 6.3

Table 6.4: Extraction of components

Total Variance Explained

5,754 47,952 47,952 5,754 47,952 47,952

1,445 12,039 59,990 1,445 12,039 59,990

1,183 9,859 69,849 1,183 9,859 69,849

,801 6,672 76,521

,618 5,152 81,673

,540 4,504 86,178

,426 3,550 89,727

,390 3,248 92,975

,262 2,182 95,157

,242 2,014 97,171

,181 1,512 98,683

,158 1,317 100,000

Component1

2

3

4

5

6

7

8

9

10

11

12

Total % of Variance Cumulative % Total % of Variance Cumulative %

Initial Eigenvalues Extraction Sums of Squared Loadings

Extraction Method: Principal Component Analysis.

Note: The last three columns show the solution that has been chosen for the analysis – being a 3 component solution. Eigen values above 1 is extracted along with a cumulative variance above 60%

25 For component number 1 it is: 12*0.47952=5.754


40

The second extraction method is also based on table 6.4, as it is based on the percentage

of cumulative variance given by the different component solutions. It is suggests that

one should not settle for less than 60% of the total variance to be included in the

component solution. In this case, the numbers of components to extract would also be

three as the percentage of cumulative variance is 69.849 for that solution.

The last extraction method called a scree test criterion in which a scree plot extracted

from SPSS is evaluated. The scree plot can be used to identify the number of

components to extract before the amount of unique variance start to dominant the

variance structure. The unique variance is the part of the total variance, which is not

shared between the variables. Thus, if the later components are more based upon unique

variance, they will not be included in the component solution, which should illustrate

the correlation and variance shared among the variables.

Note: The extraction is done based on when unique variance starts dominating when the curve straightens out.

As can be seen in figure 6.1, four components should be extracted, as that is the point

on the curve where the curve starts to straighten out, which indicates where unique

variance starts dominating the components. This is one more component than the two

previous methods suggest should be extracted. However, this is not uncommon in factor

analysis that the scree test criterion yields one or two more components than the other

methods.

As two out of three of the methods suggest three components, that is the number of

components, this analysis will extract and use in the following analysis of project

finance and political risk.

Figure 6.1: Scree plot


41

This is also confirmed when looking at the communalities, which measures how well

the individual variables are explained by the three factors, by measuring the amount of

variance in each variable the factor solution account for. These measures should be

above 0.5, which they all are as can be seen in table 6.5. Hence the factor solution with

three components will explain enough of the variation of each of the 12 variables.

Table 6.5: Communalities

Note: Communalities is the amount of variance in each Variable the factor solution accounts for. They should be above 0.5 to be included in the solution The next step is to find out which variables represent which components. This is done

by initially looking at the component matrix showing the loadings for a variable on the

different components, the loadings being the correlations between the variable and the

component.

The un-rotated component matrix does not provide a clear solution, as to which

variables represent which component. However, by rotating the solution one can reduce

the ambiguity that comes with cross loadings and should thus be able to simplify the

factor structure making it more clear which variables represent each component. The

idea behind factor rotation is thus to ensure that the variables have one high loading on

one component, which makes it clear which variables represent each component. There

are two overall rotation methods, namely orthogonal and oblique rotation. The first is,

as the name suggests, a rotation in which the axes are rotated with a 90 degree angle so

that the components are mathematical independent. The other rotation method, oblique,

does not rotate the axes in a right-angle, meaning that the components do not become

uncorrelated. This has the advantage of clustering the variables more accurately but as

Initial Extraction

Bureaucratic quality 1 0,829

Corruption 1 0,778

Democratic accountability 1 0,543

Ethnic tension 1 0,585

External conflicts 1 0,574

Government stability 1 0,817

Internal conflict 1 0,793

Investment profil 1 0,809

Law & order 1 0,702

Military in politics 1 0,701

Religious tensions 1 0,565Socioeconomic conditions 1 0,687


42

the components are still correlated it makes the component solution less useful in

further analysis. In figure 6.2 is given the rotated solutions

Figure 6.2: Rotated component solutions Varimax rotation Oblimin rotation

1 2 3 1 2 3

Bureaucratic quality 0,873 0,157 0,204 Bureaucratic quality 0,919 -0,054 0,059

Corruption 0,805 0,302 -0,195 Corruption 0,822 0,146 -0,350

Democratic accountability 0,661 0,321 0,043 Democratic accountability 0,649 0,188 -0,087

Law & order 0,604 0,482 0,323 Law & order 0,535 0,361 0,192

Military in politics 0,686 0,421 0,232 Military in politics 0,644 0,279 0,093

Socioeconomic conditions 0,727 0,017 0,399 Socioeconomic conditions 0,784 -0,185 0,292

Ethnic tension 0,143 0,725 0,199 Ethnic tension -0,031 0,748 0,115

External conflicts 0,195 0,690 0,244 External conflicts 0,033 0,695 0,156

Religious tensions 0,212 0,713 -0,106 Religious tensions 0,059 0,738 -0,205

Internal conflict 0,390 0,709 0,372 Internal conflict 0,239 0,660 0,255

Investment profil 0,358 0,102 0,819 Investment profil 0,339 -0,027 0,769Government stability -0,030 0,263 0,864 Government stability -0,132 0,242 0,860

Extraction Method: Principal Component Analysis. Extraction Method: Principal Component Analysis.

Rotation Method: Varimax with Kaiser Normalization. Rotation Method: Oblimin with Kaiser Normalization.

Rotation converged in 5 iterations. Rotation converged in 8 iterations.

Component Component

Note: The left table is an orthogonal rotation method and the right one is a oblique rotation solution. The bold numbers are the highest loadings for each variable. The Varimax solution has some cross loading on Law & order and Military in politics.

The first presented is an orthogonal solution called Varimax and the second is an

oblique solution called Oblimin. The most wide spread method is the Varimax solution,

as it seems to give a more clear separation of the components than other orthogonal

solutions. As can be seen in figure 6.2, the Oblimin solution has no cross loading of

significance, whereas the Varimax shows cross-loadings on the “Law and order” and

“Military in politics” variables. However, as mentioned earlier, it is better for further

use if the components are independent and hence the Varimax solution is preferable.

The variables thus represent the three components in the following way based on high

loadings:


43

Table 6.6: Components extracted

Component 1 (QI) Component 2 (TC) Component 3 (IC)“Quality of Institutions” “Conflicts and Tensions” "Policy Quality"

Bureaucratic quality Ethnic Tension Government StabilityCorruption Religious Tension Investment Profile

Democratic accountability External ConflictSocial conditions Internal Conflict

Law & Order Military in politics

These results are consistent with the ones found by Berggren, Bergh and Bjørnskov

(2008) in their paper “The growth effects of institutional quality, instability and

change”.

Each of the components should be named in a way that reflects the variables

representing the component. The first component will be named “Quality of

Institutions”, as this should capture how well the institutions work by being represented

by the quality of the bureaucracy and the legal system, the level of corruption in the

governmental institutions and the accountability of the government to the people. The

social conditions of the country may seem to have less theoretically grounds for being

placed in this component. However, as it is a measure of unemployment and poverty, it

does indeed reflect the governmental institutions’ ability to provide sound economic

conditions in the country.

The next component will be named “Tension & Conflicts”, as it reflects both internal

and external conflict potentially leading to violence or war and ethnic and religious

tensions potentially causing investment hesitation to an investor.

The last component is named “Policy Quality”, as it reflects the quality of the policies

that the government makes. This covers the investment profile of the country, which is

comprised of the expropriation risk and currency convertibility that are related to

policies made by the government in office. The last one measures the government’s

ability to stay in office and carry out its programs and could thus be an indication of

whether it is creditworthy and has the quality to stay in power.

These three components can all be linked to the risks faced by project finances lenders

and sponsors. The quality of institutions is important for governing the many contracts

undertaken and support the project through efficient bureaucracy and democracy. As

mentioned in the theoretical part on project finance the many contracts written makes


44

legal systems and support of the institutions highly important. This component should

capture some of the risks of changing legislation; like import duties, price controls and

permits given, and security of property rights along with efficient treatment of contract

breaches.

Tension & Conflicts may harm any investment through a possible discontinuation,

which to an investment financed through project finance means that the bank will

receive its payments as cash flows will stop. This covers the risk of war and conflicts

mentioned in Yescombes (2002) investment risks.

Policy Quality is also highly important to the investor and lender as it reflects the

policies under which the project must work and the credibility of the government and

their promises. It covers the investment risk related to currency convertibility and

transfer of the profit and the stability of the government, which is not directly

mentioned as a risk by writers of project finance text books. It may however, be an

indication of the policies quality whether the government’s actions are predictable

through legislative strength and popular support of the government.

Validation and further use

The result should be validated beyond the fact that others have found same result. This

has been done as mentioned earlier, by doing the analysis both based on 5-year averages

and on yearly data. The analyses yield the same result26. The data has also been tested

by dividing it into three sub-samples based on the country names alphabetically, so that

first sub-sample includes countries from Angola to Guinea on the list in appendix 4;

sub-sample two includes countries from Gambia to Niger and the rest is in sub-sample

three. These three sub-samples all yield the same division of the variables into the

components27. If the sample is divided by income levels (High, Upper middle, Lower

middle and Low) or the years (1990, 1995, 2000 and 2005), the same result is not

obtained. However, as the regression to be made in section 7, in which these results are

to be used, is an aggregate for all countries in all income groups and over all the years,

the result of the initial component analysis will be deemed satisfactory, as that is also an

aggregate of all countries and all years.

26 See appendix 6 for the results on the yearly data set 27 See appendix 7 on these results. The


45

The last part of the analysis is to determine how to use the results in further analysis.

One can either choose to use the variable with the highest loading as a surrogate or a

new variable can be created based on summated scale or factor scores. As none of the

variables have a loading substantially higher than the others in either component the

surrogate variable option is abandoned. Using a summated scale to create a new variable

is done by summing the variables and taking an average meaning that each variable in

the component will have a weight determined by how many other variable it is grouped

with. However in this paper it has chosen to use the last option being factor scores. This

means that the factor loadings given in the Varimax rotation solution will be used to

calculate the components. Each variable will thus have a different weight according to

how high the loading is. This type of weighting seems more appropriate for use in this

analysis.


46

7. Project finance and political risk analysis

After having determined 3 appropriate measures of political risk the model of double

moral hazard proposed by Hainz and Kleimeier (2006) can now be tested.

This section will begin with a review of the method to be used for the analysis followed

by a review of the data used. Finally the results of the analysis will be presented and

discussed.

7.1 Methodology

This section will first look into the model that is going to be analysed and will be

followed by a section on method for estimation.

7.1.1 The model The model to be tested is the one described in section 5.3.2. The 2 different components

of each moral hazard were discussed and it was argued how they could be interpreted.

This is useful for finding ways to measure each component. Below is given a table that

describes the notation of the equation from 5.3.2 and the interpretation of each part of

the model.

Table 7.1: Variables in the model of double moral hazard

Economic health of country

Political risk

(negative relationship) (positive relationship)

Countries corporate governance system

Bank influence

(Negative relationship) (positive relationship)

Manager influence on the probability of success

Bank influence on the probability of success

Firm effort cost Bank effort cost

Firm moral hazard Interpretation Bank moral hazard Interpretation

Source: Hainz & Kleimeier (2006) Note: The parenthesis indicates the expected relationship between the variable and the use of project finance loans over syndicated loans. Hence, the model to be tested relates the fraction of project finance loans out of total

syndicated loans granted to country i in year t to bank moral hazard and firm moral

hazard in the following way:

0 1 2 3

4

_ _ _

_it it it it

it it

PF Economic health Corporate Governance Political risk

Bank influence

β β β ββ ε

= + + ++ +

(1)


47

As discussed earlier, political risk has several aspects that should be included separately

thus the three components (Quality of Institutions, Tension & Conflicts, Policy Quality)

found in the political risk analysis in section 6 are used separately in the model.

Economic health of the country is measured by using the GDP pr capita28 and corporate

governance will be measured as a combination of stock market capitalizations as a

percentage of GDP and domestic credit to private sector as a percentage of GDP as done

by Hainz and Kleimeier (2006)29. In order to take into account the world wide business

cycles, 4 dummy variables are constructed for the time period: 1987-1990, 1991-1995,

1996-2000 and 2001-200530. Failing to take account of the common effects of particular

years or periods across, all cross-sectional units, will lead to contemporaneous

correlation that will affect the standard errors of the coefficients. Thus, also yearly

dummies are constructed and tested in the model, to see if they give a different result

than the period dummies. At the same time, such dummies will take account of the

change in the sample across years, as more countries are represented in the later years

compared to the first years.

The model to be tested then looks like this:

0 1 2 3

4 5 6

7

_ _

_ _ _

_

it it it it

it it it

it it

PF Corporate Governance GDP Quality institutions

Tension Conflict Policy Quality Bank influence

Time Dummies

β β β ββ β ββ ε+

= + + ++ + ++ +

(2)

Note: The subscript 7+ of the last beta indicates all the Time dummies that will be included.

For testing the differences between the different income groups classified by the World

Bank, dummies for income group and interaction terms between income groups and

political risk will be added and the model will look like this:

28 Hainz and Kleimeier (2006) use Euromoney’s economic performance index which include GDP and future projections for the economy. However as GDP is the standard measure to use for economic health of a country this will be sufficient for this paper. 29 As they note these measures are measures of the financial development of the country based on the stock market and the banking system (Hainz & Kleimeier 2006). 30 Only three are included to avoid perfect multicollinearity


48

0 1 2 3

4 5 6

7 9 10 12 13 15

16 18

_ _

_ _ _

_ _ _

_

it it it it

it it it

it it it

PF Corporate Governance GDP Quality institutions

Tension Conflict Policy Quality Bank influence

Time Dummies Income Dummies QI income

TC income

β β β ββ β ββ β ββ

− − −

−

= + + ++ + ++ + ++ 19 21 _it it itPQ incomeβ ε−+ +

(3)

Note: The subscriptions referring to several numbers indicate several variables. There are 3 time dummies (87-90, 91-95, 96-00), 3 income dummies (upper middle, lower middle, low). Each of the income dummies are multiplied with the political risk variable creating 3 interaction terms for each political risk variable (QI=Quality of Institutions, TC=Tension & Conflict, PQ=Policy Quality)

7.1.2 Method The data is comprised of several countries over several years and are thus a time series

cross section data set also referred to as panel data. One could use standard Ordinary

Least Square estimation (OLS) for estimating such a multiple regression. However,

there may be potential problems related to the assumptions of OLS. For OLS to be the

best unbiased estimate, the errors of the model should be uncorrelated and

homoscedastic. If this is not true, the estimates of beta will not be biased but the

standard error of the betas will be biased. This will lead to confidence interval being too

wide or too narrow and thus an incorrect conclusion (Wooldridge 2006).

In the model to be tested here there might be some dependence over the years if several

years are used for the same country. If a project finance loan is granted to a project in a

country and it turns out positively, there might be a better chance for other projects in

the same country to be granted such a loan. This will create correlation between the

years. When running the Wooldridge test for auto correlation in panel data in Stata, the

null-hypothesis of no first-order autocorrelation is rejected by a p-value of 0.009431.

Thus, autocorrelations should be taken care of in the model before being able to rely on

t-test for the coefficients due to the incorrect standard errors.

There might also be a difference in the variance of the errors across the panels (the

countries). This could be due to coincidence prevailing more in smaller countries which

would make the variance larger here than in larger countries. It is therefore necessary to

take account of the heteroscedasticity in the errors across the panels, when running the

model.

31 See appendix 8 for the test results from Stata


49

Contemporaneous correlation might also be an issue in this model. It can happen if there

is correlation across cross-sections at any point in time, which might be the case for

project finance loan in economic expansionary times, where all countries may see an

increase in the amount of loans granted. The fraction of project finance loans may then

develop very simultaneously over time. These “time shocks” have, however been dealt

with through the inclusion of year or period dummies. Testing for contemporaneous

correlation is not possible, as the data set is highly unbalanced with too few common

observations across the panels to perform a Pesaran's test in Stata.

In order to take all this into account, the regression will be done by running a regression

with panel-corrected standard errors (PCSE). This type of model is capable of

correcting the standard errors for autocorrelation and heteroscedasticity along with

contemporaneous correlation. Such analysis can be found in Stata under xtpcse32.

Xtpcse calculates panel-corrected standard error estimates for linear cross-sectional

time-series models where the parameters are estimated by OLS or Prais-Winsten

regression (Stata help for xtpcse).

Under estimation method, the form of the autocorrelation can be chosen along with a

method for calculating autocorrelation and whether the errors are correlated and

heterosecedastic or independent across panels. The heteroscedasticity is taken care of by

choosing the “hetonly” option in Stata. The form of the autocorrelation is assumed to be

the same for each panel, which means common AR(1). Although this may not be the

case, Beck and Katz (1995) state in their paper “What to do (and not to do) with Time-

Series Cross-Section Data” that estimating an individual autocorrelation for each panel

will make the estimation too uncertain as too many parameters are estimated33. They

have done Monte Carlo simulations to show that even though the autocorrelation

coefficient (rho) is not the same for all panels, it does not do too much harm to assume

the same correlation across all panels (Beck & Katz 1995). The calculation method for

the autocorrelation is the standard OLS single lag regression, which is default in Stata.

Once it is specified that there is autocorrelation in the data, the model is estimated by

Prais–Winsten FGLS assuming an AR(1) process in the disturbances34 and still

32 xt is the notation for panel data, pcse is panel corrected standard errors. 33 For this model, it would mean 69 estimates for autocorrelation on top of the 10-22 beta coefficients 34 See: Stata FAQ Comparing xtgls with regress, cluster


50

including Beck & Katz’s (1995) panel correction for standard errors. By using this type

of regression, the standard errors of the coefficients will be corrected; however there

might still be some unobserved country effects that are not captured. To correct for

these unobserved effects, random effect estimation or fixed effect estimation would

have to be used. However, these are ordinary General Least Square (GLS) with too low

standard errors, leading to a incorrect conclusion about the coefficients. As both

problems cannot be corrected, the most severe one is chosen and in this context the low

standard errors produced by GLS and possible incorrect conclusion, is perceived most

severe. Thus, PCSE method is chosen.

7.2 Data The dependent variable is measured as the fraction of project finance loans to

syndicated loans. Both syndicated loans and project finance loans have been gathered

from the Retures DealScan database, to which special access have been granted. The

fraction is measured at two levels: the number of loans and the volume of the loans.

These two may differ in the effect of the independent variables, as the number of loans

can increase from one year to another, while the volume is decreasing if the loans given

are smaller than previously. The fraction of project finance loans is used in order to take

into account the total amount of loans given to a country, as this investment level may

differ from one country to another. The variables are calculated as

#_ $ __ 100 _ 100

#_ $ _it it

it itit it

PF PFPF no PF vol

SL SL= ⋅ = ⋅ ,

respectively, where #_PFit is the number of project finance loans given to country i in

year t and $_PFit is the corresponding measure for the volume of project finance loans

measured in US dollars. #_SLit is the total number of syndicated loans (incl. project

finance loans) given to country i in year t and $_SLit the corresponding measure for the

volume35. They both vary from less than 1% in the US in 1988 to 100% in 35

observations, e.g. Tanzania 2000, Qatar 1998 and Indonesia 2003. In the sample

Western Europe, US, Canada, Australia and New Zealand accounts for 29.5% of the

observations and if countries are classified according to the World Bank’s income

35 3 countries had to be removed from the sample due to problems. Switzerland had no data given for total amount of syndicated loans, Nigeria seemed to have a project finance loan of 2 US dollars which is probably a mistake in the data and there where one post under project finance loans called “other” to which no country was specified.


51

classification of 4 levels36 about 39.5% of the observations are from high income group,

25.8% from upper-middle income, 23.7% from lower-middle and the last 11% from low

income countries37. Thus there is a high weight of high income countries from

industrialised countries, which also has to do with the data availability of such

countries38. The period analysed is 1987-2005. However, not all countries receive

project finance loans every year and so there are only one observation for 1987 and

1988 and 55.6% of the observations are from 1996-2000. This means that the sample

will be a highly unbalance panel data sample as noted earlier.

Both measures follow a highly non-normal distribution and hence both values have

been transformed by the use of log in order to make them more normally distributed39.

This will also help to diminish the effect of potential outliers in the sample, as they will

be given less weight if a log transformation is used.

As in the paper by Hainz & Kleimeier (2006), the corporate governance variable is an

equally weighted combination of stock market capitalization and domestic credit to the

private sector. Both are measured as a percentage of GDP and are obtained from the

World Bank Indicators database40. This variable is expected to be negatively related to

the fraction of project finance loans granted, as the better the corporate governance

system, the less private benefits there are to the managers and it is thus easier to solve

the model of double moral hazard for the firm by granting full-recourse loans. This in

turns means less project finance loans. As with the dependent variable, this variable is

log transformed in order to get a more normally distributed variable41.

36 The four levels are: High income, Upper middle income, Lower middle income and Low income. See World Bank analytical classification. The bounds for each income class have changed over the years which is why this excel sheet have been used to give the right classification to the country in the specific year in which project finance was granted. 37 To see a full set of all the countries included along with the years, the income classification and average risk please refer to appendix 9 38 Missing observation are also more prevalent in low income countries – only 58.3% of the observations from low income countries are complete for use in the regression. For all income groups it is mainly the variable “corporate governance” that is missing values. 39 See appendix 10 for the graphs of the dependent variables with and without log-transformation. 40 This database has no data for Taiwan and all data is missing for Monaco, Cayman Island and Laos either due to bank confidentiality or the inability to measure the variables. Thus all three countries have been eliminated from the sample entirely. Data on market capitalization is also missing for Belgium, Cameroon, Kuwait, Mali and Vietnam in the years requested thus also these are eliminated when running the regression. Thus a total of 69 countries are included in the sample. 41 The two graphs (log and level) can be found in appendix 11 along with the graphs for GDP.


52

GDP pr capita is the measure for economic health of the country and has been measured

in purchasing power parity (PPP) meaning it also takes into account the price level of

the country and thus the wealth of the people. For all countries it is measured in US

dollars for comparison reasons. It has been obtained from the World Bank Indicators

database just like the other firm moral hazard variables and is also log-transformed in

order to get a more normally distributed variable42. This variable is also expected to

have a negative sign since the model of double moral hazard would be better solved for

firm moral hazard by granting full-recourse loans if the health of the economy increases.

GDP ranges from USD 536 in Tanzania in year 2000 to USD 54707,5 in Luxembourg

in 1999 and thus include all levels of the World Bank classification of income from high

to low.

Bank influence is measured as the share of loans granted by a development bank to

country i in year t, where development banks are those defined by the World Bank as a

multilateral development bank or a multilateral financial institution43. These have been

chosen, as they are multilateral and large enough to represent an influence over the host

government, as the model of double moral hazard predicts. It is a fraction of total

syndicated loans given to the country in the specific year to represent the amount of

influence the bank will have in the country in that year and is measured at both number

and volume just like the dependent variable and the same “scale” is used in the

regression. Thus, when the dependent variable is measured at volume so is the bank

influence in order to keep to the same measuring method44. Bank influence is expected

to be positive as more influence will make it better to solve for the bank moral hazard,

as the cost of effort decreases. Hence, more influence leads to more project finance

loans.

42 See appendix 11 43 Those found in the sample are: The Asian Development Bank, The European Bank for Reconstruction and Development, The Inter-American Development Bank, The Islamic Development Bank, Corporacion Andina de Fomento, International Development Bank and International Finance Corporation (World Bank – multilateral development banks). 44 For validation reasons it has been tested by using the “wrong” scale and for both regressions it is the variable with the same scale that has the highest influence.


53

The last three variables in the regression is the political risk variables that where derived

from the political risk analysis in section 645. Those are calculated based on the loadings

from the Varimax component analysis and the ICRG political risk points assigned.

However, the points given by ICRG have been inverted in order to have more risk with

higher values. This way, there should be a positive relationship between the dependent

variable and the political risk variables, as more risk leads to more project finance loans

according to the model of double moral hazard.

In the model of double moral hazard it is predicted that banks have some influence on

the political risk of the project, as they can help deterring it be merely being part of the

deal. Therefore, the political risk variables that should be significant in explaining the

use of project finance should be the ones that a bank can indeed deter and thus have an

influence upon. This requires the variables to be related to actual government actions

interfering more directly with the project, actions which the government may refrain

from doing if they fear repercussion from influential banks. In relation to the ICRG

political risk variables, this means that component Tension & Conflicts, is not expected

to be significant in explaining the use of project finance in the model of double moral

hazard, as it is not depicting direct government actions which the bank can help defer if

given the right incentive.

On the other hand, Policy Quality contains actions by government, which could

interfere with the project, through the policies made, and that the bank might have

influence over, as it can with the Quality of Institutions, which covers legal institutions,

bureaucratic quality and corruption. Thus, both Policy Quality and Quality of

Institutions are expected to be significant and positive in the regression according to the

model of double moral hazard.

7.3 Empirical evidence Before running the regression descriptive statistics for all of the variables are presented

in table 7.2. Both dependent variables and corporate governance is given in percentage.

Bank influence is given in decimals.

45 Some countries that have been found to have received project finance loans are missing in the ICRG tables given for use in this paper. Those are: Bermuda, Macau, Seychelles and Macedonia. Therefore, these have been left out of the sample.


54

Both dummy variables for time and income level will be used as described earlier along

with interaction terms between political risk components and income levels.

To test the model, the regression described earlier, is run and the results can be found in

table 7.3 below. For each regression there are 289 observations from a total of 69

countries.

Table 7.2: Desriptive statistics for all variables

Variable Observations Mean Std.dev. Minimum Maximum

329 25.532 29.2749 0.57 100

329 29.385 31.8910 0.1425 100

Corporate Governance 294 47.117 30.7930 3 192

GDP 329 13432.88 10160.02 536 54707.5

Quality of Institutions 321 20.258 10.2581 1.6935 41.2117

Tension & Conflicts 321 12.998 6.9937 0.5294 33.3383

Policy Quality 321 11.407 4.2677 1.5369 22.3380

Bank Influence (volume) 329 0.0355 0.1242 0 1

Bank Influence (numbers) 329 0.0511 0.1542 0 1

D 87_90 329 0.0334 0.1800 0 1

D 91_95 329 0.2127 0.4099 0 1

D 96_00 329 0.5562 0.4976 0 1

D 01_05 329 0.1976 0.3988 0 1

D high 329 0.3951 0.4896 0 1

D upper 329 0.2584 0.4384 0 1

D lower 329 0.2371 0.4259 0 1

D low 329 0.1094 0.3126 0 1

QI_high 321 4.4661 6.8701 0 28.5364

QI_upper 321 5.4524 10.1109 0 41.0562

QI_lower 321 7.1025 12.8547 0 41.2117

QI_low 321 3.2369 9.5538 0 38.7203

TC_high 321 3.0789 4.6605 0 26.7062

TC_upper 321 3.1574 6.1108 0 31.3072

TC_lower 321 4.6335 8.4846 0 30.2995

TC_low 321 2.1267 6.4603 0 33.3383

PQ_high 321 3.4596 4.7675 0 17.1655

PQ_upper 321 2.8448 5.2356 0 19.2172

PQ_lower 321 3.5091 6.4359 0 22.3380

PQ_low 321 1.5931 4.7061 0 21.8583

Share of Project finance loans in numbers

Share of Project finance loans in volume

Note: D indicates a dummy variable and the political risk variables have been abbreviated in the interaction terms as follows: QI: Quality of Institutions, TC: Tension and Conflicts, PQ: Policy Quality. For the income dummies, upper refers to upper middle income and lower refers to lower middle income


55

As can be seen, most of the predictions from the previous section are fulfilled. For

regressions (1) on both dependent variables, corporate governance is significant and has

a negative influence, on the number and volume of loans given as project finance loans,

rather than full recourse syndicated loans. As both the dependent variable and the

independent variable is measured in log the coefficient is a measurement of elasticity.

Thus the elasticity of project finance with respect to corporate governance is -0.235 for

the fraction measured in numbers and -0.333 for the fraction measured in volume.

This means that a better corporate system leads to less project finance as governance is

in place to deal with firm moral hazard and thus making it better to solve the double

moral hazard for the firm by granting full-recourse loans.

Quality of Institutions is also positive as projected, meaning that a higher level of risk in

the Quality of Institutions leads to a higher fraction of project finance loans. Thus if

quality of institutions is increased by one point the country will receive 3.3% more

loans as project finance loans and the size of the loans will increase by 7.4%. Thus

Corporate Governance -0.235 (0.086)*** -0.333 (0.123)** *

GDP -0.175 (0.101) -0.301 (0.147)**

Quality of Institutions 0.053 (0.017)*** 0.074 (0.025)* **

Tension & Conflicts -0.009 (0.025) -0.024 (0.036)

Policy Quality -0.102 (0.025)*** -0.107 (0.036)***

Bank Influence 0.932 (0.329)*** 0.609 (0.532)

D 87_90 -0.189 (0.381) -0.046 (0.514)

D 91_95 0.136 (0.243) 0.568 (0.348)

D 96_00 -0.165 (0.163) 0.060 (0.231)

Constant 5.345 (1.167) 6.374 (1.673)

Observations 289 289

Countries 69 69

R squared 0.4782 0.2804

Wald Chi squared 77.59 (0.000) 62.09 (0.000)

(1)

Project finance share based on number of loans

Project finance share based on volume of loans

(1)

Table 7.3: Results from regressions

Note: Base period is 2001-2005 for the regressions. D indicates a dummy variable. The numbers in the parentheses are the std. errors of the coefficients. The numbers in bold are significant at a 1%, 5% and 10% level reported by *, ** and *** respectively. The bank_influ variable is bank influence measured at the same scale as the dependent variable – in numbers or in volume. The number in parentheses after Wald Chi square is the significance level of that test. The regressions have also been run using year dummies rather than period dummies and yield an almost similar result although GDP becomes significant even in the regression with project finance loans measured in numbers. In both regressions GDP will be significant at a 10% level.


56

although investors perceive quality of the institutions as a risk to the project the model

of double moral hazard explains that such risk can be influenced by the bank and that

non-recourse financing through project finance deals are more efficient in such settings.

The more interesting results from the regression is that Policy Quality is negative,

which means that more risk in this component would lead to a smaller fraction of

project finance loans; an effect that is more substantial than the positive effect of

Quality of Institutions namely 10.2% less loans and 10.7% smaller volume of those

loans. Thus, not only will fewer loans be given but the loans will also be smaller. This is

in direct contradiction to the predictions of the model of double moral hazard. It

suggests that although it was found that project finance helps deter the risk of bad

quality of institutions it does not seem to deter the risk of governments making

unexpected changes, expropriating investments or restricting currency convertibility. In

other words, this study finds that project finance loans help deter the risk of poor

institutions and legal systems through banking influence as stated by the model of

double moral hazard but that worse policies and their predictability will lead to a

smaller fraction of loans being made as project finance loans. A possible explanation to

this could be that the banks influence is not usable for influencing the quality of policies

or the stability of the government for one project, as these aspects will affect all

investors. On the other hand they are able to help secure a fair treatment of the project if

the quality of the institutions is bad.

As expected Tension & Conflicts are not significant in the model of double moral

hazard since this is not part of government actions that the bank can deter. Hence, all

investors perceive this as a risk to an investment and as mentioned under “Risks of

project failure” it is a risk that will affect all investments and may thus deter all

investments from such countries. In other words, both the total syndicated loans and the

project finance loans may decrease but the fraction of project finance loans granted will

then be unchanged.

This also shows that the aggregation of all political risk variables in Hainz and

Kleimeier’s (2006) paper is problematic. Different aspects of political risk affect the

level of project finance differently. This study has established that there are 3 opposite

effect ranging from negative, to none, to positive. There are aspects like the internal and

external conflicts of the country along with religious and ethnical tensions that are not

important to the fraction of project finance given, and the two types of governmental


57

influence have opposite effects. Direct government interference through expropriation

or currency convertibility leads to less project finance whereas government influence

through the quality of it’s institutions are related positively to the fraction of project

finance loans granted. These findings are not possible to find unless aggregated political

risk variables are sub divided into sub-groups representing different aspects of the

comprehensive variable political risk. So when Hainz and Kleimeier (2006) find a

positive effect of political risk upon the fraction of project finance loans given, then it is

the risk of poor quality of institutions that prevail. The positive effect of the risk of poor

quality of institutions is however, not the whole story, as the risk of poor quality of the

policies actually decreases the fraction of project finance loans granted. Hence, the

different aspects of the wide-ranging concept of political risk should be included in the

model of double moral hazard.

When it comes to the influence of development banks it can be seen that there is a

difference between the two dependent variables. For the fraction of project finance

loans measured at number of loans, banking influence is statistically significant and

positive in accordance with the model of double moral hazard46. However, when

measuring the fraction in volume, banking influence is not important. If interpreting the

number of loans given as an indication of whether a loan is given or not, then this would

suggest that banking influence is only important for whether the loan is given or not but

it does not affect the size of the loan. Thus, the more influence the development banks

have the more loans are given but banking influence is not significant in predicting the

volume granted.

GDP as a proxy for the economic health of the country is very close to being significant

to the number of project finance loans given at a 10% alpha level and it is significant (at

a 10% level) if year-dummies are used. However, when measuring the fraction of

project finance loans in volume GDP is significant at a 10 percent alpha level with an

expected negative sign in regression (1) with only period dummies. Thus the richer the

country, the smaller the volume of the project finance loans that is granted. This means

that the model of double moral hazard is right in its prediction of better economic health

46 As mentioned earlier even when using the other scale, being volume of fraction of loans given by development banks, it is significant although less.


58

of the country leading to fewer project finance loans compared to the number of

syndicated loans.

When comparing R-square for the two regression’s, it is clear that the regression

measured as number of project finance loans, perform better with a higher R-square and

higher Wald chi test score – a test testing the joint significance off all the variables. This

test shows that both models have variables that are jointly significant. The same

conclusion is reached by Hainz and Kleimeier (2006) and they contribute it to the lack

of industry specification. The volume of a loan is highly dependent upon which industry

it is given to. Construction, oil and gas have far higher loan amounts than projects

within telecommunication, utilities and mining, which means that the fraction of project

finance loans measured in volume, will suffer from this bias, whereas the fraction of

project finance loans measured in number of loans will not suffer from this bias. Thus

the model will be a better fit when the latter is used.

The regressions have been run for a world sample of 289 observations of which 119 was

from high income countries. This means that they have a high weight in the results

found. Therefore in order to validate the results found in regression (1) a regression has

been run for only middle and low income countries. The result is somewhat different

and can be found in table 7.4, regression (2).


59

Neither of the firm moral hazard variables is significant to regression (2) with any of

the dependent variables. However, the political risk variables shows the same picture for

the fraction of project finance deals measured in numbers, meaning that poor quality of

institutions would lead to a 5,2% higher fraction of loans made of project finance deals

and a worse quality of policies would lead to 7,8% smaller fraction of project finance

deals. For the volume of the loans only the quality of the policies seems to be important

to the share of project finance loans used in a country by a negative 9%. Bank influence

Note: Regression (2) is for middle and low income groups. Regression (3) is based on the world sample and includes dummies for income (D indicates a dummy variable) and interactions between income group and political risk variables. For the interaction terms abbreviations for the political risk variables are as follows: QI: Quality of Institutions, TC: Tension & Conflict, PQ: Policy Quality. Base period is 2001-2005 for all regressions and for regression (3) high income is base. The numbers in the parentheses are the std. errors of the coefficients. The numbers in bold are significant at a 1%, 5% and 10% level reported by *, ** and *** respectively. The bank_influ variable is bank influence measured at the same scale as the dependent variable – in numbers or in volume. The number in parentheses after Wald Chi square is the significance level of that test.

Table 7.4: Results from regressions with difference in income levels .

Corporate Governance -0.089 (0.107) -0.227 (0.086)*** -0.106 (0.150) -0.349 (0.124)***

GDP -0.156 (0.129) -0.045 (0.209) -0.178 (0.173) -0.119 (0.291)

Quality of Institutions 0.052 (0.022)** 0.050 (0.029)* 0.036 (0.032) 0.126 (0.045)***

Tension & Conflicts -0.027 (0.029) 0.006 (0.042) 0.007 (0.039) -0.084 (0.077)

Policy Quality -0.078 0.034)** -0.105 (0.033)*** -0.090 ( 0.048)* -0.084 (0.047)*

Bank Influence 0.943 (0.346)*** 0.807 (0.328)** 0.656 (0.563) 0.457 (0.546)

D 87_90 0.479 (0.637) -0.117 (0.394) 0.700 (0.847) 0.060 (0.520)

D 91_95 -0.018 (0.295) 0.183 (0.254) 0.436 (0.408) 0.477 (0.371)

D 96_00 -0.177 (0.194) -0.139 (0.165) 0.242 (0.265) 0.051 (0.240)

D upper 0.370 (0.580) 0.954 (0.734)

D lower 0.844 (0.842) 1.808 (1.256)

D low -2.248 (1.702) -1.946 (2.683)

QI_upper -0.037 (0.056) -0.096 (0.072)

QI_lower -0.013 (0.046) -0.142 (0.074)*

QI_low 0.091 (0.046)** 0.002 (0.075)

TC_upper 0.012 (0.064) 0.099 (0.096)

TC_lower -0.018 (0.063) 0.147 (0.102)

TC_low -0.083 (0.063) 0.015 (0.116)

PQ_upper 0.036 (0.056) -0.015 (0.074)

PQ_lower 0.004 (0.051) -0.036 (0.076)

PQ_low 0.102 (0.121) 0.119 (0.215)

Constant 4.711 (1.331) 3.917 (2.228) 4.703 (1.872) 4.283 (3.108)

Observations 170 289 170 289

Countries 44 69 44 69

R squared 0.4303 0.4892 0.1233 0.3951

Wald Chi squared 25.49 (0.003) 103.82 (0.000) 9.59 (0.3847) 92.99 (0.000)

Project finance share based on number of loans

Project finance share based on volume of loans

(2) (3) (2) (3)


60

also shows the same picture by being significant to the number of loans given but not

the volume of those loans.

The model with project finance loans measured at volume thus only has one significant

variable, combined with a very low R-square and is insignificant as the Wald Chi test

produces a p-value of (0.3847). This means that all the variables are jointly

insignificant. For project finance loans measured at number of loans the corresponding

p-value is (0.003) which means it is far better at explaining the use of project finance

deals in middle and low income countries. These results would therefore suggest that

investors still look at political risk when deciding whether to set up a project in a middle

or low income country, but once it has been determined to establish a project, the

volume of the associated project finance loan is not necessarily determined by the

factors of the model of double moral hazard that seems of little significance as only

quality of the policies in place may lower the amount given.

The sample for these regression are approximately 40% smaller than the world sample

and for at least lower middle and low income countries, the data available tends to be

less reliable, which all in all may be part of the change in significance.

There is however, most likely difference between the income groups in the effect of the

independent variables and as political risk is of special interest to this study, dummies

for the income levels and interaction terms between the political risk variables and the

income levels have been added in regression (3), in order to see if there is a difference

in the effect of political risk between the high income group and the other income

groups. Including income dummies do however, pose another problem since the income

level variable is highly correlated with GDP (correlation is -0.9439), meaning that there

is multi collinearity between these two variables. That will cause a higher variance of

the coefficients leading to less acceptance of significance47. If the interaction terms

needs to be study the dummy variables must however, be included according to

Brambor et al. (2006) to avoid omitted variable bias and the high correlation must then

be accepted. Once the interaction terms are included it would be expected that the

political risk variables have a higher effect at lower levels of income as these tend to

also be of a higher risk. The correlation between income level and the overall ICRG

47 As expected GDP becomes insignificant in both regression once the income dummies are included.


61

political risk measure, is -0.7383, meaning that a higher level of risk leads to a lower

level of income48.

For regression (3) on the fraction of project finance loans measured in numbers of loans

there is little change in the political risk variables that now represents the high income

countries, probably because the world sample is heavily weighted by high income

countries. The Policy Quality does however have a smaller though still negative effect.

It can also be seen that the high income level differs significantly from the low income

level when it comes to the effect of Quality of Institution that is significantly higher in

low income countries. In fact, if the low income group is set as the base case in the

regression, it is found that the effect of the quality of the institutions is significantly

higher for low income countries than for all the other income groups as was expected.

This means that if the Quality of Institutions in low income countries drops the increase

in the fraction of project finance loans is higher than in other countries.

Both middle income countries are negatively different from high income dragging the

effects close to zero and although the differences are insignificant, it turns out that the

Quality of Institutions is insignificant to the fraction of project finance loans granted in

both middle income groups49.

When looking at the regression for volume of project finance loans, including the

interaction terms have three implications. First of all, for the political risk variables the

effect of Quality of Institutions is higher for high income countries than for the world

sample, whereas the Policy Quality is less negative. At the same time GDP become

insignificant, which may be due to the inclusion of the income dummies and the

subsequent multi-collinearity. The third observation is that Quality of Institution is

significantly more negative in lower middle income countries than in high income

groups, whereas neither the upper middle nor low income countries differ from the high

income in the effect of this variable. Thus, for lower middle income group the effect of

institutional quality is less positive, and actually insignificant if the regression is run

with lower middle income group as base case of the regression. The same is true for the

48 Also the correlation between the income level variable and the three risk components is high ranging from 0.6054 to 0.7649. 49 These results are found by making either of the income groups base case of the regression so that the political variables represent that income level and has the std.error and t-test for the coefficient given. To see these results please refer to appendix 12.


62

upper middle income group. Although none of the income groups seems to differ from

the high income group in effect of Policy Quality, it turns out that once upper middle or

low income group is made the base, Policy Quality is insignificant.

This would suggest that the positive significant relationship predicted by the model of

double moral hazard between the political risk component representing the quality of

institutions is only true for the high and low income group.

Base on these regressions it can be seen that the income groups differs highly, with

respect to which types of political risk that is significant in explaining the level of

project finance loans granted. This is true even though only one income group differs

from the high income in each regression. To summaries which of the political risk

variables that are significant for the different income groups the following table has

been produced.

Table 7.5: The significant political risk variables50

It can be seen that high income is the only income group having the same results for

political risk variable significance as the initial regression, which is not surprising given

that 41.1% of the total sample is from high income countries. For the upper middle

income group none of the political risk variables are significant to the fraction of project

finance loans given to a country, which means a failure of the model of double moral

hazard for this subgroup. For the lower middle income group only Policy Quality is

significant meaning that in contradiction to the model of double moral hazard political

risk is negatively related to project finance loans granted. Hence for lower middle

income countries it seems that a higher risk of bad policies will lead to less project

finance loans, which means that the model of double moral hazard has been solved for

the firm’s moral hazard. For the low income countries the model is right in its

predictions that political risk will lead to more project finance loans relative to full- 50 Please refer to appendix 12 for the different regressions.

Note: The stars in the parenthesis is the significance level with 1, 5 and 10 % represented by ***, ** and * respectively.

High Quality of Institution (*) Policy Quality (***) Quality of Institution (***) Policy Quality (*)Upper middle - - - -Lower middle - Policy Quality (**) - Policy Quality (*)Low Quality of Institution (***) - Quality of Institution (**) -

Share of project finance loans measured in volumeIncome

Share of project finance loans measured in numbers


63

recourse loans, however only for political risk related to the quality of institutions being

corruption, bad property rights and legal systems and less responsive political systems.

The poor quality of policies in these countries can affect all investment types and could

hence deter all types of syndicated loans leading to a lower level of loans but no change

in the fraction of project finance loans granted.

It should here be noted once again that the conclusions are only as good as the data used

to draw them from and data from the lower middle and low income countries tends to

be of worse quality and thus less reliable. At the same time the sub-samples of the upper

middle, lower middle and low income groups are smaller than the high income, 73, 76

and 21 observations respectively, which at least for the low income group is relatively

low for inference from regressions.

In summary two aspects of political risk is of importance to the share of loans given to a

country as project finance loans. A poor quality of the institutions will lead to more

project finance loans being granted as a fraction of total syndicated loans, whereas a

inferior quality of the policies will have the opposite effect. Banking influence is

important only to the decision about giving the loan or not but once it has been decided

the bank has no influence on the volume of the loan. For the firm moral hazard variables

only the corporate governance of the country is significant in all regression in

explaining the fractions of project finance loans granted, whereas the economic health

of the country is of importance to the volume of loans granted and to the number of

loans once year dummies are used.

Thus the model of double moral hazard fails in certain places in explaining the use of

project finance rather than full-recourse syndicated loans. For the bank moral hazard it

seems that it is the increased risk of bad institutions that leads to solving for this moral

hazard, whereas the increased risk of inferior policies seems to lead to a solving of the

double moral hazard for the firm by granting more full-recourse syndicated loans. The

banks effort measured by its influence over the host government is important in

determining whether the loans are given but not the volume of the loans.

The failure of the model of double moral hazard is less severe for the firm moral hazard

part. It seems that mainly the cost of effort measured as the level of corporate

governance, is important to the firm moral hazard. The increase in the probability of


64

success, from the firm exercising an effort, is also important at least as long as the time

dummies are incorporated for each year.

When excluding high income countries from the sample, only political risk variables are

significant and when looking at the difference in the effect of political risk, it is only in

high and low income countries that Quality of Institutions shows the positive significant

effect predicted by the model. For the lower middle income group it is only the negative

effect of Policy Quality that is significant and for upper middle income levels neither of

the political variables is important.

This insight has important implication for lenders and sponsors. Political risk affects the

loans differently and an efficient solving of the model of double moral hazard has is

dependent upon which type of risk is prevailing. If the quality of institutions is bad then

it is more efficient to solve for the bank’s moral hazard by given a non-recourse loan.

However, if the main risk in the country is related to the quality of the policies then it

will be more efficient to give a full-recourse loan and thus solve the moral hazard of the

firm.


65

8. Conclusion This paper has analysed the concept of project finance and political risk in order to test

the ability of the model of double moral hazard to explain the relationship between these

two variables. It was initially found that many risks can affect a project and that lenders

of non-recourse loans are specially concerned by these as their loans can only be

serviced through the cash flows generated by the project itself. As a number of the

studies on project finance have discussed, the use of project finance is often used as a

mean of deterring especially political risk. The main political risks to a project were

“changing legislation”, “war and conflicts” and “legal system and property rights”.

The best way of measuring political risk is through a measure of subjective perception

of political risk. The perceptive measure WGI was used by Hainz and Kleimeier (2006)

but possesses undesirable characteristics as it is a non-consistent aggregation of several

other primary measures making it conceptually flawed and unusable for comparison

across time and countries. Instead the ICRG’s political risk index of 12 variables was

used and a principal component analysis was conducted on these variables, through

which 3 components were extracted. These components represent: Quality of

Institutions, Tension & Conflicts and Policy Quality, these are all consistent with the

variables found to be of importance to a project investor and lender, in the beginning of

the paper.

The use of these three components lead to the main finding of this paper, that political

risk in regards to project finance cannot merely be represented by a single variable.

Different aspects of political risk have different effects upon the share of project finance

loans granted. The inadequacy of institutions does lead to more project finance loans,

which leads to the conclusion that such syndicated loan types can help deter some of

those risks. Policy inferiority on the other hand will lead to less project finance loans

being granted, suggesting that the perception that project finance loans helps deter

government actions such as expropriation or simply the uncertainty of government

policy is wrong. The opposing effects that the two significant political risk variables

obtain in the regression, thus leads to the conclusion that banks influence is not usable

for influencing the quality of policies or the stability of the government, as these risks

will affect all investors rather than only the project the bank is engaged in. On the other


66

hand they are able to help secure a fair treatment of the project if the quality of the

institutions is bad.

The last component of political risk is termed Tension & Conflict. This variable has no

effect on whether project finance loans are preferred to full-recourse syndicated loans.

Thus, a measure of political risk in the model of double moral hazard must incorporate

more than a single measurement. It is necessary for such a measurement to incorporate a

number of angles to be able to make a stronger conclusion and fully capture the

relationship between political risk and project finance.

At the same time it was found that development bank influence is only significant to

whether the loan is given or not. It does however not affect the actual size of the loan

granted. Hence if the bank is to secure the loans against political risk they will do so, no

matter the size of the loan.

It thus seemed that the model of double moral hazard failed in several places in

explaining the use of project finance. Only certain types of political risk show the

predicted positive relationship between more risk and solving for the bank moral hazard

by granting a non-recourse loan. The bank’s effort in helping the project succeed is thus

limited to the areas of political risk where their influence can have a direct influence on

the risk of the project.

It was also shown that the model has several short comings when excluding the high

income countries from the sample. The firm moral hazard was of no importance and

only political risk and bank influence can explain the use of project finance. Thus, in

middle and low income countries, it is necessary to use other or additional variables to

capture the use of project finance loans over other syndicated loans and hence, explain

when moral hazard is solved for the firm. This would be a possible subject for further

research.

There is some ground for the model of double moral hazard as long as the user/lender

pays attention to the type of political risk facing the investment as there are different

effects of the different types of political risk. In addition it is important that the income

level of the country is taken into account.


67

9. Critical perspective This study offers an assessment of the model of double moral hazard and its ability to

explain the relationship between political risk and the use of project finance loans once

different aspects of political risks are incorporated. The relationship between project

finance and political risk is not as easily explained as previous literature would have us

think. Political risk cannot merely be represented by a single measure, as different

aspects of political risk have differing effects on project finance utilization. This

suggests that a more complex model is necessary to explain the choice between full-

recourse loans and non-recourse loans than the model of double moral hazard consisting

of effort and increase in probability of success from the two participating parties.

Project specific characteristics such as the industry of the project and the sponsoring

firms experience may be influential as to whether the loan is given as a non-recourse

loan or a full-recourse syndicated loan. Also unobserved country effects may, as

mentioned in the paper, influence the loan process. It would however require a more

balanced data set than the one presented here, to capture this.

Another field of future research would be the difference presented between income

levels. Taking into consideration the higher share of project finance loans in lower

middle and low income countries, this difference needs to be explored more in depth to

fully understand the reasons behind the use of project finance loans where they are

highly used. This paper concluded that the effect of the three political risk components

that were derived in the principal component analysis differed across the various income

levels. Thus, the parameter heterogeneity makes it necessary to explore those

differences more properly including possibly making individual principal component

analyses for each income level as it was found that those differed.

A weakness of this study is the low number of observations from the post-period of the

terror attack on the World Trade Centre in 2001. This event may very well have

changed the investors’ perception of risk along with the increased war on terror. Thus

parameter heterogeneity may prove to be a problem over time. This has not been part of

this study due to the limited amount of data on project finance loans after 2002.

At the same time there are a small number of countries with only one entry of a project

finance loan. These may be more arbitrary once again reflecting the need for a more

complex model to explain the use of project finance.


68

A more comprehensive model may help guide sponsors and lenders in the use of project

finance and may also prove to be beneficial to less developed countries as better

knowledge of the access to foreign capital markets may help their growth path. Such

investigations are left for further research.


69

Bibliography Ahmed, Priscilla A., 1999, Project Finance in Developing Countries, Lesson of experience series, no 7, International Finance Corporation, Washington D.C., Akerlof, George A. 1970, The market for “Lemons”: Quality, Uncertainty and the market mechanism, The Quarterly Journal of Economics, Vol. 84, No.3, pp.488-500 Beck, N. & Katz, J.N., 1995, What to do (and not to do) with Time-Series Cross-Section Data, The American Political Science Review, Vol. 89, No. 3, pp. 634-647 Berggren N., Bergh A., Bjørnskov C., 2008, The growth effects of institutional quality, instability and change, Mimeo ASB Brambor, T., Clark, W.R., Golder, M. 2006 Understanding interaction models: Improving Empirical Analyses, Political analysis, No 14, pp.63-82 Brunelli, A., 1997, Political variables in cross-country growth analysis, Journal of economic surveys, vol.11, no.2 Brunetti, A. & Weder, B., 1995, Political sources of growth: A critical note on measurement, Kluwer Academic Publishers, Public choice 82, pp.125-134 Eatwell, J., et.al. 1987-94 “The new Palgrave: a dictionary of economics (K-P)”, Macmillan Press Limited pp. 549 – Moral Hazard by Y. Kotowitz Esty, Benjamin C., 2004, Modern Project Finance – A case book, John Wiley & Sons Feng, Y., 2001, Political freedom, political instability, and policy uncertainty: A study of political institutions and private investment in developing countries, International studies quarterly, 45, pp.271-294 Fight, A., 2006, Introduction to project finance, Essential capital markets, Elsevier, Butterworth-Heinemann Finnerty, John D., 2007, Project Finance: Asset-Based Financial Engineering, 2nd edition, John Wiley & Sons Hainz C., & Kleimeier S., Dec.2006, Project Finance as a Risk-Management Tool in International Syndicated Lending, Discussion paper No.183, Governance and the efficiency of economic systems (GESY) Hair, J.F., et.al. 2005, Multivariate data analysis, chap.3 (pp.101-166), 6th edition, Pearson Education Inc – Prentice Hall Harms, P., 2002 Political risk and equity investment in developing countries, Applied economics letters, 9, pp.377-380


70

Howell, L.D, & Chaddick, B. 1994, Models of political risk for foreign investment and trade - An Assessment of three Approaches, The Columbia Journal of World Business, Vol. 29, Issue 3, pp.70-91 Isham, J. et al., 1997, Civil liberties, democracy and the performance of government projects, The world bank economic review, vol.11, no.2, pp.219-42 Iqbal, Kazi & Shah, Anwar, 2008, Truth in Advertisement: How do Worldwide Governance Indicators Stack Up?, Unpublished, presented at the CEA Session: Institutions and Measurement, Vancouver, June 6-8. Knack, S. & Keefer, P., 1995, Institutions and economic performance: Cross-country tests using alternative institutional measures, Economics and Politics, Vol. 7, No. 3, pp.207-227 Murshed, S. Mansoob & Sen, Somnath, 1995, Aid Conditionality and Military Expenditure Reduction in Developing Countries: Models of Asymmetric Information, The Economic Journal, Vol. 105, No. 429 (Mar., 1995), pp. 498-509 Nevitt, Peter K., 2000, Project Financing, 7th edition, Euromoney Books PRS group, 2008 International Country Risk Guide Methodology, (Web) Ray, D. 1998, Development Economics, 1st edition, Princeton University Press, Princeton, New Jersey, Chapter 3, pp51-71 Tinsley, Richard, 2000, Advanced Project Finance - Structuring risk, 1st edition, Euromoney Books Yescombe, E.R, 2002, Principles of project finance, Academic Press Wooldridge, J. 2006, Introductory econometrics – a modern approach, 3rd edition, international student edition, Thomson, South Western Web pages: Political Risk Service group – ICRG methodology http://www.prsgroup.com/ICRG_Methodology.aspx (Accessed 11th of June 2008) STATA FAQ online: Testing for panel-level heteroskedasticity and autocorrelation: http://www.stata.com/support/faqs/stat/panel.html (Accessed 9th July 2008) Comparing xtgls with regress, cluster:

http://www.stata.com/support/faqs/stat/xtgls_rob.html (Accessed 9th of July 2008) STATA help online Xtpcse: http://www.stata.com/help.cgi?xtpcse (Accessed 9th of July 2008)


71

World Bank Multilateral development banks: http://web.worldbank.org/WBSITE/EXTERNAL/EXTABOUTUS/0,,contentMDK:20040612~menuPK:41694~pagePK:43912~piPK:44037~theSitePK:29708,00.html (Accessed 14th of July 2008) World Bank analytical classification http://go.worldbank.org/0CO1RKFBP0 (Accessed 20th of July 2008)


72

Appendix 1 The development in number of project finance loans over time and across income and risk levels.

Average percentage of syndicated loans made as proj ect finance loans

for the different political risk level over the 4 p eriods analysed

5037

34

36

35

50

35 25

49

32

10

1316

25

16

36

01-05

20

87-05

Very low

Low

Moderate

High

87-90 96-00

1811

91-95

Average percentage of syndicated loans made as proj ect finance loans for the different income level over the 4 periods a nalysed

50

22

32

31

30

5036

22

33

26

10 16 13

32

17

47

01-05 87-05

High

Upper middle

Lower middle

Low

87-90

39

91-95

44

96-00

66


73

Appendix 2 Yescombe’s (2002) risk matrix specified

Commercial viability

Site acquisition and accessPermits

EPC contractor (engineering, procurement and construction contract)Construction cost overruns

Revenue during constructionDelay in completion

Inadequate performance on completion Third party risks

Projects without a fixed-price, date-certain, EPC contract

Environmental risks

TechnologyGeneral project operation

General operating cost overrunsProject availability

MaintenanceDegradation

Off-take contractsConcession agreements

Hedging contractsContracts for differences (electricity industry)

Long-term sales contractsPrice and volume risks

Risks for the off-taker or contracting authority

Input supply risks Input supply contracts

TemporaryPermanent

Contract mismatch

Sponsor support

Inflation

Interest rate risks Interest rate swaps

HedgingFinance in more currency

Conversion of local revenuesDevaluation

Currency convertibility and transferExpropriation

War and civil disturbanceEnvironmental requirements

Price controlsRemoval on price controls on input material

Import duties or controlsIncrease in taxes (general corporate or special or on dividends)

Deregulation or privatisationEmployment, health, safety rules

Amendment or withdrawal of permitsInvalidation or project

Breach of contract and court decisionsSub-sovereign risks

Creeping expropriation

Commercial risks

Completion risks

Operating risks

Revenue risks

Force majeure risks

Change of law risks (new legislation, new

regulation, new interpretation)

Investment risks

Macro-economic risks

Quasi-political risks

Political risks

Exchange rate risks

Source: Yescombe 2002


74

Appendix 3 Political risk components from ICRG political risk index51 The following risk components, weights, and sequence are used to produce the political risk rating:

Sequence Component point(max)

A Government Stability 12B Socioeconomic Conditions 12C Investment Profile 12D Internal Conflict 12E External Conflict 12F Corruption 6G Military in Politics 6H Religious Tensions 6I Law and Order 6J Ethnic Tensions 6K Democratic Accountability 6L Bureaucracy Quality 4Total 100

POLITICAL RISK COMPONENTS

Source: ICRG methodology Government Stability – 12 Points This is an assessment both of the government’s ability to carry out its declared program(s), and its ability to stay in office. The risk rating assigned is the sum of three subcomponents, each with a maximum score of four points and a minimum score of 0 points. A score of 4 points equates to Very Low Risk and a score of 0 points to Very High Risk. The subcomponents are: • Government Unity • Legislative Strength • Popular Support Socioeconomic conditions – 12 Points This is an assessment of the socioeconomic pressures at work in society that could constrain government action or fuel social dissatisfaction. The risk rating assigned is the sum of three subcomponents, each with a maximum score of four points and a minimum score of 0 points. A score of 4 points equates to Very Low Risk and a score of 0 points to Very High Risk. The subcomponents are: • Unemployment • Consumer Confidence • Poverty Investment Profile – 12 Points This is an assessment of factors affecting the risk to investment that are not covered by other political, economic and financial risk components. The risk rating assigned is the sum of three subcomponents, each with a maximum score of four points and a minimum score of 0 points. A score of 4 points equates to Very Low Risk and a score of 0 points to Very High Risk. The subcomponents are: • Contract Viability/Expropriation • Profits Repatriation • Payment Delays

51 The entire appendix is an exact reproduction of the ICRG methodology.


75

Internal Conflict – 12 Points This is an assessment of political violence in the country and its actual or potential impact on governance. The highest rating is given to those countries where there is no armed or civil opposition to the government and the government does not indulge in arbitrary violence, direct or indirect, against its own people. The lowest rating is given to a country embroiled in an on-going civil war. The risk rating assigned is the sum of three subcomponents, each with a maximum score of four points and a minimum score of 0 points. A score of 4 points equates to Very Low Risk and a score of 0 points to Very High Risk. The subcomponents are: • Civil War/Coup Threat • Terrorism/Political Violence • Civil Disorder External Conflict – 12 Points The external conflict measure is an assessment both of the risk to the incumbent government from foreign action, ranging from non-violent external pressure (diplomatic pressures, withholding of aid, trade restrictions, territorial disputes, sanctions, etc) to violent external pressure (cross-border conflicts to all-out war). External conflicts can adversely affect foreign business in many ways, ranging from restrictions on operations, to trade and investment sanctions, to distortions in the allocation of economic resources, to violent change in the structure of society. The risk rating assigned is the sum of three subcomponents, each with a maximum score of four points and a minimum score of 0 points. A score of 4 points equates to Very Low Risk and a score of 0 points to Very High Risk. The subcomponents are: • War • Cross-Border Conflict • Foreign Pressures Corruption – 6 Points This is an assessment of corruption within the political system. Such corruption is a threat to foreign investment for several reasons: it distorts the economic and financial environment; it reduces the efficiency of government and business by enabling people to assume positions of power through patronage rather than ability; and, last but not least, introduces an inherent instability into the political process. The most common form of corruption met directly by business is financial corruption in the form of demands for special payments and bribes connected with import and export licenses, exchange controls, tax assessments, police protection, or loans. Such corruption can make it difficult to conduct business effectively, and in some cases my force the withdrawal or withholding of an investment. Although our measure takes such corruption into account, it is more concerned with actual or potential corruption in the form of excessive patronage, nepotism, job reservations, ‘favor-for-favors’, secret party funding, and suspiciously close ties between politics and business. In our view these insidious sorts of corruption are potentially of much greater risk to foreign business in that they can lead to popular discontent, unrealistic and inefficient controls on the state economy, and encourage the development of the black market. The greatest risk in such corruption is that at some time it will become so overweening, or some major scandal will be suddenly revealed, as to provoke a popular backlash, resulting in a fall or overthrow of the government, a major reorganizing or restructuring of the country’s political institutions, or, at worst, a breakdown in law and order, rendering the country ungovernable. Military in Politics – 6 Points The military is not elected by anyone. Therefore, its involvement in politics, even at a peripheral level, is a diminution of democratic accountability. However, it also has other significant implications.


76

The military might, for example, become involved in government because of an actual or created internal or external threat. Such a situation would imply the distortion of government policy in order to meet this threat, for example by increasing the defense budget at the expense of other budget allocations. In some countries, the threat of military take-over can force an elected government to change policy or cause its replacement by another government more amenable to the military’s wishes. A military takeover or threat of a takeover may also represent a high risk if it is an indication that the government is unable to function effectively and that the country therefore has an uneasy environment for foreign businesses. A full-scale military regime poses the greatest risk. In the short term a military regime may provide a new stability and thus reduce business risks. However, in the longer term the risk will almost certainly rise, partly because the system of governance will be become corrupt and partly because the continuation of such a government is likely to create an armed opposition. In some cases, military participation in government may be a symptom rather than a cause of underlying difficulties. Overall, lower risk ratings indicate a greater degree of military participation in politics and a higher level of political risk. Religious Tensions – 6 Points Religious tensions may stem from the domination of society and/or governance by a single religious group that seeks to replace civil law by religious law and to exclude other religions from the political and/or social process; the desire of a single religious group to dominate governance; the suppression of religious freedom; the desire of a religious group to express its own identity, separate from the country as a whole. The risk involved in these situations range from inexperienced people imposing inappropriate policies through civil dissent to civil war. Law and Order – 6 Points Law and Order are assessed separately, with each sub-component comprising zero to three points. The Law sub-component is an assessment of the strength and impartiality of the legal system, while the Order sub-component is an assessment of popular observance of the law. Thus, a country can enjoy a high rating – 3 – in terms of its judicial system, but a low rating – 1 – if it suffers from a very high crime rate of if the law is routinely ignored without effective sanction (for example, widespread illegal strikes). Ethnic Tensions – 6 Points This component is an assessment of the degree of tension within a country attributable to racial, nationality, or language divisions. Lower ratings are given to countries where racial and nationality tensions are high because opposing groups are intolerant and unwilling to compromise. Higher ratings are given to countries where tensions are minimal, even though such differences may still exist. Democratic Accountability – 6 Points This is a measure of how responsive government is to its people, on the basis that the less responsive it is, the more likely it is that the government will fall, peacefully in a democratic society, but possibly violently in a non-democratic one. The points in this component are awarded on the basis of the type of governance enjoyed by the country in question. For this purpose, we have defined the following types of governance: Alternating Democracy The essential features of an alternating democracy are: • A government/executive that has not served more than two successive terms. • Free and fair elections for the legislature and executive as determined by constitution or statute; • The active presence of more than one political party and a viable opposition;


77

• Evidence of checks and balances among the three elements of government: executive, legislative and judicial;

• Evidence of an independent judiciary; • Evidence of the protection of personal liberties through constitutional or other legal guarantees. Dominated Democracy The essential features of a dominated democracy are: • A government/executive that has served more than two successive terms. • Free and fair elections for the legislature and executive as determined by constitution or statute; • The active presence of more than one political party • Evidence of checks and balances between the executive, legislature, and judiciary; • Evidence of an independent judiciary; • Evidence of the protection of personal liberties. De-facto One-Party State The essential features of a de-facto one-party state are:

• A government/executive that has served more than two successive terms, or where the political/electoral system is designed or distorted to ensure the domination of governance by a particular government/executive.

• Holding of regular elections as determined by constitution or statute • Evidence of restrictions on the activity of non-government political parties (disproportionate media access between the governing and non-governing parties, harassment of the leaders and/or supporters of non-government political parties, the creation impediments and obstacles affecting only the non-government political parties, electoral fraud, etc).

De jure One-Party state The identifying feature of a one-party state is: • A constitutional requirement that there be only one governing party. • Lack of any legally recognized political opposition. Autarchy The identifying feature of an autarchy is:

• Leadership of the state by a group or single person, without being subject to any franchise, either through military might or inherited right.

In an autarchy, the leadership might indulge in some quasi-democratic processes. In its most developed form this allows competing political parties and regular elections, through popular franchise, to an assembly with restricted legislative powers (approaching the category of a de jure or de facto one party state). However, the defining feature is whether the leadership, i.e. the head of government, is subject to election in which political opponents are allowed to stand.

In general, the highest number of risk points (lowest risk) is assigned to Alternating Democracies, while the lowest number of risk points (highest risk) is assigned to autarchies. Bureaucracy Quality – 4 Points The institutional strength and quality of the bureaucracy is another shock absorber that tends to minimize revisions of policy when governments change. Therefore, high points are given to countries where the bureaucracy has the strength and expertise to govern without drastic changes in policy or interruptions in government services. In these low-risk countries, the bureaucracy tends to be somewhat autonomous from political pressure and to have an established mechanism for recruitment and training. Countries that lack the cushioning effect of a strong bureaucracy receive low points because a change in government tends to be traumatic in terms of policy formulation and day-to-day administrative functions.


78

Appendix 4 Countries in the ICRG political risk data set used in the PCA analysis

COUNTRIES COUNTRIES COUNTRIES COUNTRIES

ANGOLA ECUADOR KUWAIT PARAGUAYALBANIA EGYPT LEBANON QATAR

UAE SPAIN LIBERIA ROMANIAARGENTINA ESTONIA LIBYA RUSSIA

ARMENIA ETHIOPIA SRI LANKA SAUDI ARABIAAUSTRALIA FINLAND LITHUANIA SUDAN

AUSTRIA FRANCE LUXEMBOURG SENEGALAZERBAIJAN GABON LATVIA SINGAPORE

BELGIUM UNITED KINGDOM MOROCCO SIERRA LEONEBURKINA FASO GHANA MOLDOVA EL SALVADORBANGLADESH GUINEA MADAGASCAR SOMALIA

BULGARIA GAMBIA MEXICO SURINAMEBAHRAIN GUINEA-BISSAU MALI SLOVAK REPUBLICBAHAMAS GREECE MALTA SLOVENIABELARUS GUATEMALA MYANMAR SWEDENBOLIVIA GUYANA MONGOLIA SYRIABRAZIL HONG KONG MOZAMBIQUE TOGOBRUNEI HONDURAS MALAWI THAILAND

BOTSWANA CROATIA MALAYSIA TRINIDAD & TOBAGOCANADA HAITI NAMIBIA TUNISIA

SWITZERLAND HUNGARY NIGER TURKEYCHILE INDONESIA NIGERIA TANZANIACHINA INDIA NICARAGUA UGANDA

COTE D'IVOIRE IRELAND NETHERLANDS UKRAINECAMEROON IRAN NORWAY URUGUAY

CONGO IRAQ NEW ZEALAND UNITED STATESCOLOMBIA ICELAND OMAN VENEZUELA

COSTA RICA ISRAEL PAKISTAN VIETNAMCUBA ITALY PANAMA YEMEN

CYPRUS JAMAICA PERU SERBIA-MONTENEGROCZECH REPUBLIC JORDAN PHILIPPINES SOUTH AFRICA

GERMANY JAPAN PAPUA NEW GUINEA CONGO DRDENMARK KAZAKSTAN POLAND ZAMBIA

DOMINICAN REPUBLIC KENYA KOREA, DPR ZIMBABWEALGERIA SOUTH KOREA PORTUGAL

Source: ICRG data set


79

Appendix 5 Descriptive statistics for the principal component analyses Table 1: Data on 5 year average

Descriptive Statistics

499 1,33 12,00 6,9122 3,09254

530 ,17 12,00 6,2896 2,58329

532 ,33 12,00 7,3271 3,09033

531 ,33 12,00 7,9574 2,76606

532 2,80 12,00 9,7307 1,97036

533 1,00 11,48 7,5489 2,12132

533 ,75 12,00 8,7769 2,48843

533 ,83 11,96 6,8424 2,28528

533 1,83 12,00 7,3946 2,88815

521 ,25 12,07 7,6334 3,36294

531 1,84 12,00 9,2233 2,53768

533 ,60 11,00 5,5710 2,06685

484

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond

Valid N (listwise)

N Minimum Maximum Mean Std. Deviation

Table 2: Yearly data


2601 ,25 12,00 7,0397 3,13227

2822 ,17 12,33 6,3755 2,66786

2814 ,08 12,00 7,3612 3,16969

2837 ,17 12,00 7,9408 2,88294

2856 2,00 12,00 9,6267 2,23074

2871 ,67 12,00 7,4478 2,33427

2858 ,17 12,00 8,7025 2,71771

2866 ,08 12,00 6,8078 2,41359

2864 ,83 12,00 7,3107 3,04186

2751 ,17 12,33 7,6828 3,43629

2853 ,17 12,00 9,1861 2,64875

2869 ,08 11,00 5,6470 2,17262

2450

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond

Valid N (listwise)



80

Appendix 6 Results for the principal component analysis on political risk on yearly data Table1. Descriptive Statistics


2601 ,25 12,00 7,0397 3,13227

2822 ,17 12,33 6,3755 2,66786

2814 ,08 12,00 7,3612 3,16969

2837 ,17 12,00 7,9408 2,88294

2856 2,00 12,00 9,6267 2,23074

2871 ,67 12,00 7,4478 2,33427

2858 ,17 12,00 8,7025 2,71771

2866 ,08 12,00 6,8078 2,41359

2864 ,83 12,00 7,3107 3,04186

2751 ,17 12,33 7,6828 3,43629

2853 ,17 12,00 9,1861 2,64875

2869 ,08 11,00 5,6470 2,17262

2450

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond

Valid N (listwise)


Table 2. Overall MSA and Bartlett Test of Sphericity

KMO and Bartlett's Test

,857

15176,907

66

,000


Approx. Chi-Square

df

Sig.


Table 3. Partial Correlations and Individual MSA’s

Anti-image Matrices

,340 -,102 -,112 ,033 -,003 ,001 ,030 -,019 -,065 -,093 ,023 -,117

-,102 ,437 -,107 -,015 ,006 ,037 ,019 ,102 -,081 -,031 -,087 -,093-,112 -,107 ,508 -,003 -,075 ,041 ,006 -,102 -,015 -,063 -,037 ,122

,033 -,015 -,003 ,617 -,010 -,055 -,122 ,042 -,065 -,002 -,101 -,018

-,003 ,006 -,075 -,010 ,587 -,039 -,151 -,005 ,025 -,036 -,097 ,042,001 ,037 ,041 -,055 -,039 ,522 -,030 -,255 -,079 ,046 ,040 ,084

,030 ,019 ,006 -,122 -,151 -,030 ,321 ,003 -,126 -,082 -,063 -,031

-,019 ,102 -,102 ,042 -,005 -,255 ,003 ,383 ,025 -,055 -,036 -,186

-,065 -,081 -,015 -,065 ,025 -,079 -,126 ,025 ,336 -,033 ,017 -,040-,093 -,031 -,063 -,002 -,036 ,046 -,082 -,055 -,033 ,411 -,025 -,008

,023 -,087 -,037 -,101 -,097 ,040 -,063 -,036 ,017 -,025 ,705 ,029

-,117 -,093 ,122 -,018 ,042 ,084 -,031 -,186 -,040 -,008 ,029 ,448,881a -,264 -,270 ,072 -,006 ,003 ,090 -,052 -,191 -,249 ,047 -,299

-,264 ,862a -,227 -,029 ,012 ,077 ,051 ,248 -,212 -,073 -,157 -,210

-,270 -,227 ,867a -,006 -,137 ,079 ,015 -,231 -,037 -,137 -,061 ,255,072 -,029 -,006 ,908a -,016 -,097 -,274 ,087 -,143 -,003 -,153 -,035

-,006 ,012 -,137 -,016 ,894a -,071 -,349 -,011 ,056 -,074 -,151 ,083

,003 ,077 ,079 -,097 -,071 ,690a -,074 -,569 -,189 ,099 ,066 ,174,090 ,051 ,015 -,274 -,349 -,074 ,863a ,008 -,382 -,225 -,133 -,081

-,052 ,248 -,231 ,087 -,011 -,569 ,008 ,700a ,069 -,139 -,069 -,450

-,191 -,212 -,037 -,143 ,056 -,189 -,382 ,069 ,902a -,088 ,035 -,102-,249 -,073 -,137 -,003 -,074 ,099 -,225 -,139 -,088 ,937a -,047 -,018

,047 -,157 -,061 -,153 -,151 ,066 -,133 -,069 ,035 -,047 ,916a ,051

-,299 -,210 ,255 -,035 ,083 ,174 -,081 -,450 -,102 -,018 ,051 ,797a

bur_qua

prsdem_acc

eth_ten

ext_confgov_stab

int_conf

inv_proflaw_ord

milit

rel_tensoc_cond

bur_qua

prsdem_acc

eth_ten

ext_conf

gov_stabint_conf

inv_prof

law_ordmilit

rel_ten

soc_cond

Anti-image Covariance

Anti-image Correlation

bur_qua prs dem_acc eth_ten ext_conf gov_stab int_conf inv_prof law_ord milit rel_ten soc_cond

Measures of Sampling Adequacy(MSA)a.


81

Table 4. Extraction of components Total Variance Explained

5,450 45,419 45,419 5,450 45,419 45,419 3,428 28,567 28,567

1,402 11,687 57,105 1,402 11,687 57,105 2,797 23,310 51,877

1,272 10,601 67,706 1,272 10,601 67,706 1,900 15,829 67,706

,789 6,577 74,283

,644 5,363 79,646

,565 4,708 84,354

,446 3,721 88,075

,414 3,453 91,528

,322 2,680 94,208

,276 2,300 96,508

,215 1,788 98,296

,205 1,704 100,000

Component1

2

3

4

5

6

7

8

9

10

11

12

Total % of Variance Cumulative % Total % of Variance Cumulative % Total % of Variance Cumulative %

Initial Eigenvalues Extraction Sums of Squared Loadings Rotation Sums of Squared Loadings

Extraction Method: Principal Component Analysis. Table 5. Communalities

Communalities

1,000 ,804

1,000 ,751

1,000 ,520

1,000 ,572

1,000 ,580

1,000 ,793

1,000 ,761

1,000 ,822

1,000 ,690

1,000 ,669

1,000 ,510

1,000 ,651

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond

Initial Extraction

Extraction Method: Principal Component Analysis. Table 6. Rotated component solutions

Rotated Component Matrix a

,866

,801

,639

,735

,723

,856

,750

,841

,603 ,515

,674 ,420

,679

,701

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond

1 2 3

Component

Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization.

Rotation converged in 5 iterations.a.


82

Appendix 7 Validation of Principle Component analysis Group 1: Angola to Guinea

KMO and Bartlett's Test a

,829

1268,790

66

,000


Approx. Chi-Square

df

Sig.


Split_code = 1,00a.

Communalities a

1,000 ,820

1,000 ,780

1,000 ,639

1,000 ,592

1,000 ,569

1,000 ,837

1,000 ,738

1,000 ,834

1,000 ,710

1,000 ,744

1,000 ,505

1,000 ,779

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond

Initial Extraction


Split_code = 1,00a.

Total Variance Explained a

5,799 48,325 48,325 5,799 48,325 48,325 4,068 33,898 33,898

1,582 13,187 61,513 1,582 13,187 61,513 2,545 21,204 55,102

1,165 9,712 71,225 1,165 9,712 71,225 1,935 16,123 71,225

,775 6,458 77,684

,586 4,887 82,570

,489 4,072 86,643

,450 3,754 90,397

,355 2,959 93,356

,291 2,425 95,781

,222 1,851 97,631

,165 1,376 99,008

,119 ,992 100,000

Component1

2

3

4

5

6

7

8

9

10

11

12




Split_code = 1,00a.

Rotated Component Matrix a,b

,868

,814

,666 ,437

,767

,671

,896

,474 ,639

,422 ,810

,684 ,434

,736

,621

,786

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond

1 2 3

Component



Split_code = 1,00b.

Descriptive Statistics a

7,5955 3,13354 172

7,0042 2,54071 172

7,9238 2,92269 172

8,3311 2,30268 172

10,0955 1,70271 172

7,8641 1,87879 172

9,3356 2,11253 172

7,3665 2,19464 172

8,1745 2,71649 172

8,1298 3,28064 172

9,8364 2,22313 172

6,1232 1,97593 172

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond

Mean Std. Deviation Analysis N

Only cases for which Split_code = 1 are usedin the analysis phase.

a.


83

Group 2: Gambia to Niger


6,9212 2,85994 153

6,4101 2,32847 153

7,7791 2,85588 153

7,8606 2,88306 153

9,7434 2,09993 153

7,6496 2,00435 153

8,8022 2,47823 153

7,0501 2,05237 153

7,3250 2,83434 153

8,1307 3,15921 153

8,6778 2,80090 153

5,6114 1,97245 153

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond


Only cases for which Split_code = 2 are usedin the analysis phase.

a.


,818

1064,626

66

,000


Approx. Chi-Square

df

Sig.


Only cases for which Split_code = 2 are used in theanalysis phase.

a.


5,675 47,294 47,294 5,675 47,294 47,294 3,460 28,832 28,832

1,502 12,520 59,813 1,502 12,520 59,813 2,649 22,077 50,910

1,146 9,553 69,366 1,146 9,553 69,366 2,215 18,457 69,366

,767 6,391 75,758

,690 5,751 81,508

,605 5,039 86,547

,452 3,770 90,317

,353 2,940 93,257

,274 2,285 95,542

,228 1,897 97,439

,175 1,458 98,897

,132 1,103 100,000

Component1

2

3

4

5

6

7

8

9

10

11

12




Split_code = 2,00a.

Communalities a

1,000 ,822

1,000 ,736

1,000 ,588

1,000 ,648

1,000 ,571

1,000 ,788

1,000 ,818

1,000 ,781

1,000 ,712

1,000 ,671

1,000 ,610

1,000 ,580

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond

Initial Extraction


Split_code = 2,00a.


,872

,709

,645

,721

,638

,821

,701 ,432

,811

,651 ,444

,666 ,442

,770

,735

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond

1 2 3

Component



Split_code = 2,00b.


84

Group 3: Nigeria to Zimbabwe


6,4752 3,10401 159

6,2623 2,51868 159

7,3244 3,08200 159

8,4404 2,61559 159

9,9218 1,78615 159

7,5970 2,00312 159

9,1169 2,12851 159

6,8435 2,24271 159

7,6401 2,67321 159

7,5348 3,19578 159

9,4152 2,28440 159

5,6801 1,83684 159

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond


Split_code = 3,00a.


,842

1183,248

66

,000


Approx. Chi-Square

df

Sig.


Split_code = 3,00a.


5,838 48,653 48,653 5,838 48,653 48,653 3,218 26,815 26,815

1,521 12,675 61,329 1,521 12,675 61,329 2,697 22,475 49,291

1,139 9,489 70,818 1,139 9,489 70,818 2,583 21,527 70,818

,906 7,547 78,365

,597 4,978 83,343

,476 3,965 87,307

,436 3,632 90,939

,353 2,942 93,881

,212 1,763 95,644

,196 1,631 97,275

,174 1,446 98,721

,153 1,279 100,000

Component1

2

3

4

5

6

7

8

9

10

11

12




Split_code = 3,00a.

Communalities a

1,000 ,849

1,000 ,806

1,000 ,605

1,000 ,569

1,000 ,640

1,000 ,781

1,000 ,793

1,000 ,702

1,000 ,707

1,000 ,710

1,000 ,624

1,000 ,712

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond

Initial Extraction


Split_code = 3,00a.


,873

,821

,659 ,408

,490 ,564

,721

,870

,505 ,642

,773

,488 ,542 ,418

,610 ,476

,770

,657 ,528

bur_qua

prs

dem_acc

eth_ten

ext_conf

gov_stab

int_conf

inv_prof

law_ord

milit

rel_ten

soc_cond

1 2 3

Component



Split_code = 3,00b.


85

Appendix 8 Wooldridge test for auto correlation in panel data52

The first test was run with only the main variables whereas the next test included the period dummies. Both test are testing the hypothesis H0: No first order autocorrelation Ha: Autocorrelation As both test have very small p-values they are both rejected and thus confirms the suspicion of autocorrelation in the error terms.

52 The method is obtain from the STATA help online: http://www.stata.com/support/faqs/stat/panel.html


86

Appendix 9 Countries in the regression

United Arab Emirates ARE 1992 Middle East High Low2005

Argentina ARG 1995 Latin Amr. Upper middle Low2001

Australia AUS 1994 Australia+NZ High Very low2005

Azerbaijan AZE 2004 E.Europe+Russia Lower middle Moderate

Belgium BEL 1998 W.Europe High Very low

Bangladesh BGD 1997 Asia Low High2001

Bahrain BHR 1997 Middle East Upper middle Moderate2001 High

Brazil BRA 1997 Latin Amr. Upper middle Moderate2005 Lower middle

Canada CAN 1989 US+Canada High Very low2002

Chile CHL 1990 Latin Amr. Lower middle Low2000 Upper middle

China CHN 1991 Asia Low Moderate2005 Lower middle

Cameroon CMR 2001 Africa Low High

Colombia COL 1994 Latin Amr. Lower middle High2000

Costa Rica CRI 1998 Latin Amr. Lower middle Very low2000 Upper middle

Czech Republic CZE 1996 E.Europe+Russia Upper middle Very low2001

Germany DEU 1989 W.Europe High Very low1998

Denmark DNK 1998 W.Europe High Very low

Dominican Republic DOM 2000 Latin Amr. Lower middle Low

Ecuador ECU 1998 Latin Amr. Lower middle Moderate2002

Egypt EGY 1999 Middle East Lower middle Moderate2001

Spain ESP 1991 W.Europe High Low1999

Estonia EST 2001 E.Europe+Russia Upper middle Low

Finland FIN 1992 W.Europe High Very low2001

France FRA 1990 W.Europe High Very low2001

United Kingdom GBR 1990 W.Europe High Very low2001

Income classification

Risk (average)

Borrower CountryCountry code

Year Region


87

Ghana GHA 1997 Africa Low Moderate2000

Greece GRC 1997 W.Europe High Very low1999

Hong Kong HKG 1993 Asia High Very low1997

Croatia HRV 1997 E.Europe+Russia Upper middle Moderate2000

Hungary HUN 1995 E.Europe+Russia Upper middle Very low2001

Indonesia IDN 1993 Asia Lower middle High2005

India IND 1993 Asia Low High2002

Ireland IRL 1999 W.Europe High Very low

Iceland ISL 1998 W.Europe High Very low2001

Israel ISR 1998 Middle East High Moderate

Italy ITA 1993 W.Europe High Low2000

Jordan JOR 1996 Middle East Lower middle Low2000

Japan JPN 1999 Asia High Very low2003

Kenya KEN 2000 Africa Low High

Korea (South) KOR 1992 Asia Upper middle Low2002

Kuwait KWT 1994 Middle East High Low2001

Lithuania LTU 1998 E.Europe+Russia Lower middle Moderate1999

Luxembourg LUX 1998 W.Europe High Very low1999

Morocco MAR 1997 Middle East Lower middle Moderate

Mexico MEX 1990 Latin Amr. Upper middle Moderate2004

Mali MLI 2000 Africa Low Moderate

Malaysia MYS 1994 Asia Upper middle Low2005

Netherlands NLD 1994 W.Europe High Very low2000

Norway NOR 2000 W.Europe High Very low

New Zealand NZL 2002 Australia+NZ High Very low

Income classification

Risk (average)


Year Region

Note: The years given are the first and last in which project finance loans have been observed. The income classification is according to the World Bank and may have change over the course of time observed. This has been noted by writing two income levels.


88

Oman OMN 1996 Middle East Upper middle Low2002

Pakistan PAK 1996 Asia Low High1998

Panama PAN 1997 Latin Amr. Lower middle Low1999 Upper middle

Peru PER 1997 Latin Amr. Lower middle Moderate1999

Philippines PHL 1993 Asia Lower middle Moderate2004

Papua New Guinea PNG 1995 Asia Lower middle Moderate2004 Low

Poland POL 1996 E.Europe+Russia Upper middle Very low2000

Portugal PRT 1993 W.Europe Upper middle Low

Qatar QAT 1996 Middle East High Low2001

Romania ROM 2000 E.Europe+Russia Lower middle High

Russia RUS 1997 E.Europe+Russia Lower middle High2002

Saudi Arabia SAU 1995 Middle East Upper middle Moderate2000

Singapore SGP 1994 Asia High Very low2001

Slovakia SVK 1997 E.Europe+Russia Upper middle Low2001

Slovenia SVN 1999 E.Europe+Russia High Very low

Sweden SWE 1998 W.Europe High Very low2000

Thailand THA 1992 Asia Lower middle Moderate2005

Trinidad and Tobago TTO 1996 Latin Amr. Upper middle Moderate1997

Turkey TUR 1992 Middle East Lower middle High2001 Upper middle

Tanzania TZA 2000 Africa Low Moderate

USA USA 1987 US+Canada High Very low2003

Venezuela VEN 1993 Latin Amr. Upper middle Very low2001

Vietnam VNM 1993 Asia Low Moderate2002

South Africa ZAF 1999 Africa Upper middle Moderate2001 Lower middle

Year RegionIncome classification

Risk (average)


Note: The years given are the first and last in which project finance loans have been observed. The income classification is according to the World Bank and may have change over the course of time observed. This has been noted by writing two income levels.


89

Appendix 10 Distribution of the dependent variables

Log share of project finance loans in numbers

Log share of project finance loans in volume

Share of project finance loans in numbers

Share of project finance loans in volume


90

Appendix 11 Distribution of independent variables

Corporate Governance Log Corporate Governance

GDP Log GDP


91

Appendix 12 Result from regression run with different incomes groups as base. Share of project finance loans measured in numbers Upper middle income as base of the regression

Lower middle income group as base of the regression

Note: QI=Quality of Institutions, TC=Tension & Conflicts, PQ=Policy Quality. D refers to the variable being a dummy variable. QI_high is an interaction term between Quality of Institutions and high income dummy and so forth for the other interaction terms.


92

Low income group as base of the regression

Share of project finance loans measured in volume Upper middle income group as base of the regression


93

Lower middle income group as base of the regression

Low income group as base of the regression

Documents

Project Finance and Political Risk - AU Purepure.au.dk/portal/files/3002/Sara_Toft_THESIS_aug_2008.pdf · finance, in which more risk leads to more syndicated loans being given as