domestic investment, fdi and external debt: an …preliminary domestic investment,fdi and external...
TRANSCRIPT
Preliminary
Domestic Investment, FDI and External Debt: An Empirical Investigation
Manop Udomkerdmongkol, Holger Görg and Oliver Morrissey
School of Economics, The University of Nottingham Nottingham NG7 2RD UK
Abstract: The paper is to make predictions on the relative importance of three different sources of financing, namely domestic capital self-financing (private investment); FDI financing; external debt financing, for domestic investment under two types of political regime – politically unstable and stable regimes, based on a sample of low and middle-income countries over the period 1995-2001. Our hypotheses are that international borrowing financing would be the major source of finance in both regimes. Private investment would be least important source in unstable regime. Yet, in stable regime, it would be of similar importance to FDI financing. FDI financing would be between foreign debt financing and domestic capital self-financing in unstable regime. Findings suggest that external debt financing has no impact on domestic investment. By contrast, FDI and private investment crowd in the investment. In unstable regime, the effect of domestic capital self-financing is greater than FDI financing effect. Domestic capital self-financing, however, is of similar significance to FDI financing in stable regime.
JEL classification: E20, F23, F30
Keywords: Direct Investment, Domestic Investment, Foreign Debt, Emerging Markets
1
1. Introduction
Does FDI crowd in or crowd out domestic investment in a developing country? It is a
question for which economists try to provide the explanation. Crowding in means the
development and upgrading of domestic firms to benefit from linkages with foreign affiliates,
raises the efficiency of production, and contributes to the diffusion of knowledge and skills
from multinational enterprises to the local enterprise sector. It also includes new investment in
upstream or downstream production by other foreign or domestic producers or increases in
financial intermediation. By contrast, the form of crowding out is in terms of access to finance
and skilled labour, resulting in an uneven playing field for domestic firms. This can raise the
cost to local firms in terms of finance and skilled personnel (Kumar (2003)).
Many studies attempted to clarify the answer. However, their results seem to be
ambiguous and inconclusive. For example, Agosin and Mayer (2000) investigate the long-run
impact of FDI on domestic investment in three regions, namely Africa, Asia and Latin
America, using annual data covering 1970-96. They discover that there is strong crowding in
of domestic investment by FDI in Asia and Africa. Yet strong crowding out is the evidence in
Latin America. Using time-series estimation techniques, Kim and Seo (2003) give empirical
evidence on the dynamic relationship among inward FDI, economic growth and domestic
investment in South Korea for the period 1985-99. They suggest that FDI crowds in domestic
investment in the country.
Our main contribution is to make predictions on the relative importance of three different
sources of financing – domestic capital self-financing (private investment), FDI financing and
external debt financing – for domestic investment under two types of political regimes –
politically unstable and stable regimes. Other features of our study contribute some
improvements to the analysis of ‘FDI – crowding in or crowding out of domestic investment’
in several ways. We employ more recent annual aggregate data for 36 low and middle-income
countries covering the period 1995-2001. Secondly, not only do we evaluate the FDI effects on
domestic investment, but we also appraise the impacts of self-financing domestic capital and
2
external debt on the investment. Finally, Prais-Winsten model with panel-corrected standard
errors is utilised in our research.
Findings suggest that, in both regimes, foreign debt financing has no effect on domestic
investment. Domestic capital self-financing and FDI financing encourage the investment.
Nonetheless the impact of domestic capital self-financing is greater than FDI financing effect
in unstable regime. In stable regime, it turns out that private investment is of similar
importance to direct investment, which is in line with hypothesis.
The paper is structured as follows. Section 2 demonstrates the benchmark model and
outlines the hypotheses tested. Section 3 reviews the literature on previous studies on domestic
investment, FDI and external debt interactions. The subsequent section describes the data set
and the econometric framework, followed by a discussion of the findings. Finally, conclusions
are expressed in the last section.
2. The Benchmark Model
This study utilises the model of Dalmazzo and Marini (2000) to generate predictions on
the relative significance of three different sources of financing, say domestic capital self-
financing, FDI financing and foreign debt financing, for domestic investment under two types
of political regimes – politically unstable (i.e. a positive probability of a populist government)
and politically stable (i.e. zero probability of a populist government).
The authors consider a country with a worker, w, and a capitalist, c. To emphasize the
relationship between international trade and finance, they assume that the agents produce an
export good and consume a good imported from abroad. The agents’ utility function is
U(Ci) = Ci, i = (w,c) (1)
3
where Ci is the consumption of the imported good for agent i. The capitalist controls an
investment opportunity that costs K > 0, requires the labor services of the worker1 and
generates y units of the export good. When the investment decision is taken, there is a sunk
cost of an amount K, which could have been spent on a foreign consumption good. Each unit
of the export good can be traded for P units of the import good, so that aggregate consumption
is C = Py.
Social efficiency in capitalistic technology, i.e. C – K ≥ 0, is assumed. The details of the
distribution of surplus, the country’s political environment, the working of sanctions and the
timing of the model are summarized as follows.
Distribution: The outcome of the distribution process over C is modelled as the Nash-
solution to a bargaining game, with N ≥ 2 players: Si = C/N, where Si is agent i’s share of C.
Political Environment: The capitalist is subject to the risk that a populist government
(type w government) comes into office when the investment cost has been sunk. Once in
office, it will aim at maximizing the worker’s consumption level2, Cw. As a consequence, the
capitalist could be excluded from the bargaining process over the returns generated by his
investment. Assume that there is an exogenous probability (1 – ρ) that the populist government
will win the elections.
Sanctions: Foreign partner can impose sanctions whenever the country considered
violates some international agreement.
1 The worker always retains some bargaining power over the surplus that the project generates (Dalmazzo and Marini (2000). 2 Such an adverse attitude towards the capital share can take several forms: a hostile government can opt for outright expropriation of the capitalist’s assets, or it can impose rules, which strongly limit the capitalist’s right to manage his assets. The income from capital can also be heavily taxed, so as to redistribute surplus in favor of workers (Dalmazzo and Marini (2000)).
4
Timing: At time t = 0, the decision of whether or not to invest is taken. Once investment
is undertaken, the cost K is sunk. At time t = 1, political uncertainty is resolved with a type c
government (the government that safeguards the capitalist’s property rights on investment)
being in office with probability ρ. At time t = 2, production takes place and trade occurs if
actions leading to sanctions have not been taken (in the case of type w government).
1) The Domestic Capitalist’s Self-financing
Using own funds, the capitalist incurs the investment cost at t = 0. Depending on the
political outcome at t = 1, the capitalist will either remain in control of his assets with
probability ρ, or will be excluded from the division of the surplus C with probability
1 – ρ.
In the first case, the capitalist retains the power to deny the worker access to physical
capital. Because the agents need to find an agreement to produce at t = 2, there is a bilateral
bargaining over C generating equilibrium payoffs equal to Sw = Sc = ½ C. However, if he is
expropriated by an adverse government, his payoff is zero while the worker obtains the whole
surplus C.
Based on risk-neutrality and common knowledge of ρ, the agents’ expected shares at
t = 0 are, respectively
ESw = ρ(C/2) + (1 – ρ)C = [1 – (ρ/2)]C (2)
and
ESc = ρ(C/2) + (1 – ρ)0 = ρ(C/2). (3)
The capitalist’s expected consumption is
ECc = C – ESw – K = ρ(C/2) – K. (4)
5
Based on the social efficiency in capitalistic technology assumption, ECc is at least as
great as zero. Therefore, domestic investment is not greater than ρ(C/2) (Dalmazzo and Marini
(2000)).
2) Foreign Direct Investment Financing
They suppose that now the capitalist sells his project to a foreign company. Once the
foreign investor has sunk the investment cost, he faces the risk that a hostile government will
come into office. However, he can demand the application of trade sanctions against the
country. When sanctions can be imposed, the expected worker’s and foreign investor’s shares
are Sw = Sc = ½C.
If type c government comes into office, the agents will agree on a partition giving each of
them half of C. On the other hand, when type w government expropriates the foreign investor,
the application of sanctions enables him to block the country’s international trade. That leads
to a bargaining over C under the worker’s interest. After an agreement is reached, foreign
country lifts sanctions and trade occurs. In equilibrium, each agent obtains ½C. Thus, the
sanctions guarantee foreign investor half of the surplus. The capitalist’s expected consumption
is
ECc = C – Sw – K = (C/2) – K. (5)
Based on the assumption of social efficiency in the capitalistic technology, domestic
capital is not greater than C/2.
3) Foreign Debt Financing
According to Dalmazzo and Marini (2000), the capitalist may borrow abroad to finance
the project. The repudiation of the outstanding debt obligations makes the country’s export
liable to the application of sanctions. Foreign creditor gains bargaining power over the surplus
6
C, since the embargo will be lifted only when an agreement between foreign lender and
capitalist is reached.
Denote the amount borrowed abroad and the repayment prescribed by the debt contract
by X and D that are greater than zero. At t = 0, the investment is implemented and debt is
contracted with a foreign lender. At t = 1, political uncertainty is solved. The capitalist remains
in full control of his project (with probability ρ). Instead, if the type w government came into
office (with probability 1 – ρ), the capitalist would be excluded from the bargaining process
over C. At t = 2, production and trade are ready to take place. At this stage, the party having
the right to manage may decide to default on the foreign debt. When the capitalist retains
control, he may decide to repudiate to maximize his share, Sc. However, when the capitalist
loses control, the populist government may default to maximize the worker’s share, Sw. When
repudiation does not occur, the parties will bargain over C – D and the lender will be paid D
back. If repudiation occurs, the application of sanctions will enable the lender to participate in
the bargaining game over C. Once an agreement is eventually reached among the parties, trade
and consumption will take place.
After repudiation, the number of agents taking part in the negotiation relies on the
political outcome at t = 1. Since the worker is indispensable for producing the export good,
repudiation implies that there will be a two-party game over C between foreign lender and
domestic government whenever the type w government is in the office. However, under the
type c government, repudiation entails a three-party game among the capitalist, the worker and
the lender.
Case 1: suppose that a type c government prevailed at t = 1. At t = 2, the decision
whether to repudiate is taken by the capitalist controlling the project. If the contractual amount
D is paid back, the capitalist, the worker and the lender obtain, respectively,
½(C – D), ½(C – D) and D. However, if the capitalist chooses repudiation, the foreign lender
can apply sanctions: a three-party bargaining over C occurs, each player (the capitalist, the
worker and the lender) obtains C/3. Therefore, the capitalist will default on foreign debt when
D is greater than C/3.
7
Case 2: suppose that a type w government prevailed at t = 1. At t = 2, the decision
whether to repudiate is taken by the type w government. If the contractual amount D is paid
back, the capitalist, the worker and the lender obtain, respectively, 0, C – D and D. However,
in case of repudiation, sanctions will force the type w government to bargain with the foreign
lender. A two-party bargaining over C occurs, the worker and the lender obtain ½C. Thus,
repudiation occurs when C – D < ½C or D > ½C.
The worker’s expected share at t = 0 can be written as a function of D. The function that
is decreasing in D becomes
ESw(D) = ρ[(C – D)/2] + (1 – ρ)(C – D), if D ≤ C/3
= ρC/3 + (1 – ρ)(C – D), if C/3< D < C/2
= ρC/3 + (1 – ρ)C/2, if D > C/2. (6)
The capitalist’s expected consumption is
ECc = C – ESw – K. (7)
The capitalist invests in the country to receive the highest expected consumption
happening when ESw is the lowest value. Thus, he makes a contract requiring D > C/2 with the
external lender. Therefore,
ECc = C – ρC/3 – (1 – ρ)C/2 – K. (8)
Based on the assumption of social efficiency in capitalistic technology, domestic
investment is not greater than (1/2 + ρ/6)C.
In conclusion, the model shows that, in politically unstable regime, both foreign debt and
FDI financing yield domestic investment greater than that generated by domestic capital self-
8
financing. Moreover, due to the structure of bargaining game, foreign debt financing yields the
investment greater than that generated by FDI financing.
The Model without Populist Government
We assume that there is no populist government (ρ = 1) in the model, but other
assumptions hold.
1) The Domestic Capitalist’s Self-financing
At t = 0, the capitalist sinks the investment cost by using own funds and will certainly
remain in control of his assets at t = 1. The agents find an agreement to produce at t = 2. That
leads to a bargaining over C, which generates equilibrium payoffs equal to Sw = Sc = ½C. The
agents’ expected shares at t = 0 are, respectively
ESw = ESc = C/2. (9)
Thus, the capitalist’s expected consumption is
ECc = C – ESw – K = C/2 – K. (10)
Domestic investment is not greater than C/2, based on the assumption of social efficiency
in the capitalistic technology.
2) FDI Financing
Suppose that the entrepreneur sells his project to a foreign company. The foreign investor
can sink the investment cost without any risk that a hostile government will come into office.
The expected worker’s and foreign investor’s shares are Sw = Sc = ½C. The capitalist’s
expected consumption in case of FDI is
9
ECc = C – Sw – K = C/2 – K. (11)
Based on the assumption of social efficiency in the capitalistic technology, domestic
investment is not greater than C/2.
3) Foreign Debt Financing
Assume that the capitalist uses external borrowing strategy to finance a project. At
t = 0, the investment is implemented and debt is contracted with a foreign lender. He will
remain in full control of the project at t = 1. At t = 2, production and trade are ready to take
place. At this stage, he may default on the foreign debt to maximize his share, Sc.
The investor and the worker will bargain over C – D and the lender will be paid D back if
repudiation does not occur. However, if repudiation occurs, the application of sanctions will
enable the foreign lender to participate in the bargaining game over C. Once an agreement is
reached among the parties, trade and consumption will take place. Agents taking part in the
negotiation are the capitalist, the worker and the lender.
As a consequence, if the contractual amount D is paid back, the capitalist, the worker and
the lender obtain, respectively, ½(C – D), ½(C – D) and D. However, if the capitalist chooses
repudiation, the foreign creditor can apply sanctions: a three-party bargaining over C occurs,
each player obtains C/3. Thus, he will default on foreign debt when D is greater than C/3. The
worker’s expected share at t = 0 is
ESw(D) = (C – D)/2, if D ≤ C/3
= C/3, if D > C/3. (12)
The capitalist’s expected consumption is
ECc = C – ESw – K. (13)
10
To maximise the expected consumption, he makes a contract requiring D > C/3 with the
foreign lender (ESw is decreasing in D). Therefore, ECc = 2/3C – K. Domestic investment is
not greater than 2/3C, based on the assumption of social efficiency in capitalistic technology.
In politically stable regime, we discover that foreign debt financing provides domestic
investment greater than that generated by FDI and domestic capitalist’s self-financing. In
addition, owing to the bargaining game, FDI and domestic capitalist’s self-financing yield
same amount of domestic capital.
Table 1: A Summary of Theoretical Results
Domestic Investment (K)
Source of Financing Politically Unstable
Regime
Politically Stable
Regime
Domestic Capital Self-financing (DCS) K ≤ ρ(C/2) K ≤ C/2
FDI Financing (FDI) K ≤ C/2 K ≤ C/2
Foreign Debt Financing (FDF) K ≤ C/2 + ρ(C/6) K ≤ 2/3C
Conclusion FDF > FDI > DCS FDF > FDI = DCS
Table 1 summarises the predicted ‘ranking’ of sources of financing under each regime.
Findings allow us to identify a number of hypotheses to test.
Hypothesis 1: Foreign debt financing would be
(a) the major source of finance in both regimes.
11
Hypothesis 2: Domestic capital self-financing would be
(a) least important source in unstable regime.
(b) of similar importance to FDI financing in stable regime.
Hypothesis 3: FDI financing would be
(a) between foreign debt financing and domestic capital self-financing in unstable
regime.
(b) of similar importance to domestic capital self-financing in stable regime.
3. A Literature Review
Three groups of previous studies are reviewed, namely FDI and domestic investment
interactions (see Appendix 1 for summary), foreign debt and domestic investment linkages (see
Appendix 2 for summary), and FDI, foreign debt and domestic investment interactions.
FDI and Domestic Investment Interactions
Agosin and Mayer (2000) assess the extent to which FDI in developing countries crowds
in or crowds out domestic investment. They develop a theoretical model of total investment
including FDI and domestic investment variables. Specifically, in their model, total investment
is the sum of domestic investment, contemporaneous and lagged FDI. They keep the FDI
variables as exogenous variables since they argue that FDI relates to conditions in the world
economy and multinational enterprises’ strategies. Based on the concept of adaptive
expectations, domestic investment is a function of its lagged values, expected growth of host
country’s GDP estimated by its lagged values. These lead to the model of total investment that
12
is explained by FDI and its lagged values, domestic investment and its lagged values, and GDP
growth rate and its lagged values.
They test the theory with panel data for the period 1970-96 and then determine whether
or not, in the long run, FDI crowds in domestic investment by using Wald tests. They study
three regions (Africa, Asia and Latin America) and find that in Asia, but less so in Africa, there
has been strong crowding in of domestic investment by FDI; by contrast, strong crowding out
has been the norm in Latin America.
Using Agosin and Mayer’s (2000) model, Mišun and Tomšík (2002) estimate whether or
not, in Czech Republic, Hungary and Poland, FDI crowds out domestic investment over 1990-
2000 period. They discover evidence of a crowding-out effect in Poland, Hungary over 1990-
2000 and Czech Republic over 1993-2000.
Egger and Pfaffermayr (2005) set up a simple oligopoly model with segmented markets.
They restrict multinational enterprises to close down their foreign plants and to operate as pure
exporters, and find lower domestic investment once the firms become pure exporters. They use
a firm-level data set based on a survey of Austrian manufacturing firms over 1998-2001, and
employ several endogenous treatment estimators – matching estimators, the Heckman
approach and Wooldridge’s instrumental variable estimators – to account for the endogenous
self-selection of firms. This allows one to estimate the domestic investment to sales ratio of a
firm if it were forced to be a pure exporter. Their result suggests that multinational activity
does not harm domestic investment propensity. More particularly, the average treatment effect
of going multinational is importantly positive for investments in physical capital and research
and development (R&D).
Turning to time series approach, little work has been found to date. Kim and Seo (2003)
provide empirical evidence on the dynamic relationship between inward FDI, economic growth
and domestic investment in South Korea for the period 1985-99. By employing a vector
autoregression model and the innovations accounting techniques, they explore dynamic
interactions between inward FDI, domestic investment and output. They find that FDI has
13
some positive effects on economic growth, but its effects seem to be insignificant.
Nevertheless, economic growth is found to have statistically significant and highly persistent
effects on the future level of FDI. Moreover, their finding does not support the view that FDI
crowds out domestic investment in the country.
Foreign Debt and Domestic Investment Interactions
Borensztein (1990) finds evidence for the simple notion that increasing foreign debt
stock leads to domestic investment falling by analyzing the Philippines over 1970-90 period.
He runs domestic investment-GDP ratio on the relative price of investment goods, real
domestic interest rates, and foreign debt. He suggests that, ceteris paribus, a 1.3 billion US$
foreign debt reduction stimulated domestic investment by one percentage point in the
Philippines.
Cohen (1991) considers the relationship between domestic investment, foreign debt and
economic growth in 81 developing countries over 1980s. He runs investment-GDP ratio on
foreign debt-export ratio, population growth, inflation, exports-GDP ratio, per capita GDP, and
the share of the population in primary school. His empirical results show the debt to export
variable has negative coefficients but is not statistically significant. He concludes an increase
in foreign debt does not necessarily reduce domestic investment.
Warner (1992) fails to find support for the general view of international debt crisis – that
external debt problem can cause domestic investment to decline – analyzing 9 Latin American
countries over 1982-89 period. He suggests one way to test for foreign debt effects is to see
whether equations without any debt-related information can explain the investment declines in
the countries or not. More specifically, he estimates equations that incorporate the effects of
world variables – host country’ s terms of trade, the percentage change in an industrial
production index for developed countries, and real US interest rate – but do not incorporate
debt crisis effects. The results strongly support his claim.
14
FDI, Foreign Debt and Domestic Investment Interactions
Neumann (2003) puts forward a different argument for domestic investment and FDI
interaction by adding international debt. She argues that owing to information asymmetry,
domestic investors cannot costlessly and credibly reveal the level of first-period investment to
international creditors. Thus, they choose to incur self-monitoring costs to increase capital
flows. As an alternative to international borrowing, they may sell some ownership to foreign
investors. Once equity claims convey information, equity trade is preferred to foreign
borrowing. In addition, FDI, portfolio investment and foreign debt crowd in domestic
investment. However, domestic investment with FDI and portfolio equity financing is greater
than that if financed by international borrowing.
In sum, previous studies of ‘FDI and domestic investment interactions’ and ‘foreign debt
and domestic investment linkages’ express ambiguous results and conclusions. However,
without any political factors, Neumann (2003) argues that both FDI and foreign debt can
encourage domestic investment in a developing country, but FDI dominates foreign borrowing
as a tool to increase domestic investment in the country.
4. The Data and Econometric Framework
The Data
To test the hypotheses set out in Section 2, we employ net FDI inflows to capture the
impacts of FDI financing on domestic investment in low and middle-income countries. We
also calculate the change in private external debt to GDP as a proxy for foreign debt financing.
In addition, to test domestic capitalist’s self-financing hypotheses, we calculate self-financing
domestic capital.
15
Self-financing domestic capital = gross fixed capital formation – net FDI inflows –
public investment (14)
We control for other factors that could determine a decision of an entrepreneur to invest
in the country, as identified in the previous empirical literature. An issue examined in the
literature is the effect of labour costs on domestic investment in the country. One would expect
that, ceteris paribus, high labour costs would reduce local firms’ investment. Cohen (1991)
reports a negative relationship of labour costs and domestic investment in 81 less-developed
countries. Many scholars (Agosin and Mayor (2000), Mišun and Tomšík (2002)) confirm the
positive impact of market size – a demand side factor – on domestic investment. Agosin and
Mayer (2000), examining domestic investment in 32 developing countries, find that the market
size is a significant determinant of domestic investment, for example.
Based on the significance in all of the factors, we collect annual aggregate data
representing those variables for our estimation. The data covering 1995-2001 for 36 low or
middle-income countries are chosen from The World Bank list of low or middle-income
countries3 on basis of data availability (see Appendix 3). The countries consist of 13 Latin
American and Caribbean countries (Argentina, Bolivia, Chile, Costa Rica, El Salvador,
Guatemala, Jamaica, Mexico, Panama, Paraguay, Peru, Uruguay, and Venezuela), 8 countries
from Asia (Indonesia, India, Iran, Kyrgyz Republic, Pakistan, The Philippines, Sri Lanka, and
Thailand), 10 countries from Europe (Bulgaria, Croatia, Czech Republic, Estonia, Hungary,
Moldova, Poland, Romania, Slovak Republic, and Turkey) and 5 African countries (Cote
d’lvoire, Mauritius, Senegal, South Africa, and Tunisia).
Based on The World Bank’s governance indicators4 covering the countries for 1996,
1998, 2000, 2002 and 20045 (see Appendix 4 for details), as proxies of the political regime, we
specifically employ 3 governance measures – voice and accountability, political stability and
3 The organisation defines low and middle-income countries as developing countries with 1995 per capita incomes of less than $765 (low) and $9,385 (middle) (Neumann (2003)). 4 The data are available at http://www.worldbank.org/wbi/governance/govdata. 5 Due to this reason and data availability basis, we test the hypotheses by using the data from 1995-2001.
16
absence of violence, regulatory quality – to sort the countries into stable and unstable regimes
and test the hypotheses in our study. A high value refers to stable, low: unstable.
The proxies are reasonable in the sense that the media in high voice and accountability
country has independence to monitor own government and hold it accountable for its actions.
The government of high voice and accountability and/or regulatory quality country probably
never constitutes market-unfriendly policies and perception of the burdens imposed by
excessive regulation in FDI, foreign trade and business development. Moreover, the
government of politically stable country practicably continues business-friendly policies.
Spearman’s rank correlation is our technique used to test the direction and strength of the
relationship among three governance measures. We find that there is positive correlation
among the governance measures.
Table 2: Spearman’s Rank Correlation
Governance Indicator Spearman’s Rank Correlation Interpretation
voice and accountability
and political stability
0.76 strong positive
correlation
voice and accountability
and regulatory quality
0.68 strong positive
correlation
political stability
and regulatory quality
0.56 moderate positive
correlation
The countries classified by average percentile rank (1996-2002) of the governance
measures are as follows:
17
Voice and Accountability
High6: Argentina, Bolivia, Bulgaria, Chile, Costa Rica, Czech Republic, El Salvador, Estonia,
Hungary, India, Jamaica, Mauritius, Mexico, Panama, The Philippines, Poland, Romania,
Slovak Republic, South Africa, Thailand, Uruguay
Low: Cote d’lvoire, Croatia, Guatemala, Indonesia, Iran, Kyrgyz Republic, Moldova,
Pakistan, Paraguay, Peru, Senegal, Sri Lanka, Tunisia, Turkey, Venezuela
Political Stability
High: Argentina, Bulgaria, Chile, Costa Rica, Croatia, Czech Republic, El Salvador, Estonia,
Hungary, Jamaica, Kyrgyz Republic, Mauritius, Panama, Poland, Romania, Slovak Republic,
Thailand, Tunisia, Uruguay
Low: Bolivia, Cote d’lvoire, Guatemala, India, Indonesia, Iran, Mexico, Moldova, Pakistan,
Paraguay, Peru, The Philippines, Senegal, South Africa, Sri Lanka, Venezuela, Turkey
Regulatory Quality
High: Argentina, Bolivia, Chile, Costa Rica, Croatia, Czech Republic, El Salvador, Estonia,
Guatemala, Hungary, Jamaica, Mauritius, Mexico, Panama, Peru, The Philippines, Poland,
Slovak Republic, South Africa, Sri Lanka, Thailand, Tunisia, Turkey, Uruguay
Low: Bulgaria, Cote d’lvoire, India, Indonesia, Kyrgyz Republic, Moldova, Pakistan,
Paraguay, Romania, Senegal, Venezuela, Iran
6 High is of countries having the percentile rank that is greater than 50, Low otherwise.
18
Following are the data utilised as dependent and independent variables in the estimation.
Gross capital formation as a share of GDP (constant 2000, US dollar) is collected
from World Development Indicators 2004. It is treated as the dependent variable (see
Borensztein (1990), Cohen (1991), Warner (1992). Agosin and Mayer (2000), Mišun
and Tomšík (2002)).
Based on equation 14, self-financing domestic capital (current, US dollar) is
calculated from gross fixed capital formation (current, US dollar), net inflows of FDI
(current, US dollar), and public investment (current, local currency unit) divided by
exchange rate (local currency unit against US dollar). Then it is adjusted by GDP
(current, US dollar) to get self-financing domestic capital – GDP ratio (constant, US
dollar).
Data of gross fixed capital formation, exchange rate, and net inflows of FDI are from
World Development Indicators 2004. Public investment data are from Government
Finance Statistics, International Monetary Fund.
Net inflows of FDI (to the countries) as a share of GDP (constant 2000, US dollar) is
from World Development Indicators 2004 (see Agosin and Mayor (2000), Mišun and
Tomšík (2002)).
Percentage change in private non-guaranteed long term external debt to GDP
(constant 2000, US dollar) is calculated from change in private non-guaranteed long
term external debt (current, US dollar) adjusted by CPI (2000 = 100) and GDP
(constant 2000, US dollar). The data are from World Development Indicators (2004).
19
GDP per capita (constant 2000, US dollar) is treated as labour cost factor7 (see Cohen
(2001)). It is from World Development Indicators (2004).
Data of GDP growth (constant 2000, US dollar) treated as demand side factor are
from World Development Indicators (2004) (see Agosin and Mayor (2000), Mišun
and Tomšík (2002)).
Appendix 5 provides descriptive statistics and the correlation of those variables.
To clarify the difference between high and low countries, as shown in Table 3, two-
sample t-test with different population variances is used. The tests are of null hypotheses of no
difference between means of self-financing domestic capital, FDI inflows, and change in
private foreign debt to GDP in the two sub-samples, against the alternative that, the means are
different. The results in Table 3 suggest that on the average, self-financing domestic capital,
FDI inflows, and change in private foreign debt to GDP are different in two subgroups.
7 We measure labour costs as GDP per capita. While this is admittedly a rough proxy it may not be too problematic in our cross-country context, where differences in labour costs across country can be expected to be highly correlated with differences in GDP per capita.
20
Table 3: Testing Hypotheses about Difference between Means when Variances are Different.
Self-financing
Domestic Capital/ GDP
FDI Inflows/ GDP Private External
Debt/ GDP Governance
Indicator Χ S.D. t Χ S.D. t Χ S.D. t
high voice and
accountability
15.72 5.44 3.93 3.04 5.45* -0.52 7.49 2.34*
low voice and
accountability
14.09 4.66
2.54*
2.15 2.13 -2.55 6.22
high
political stability
16.01 5.26 3.93* 3.49 3.01 2.20* -1.97 7.82 -1.98*
low
political stability
13.51 4.69 2.72 2.48 -0.30 0.51
high
regulatory quality
15.74 5.17 3.35* 3.71 2.94 5.15* -0.39 7.79 2.00*
low
regulatory quality
13.46 4.86 2.01 2.13 -1.99 4.69
Remark: * means significant at 5 percent level (two-tailed test).
Econometric Framework
We firstly test the hypotheses by employing fixed and random effects estimation
techniques to allow for country specific factors. The Hausman test is used to justify, which
technique is more appropriate for the data. We check for heteroscedasticity problem by
Koenker-Bassett test and first-order autocorrelation problem by LM test, as it is possible that
21
the errors are not homoscedastic across panels and show first-order autocorrelation over time
within panel because of differences in domestic investment pattern among the countries and
funds projection of domestic investment projects in first few years in the country.
Fixed and random effects estimations however assume error terms are independent and
identically distributed (iid). If the errors are not ‘iid’, the estimators are still consistent but
inefficient, which could lead us to be either too or insufficiently confident in our estimators
(Beck and Katz (1995)). Prais-Winsten model with panel-corrected standard errors estimation8
(see Appendix 6 for details) is employed in this study to remedy the problems.
With the above mentioned considerations and modifications, the econometric model to
estimate is specified as follows:
lnGCFi,t = iα + 1β lnSDCi,t-1 + 2β lnFDIi,t-1 + 3β lnPFDi,t-1+ 4β lnPGDPi,t-1 + 5β lnGGDPi,t-1 +
6β DUM + 7β DUM*lnSDCi,t-1 + 8β DUM*lnFDIi,t-1 + 9β DUM*lnPFDi,t-1+ ti,ε
where GCFi,t is the gross capital formation, SDCi,t-1 is the self-financing domestic capital,
FDIi,t-1 is the FDI inflows, PFDi,t-1 is the private foreign debt, PGDPi,t-1 is the per capita GDP,
GGDPi,t-1 is the GDP growth, DUM is a dummy variable for high governance measure
countries (e.g. high voice and accountability countries). It is equal to 1 if the country relates to
high governance measure country, 0 otherwise.
The independent variables enter the logarithmic regression with a one-period lag based
on adjustment process of domestic investment. We expect positive coefficients on self-
financing domestic capital, FDI inflows, private external debt and GDP growth. Per capita
GDP is expected to be negative. As variables are in lags and coefficients are elasticities,
relative magnitudes of estimated coefficients can be interpreted as ‘rank importance’ for the
hypotheses.
8 STATA provides Prais-Winsten model with panel-corrected standard errors estimation.
22
5. Estimation Results
Based on different methodologies, Table 4 and 5 present estimation results regarding the
effects of the independent variables on the domestic investment in low and middle-income
countries from over 1996-2001 period.
On the first column in Table 4, the estimation results excluding any political factors are
based on fixed-effects estimation. We obtain evidence of positive effects of domestic capital
self-financing and FDI financing on domestic investment. The estimated coefficients of the
variables are statistically significant at 10 percent level. Higher GDP growth stimulates the
investment. The estimated coefficient is statistically significant at 1 percent level. The other
regressors are not statistically significant.
Using Koenker-Bassett test, the calculated test statistic is greater than the critical value of
1% significance value of the t distribution with 200 degrees of freedom. The null hypothesis of
homoscedastic errors is rejected. The null hypothesis of no first-order autocorrelation is
rejected since the LM test statistic is greater than the critical value of 5% critical value of the
chi-squared distribution with 1 degree of freedom. Although the regression generally performs
well with high coefficients of determinations (0.87), the tests demonstrate heteroscedasticity
and first-order autocorrelation problems9.
To remedy the problems, Prais-Winsten estimation is employed (216 observations). As
expected, we obtain evidence (appeared on the first column in Table 5) of highly significant
positive impacts of domestic capital self-financing, FDI financing and market size on the
investment. An increase in the self-financing domestic capital by 1 percentage point raises the
investment by 0.15 percent, while each extra percentage of FDI financing encourages the
investment by 0.06 percent. In addition, an increase in the GDP growth by 1 percentage point
9 Using random effects estimation (see Appendix 7), we receive similar results in general. However, the Hausman test statistic of 178.39 is greater than the critical value of the chi-squared distribution with 5 degrees of freedom at 1 percent level, showing a high importance of country-specific factors and their correlation with the explanatory variables. Thus fixed effects estimation is more appropriate than random effects estimation for the data.
23
adds the investment by 0.06 percent. The other explanatory variables are statistically
insignificant.
To clarify the importance of political factor, we add the political dummy variable (DUM)
in the regression. Based on Prais-Winsten estimation10 (see column 2-4 in Table 5), the results
suggest that the political factors – voice and accountability, political stability and regulatory
quality – do not affect domestic investment. However, the significance of domestic capital self-
financing, FDI inflows and market size in stimulating domestic investment is confirmed11.
Therefore, the assumption that the slope of the regression line is the same for each category of
the qualitative variable is not plausible. We then test the significance of the political factor
again using intercept and interactive dummy variables.
Shown in column 5-7 in Table 5, the estimation results are based on Prais-Winsten
estimation including an intercept dummy variable for political factor – voice and
accountability, political stability and regulatory quality – and slope dummy variables. The
results confirm the importance of market size determinant. An increase in the GDP growth by
1 percentage point adds domestic investment by 0.06 percent. Employing voice and
accountability dummy variable, an estimate of the extent of extra domestic investment in high
voice and accountability countries compared to low voice and accountability countries12 is
2.60. In low voice and accountability countries, an increase in self-financing domestic capital
by 1 percentage point raises the investment by 0.17 percent, whilst each extra percentage of
FDI financing encourages the investment by 0.06 percent. Thus, the impact of domestic capital
self-financing is greater than FDI financing effect in such countries (reject hypothesis 2(a) and
3(a)).13
10 Using within-groups fixed effects estimation, the dummy variable is dropped. It leads us to get the same results (see column 2-4 in Table 4). 11 Employing random effects estimation (see Appendix 7), we do not find the importance of the dummy variables. The Hausman tests moreover suggest that fixed effects estimation is more appropriate than random effects estimation. 12 ∂ FDI / ∂ DUM = 6β + 7β lnSDC + 8β lnFDI + 9β lnPFD 13 The Chi-squared test rejects the null hypothesis of no difference in effects of domestic capital self-financing and FDI financing on the investment in low voice and accountability countries.
24
In high voice and accountability countries, an increase in self-financing domestic capital
by 1 percentage point increases the investment by 0.34 (0.17+0.17) percent, while each extra
percentage of FDI inflows stimulates the investment by 0.32 (0.06+0.26) percent. Utilising the
Chi-squared test, equal effects of domestic capital self-financing and FDI financing on the
investment in the countries are found. The other independent variables are statistically
insignificant. Finding supports for hypothesis 2(b) and 3(b) – domestic capital self-financing is
of similar significance to FDI financing in stable regime.
Furthermore, we test the matter of political factor with different governance measures –
political stability and regulatory quality. The estimation results are consistent with the previous
results in general.
25
Table 4: Effects of domestic capital self-financing, FDI financing and private foreign debt financing on domestic investment in the
countries: 1996-2001
Dependent Variable: Gross capital formation as a share of GDP (logarithm, constant 2000)
FIXED EFFECTS ESTIMATION INDEPENDENT VARIABLES
(1) (2) (3) (4) (5) (6) (7)
Lagged self-financing domestic capital/ GDP (logarithm, b1) 0.06(0.08) 0.06(0.08) 0.06(0.08) 0.06(0.08) 0.09(0.13) 0.03(0.67) 0.01(0.48)
Lagged FDI inflows as a share of GDP (logarithm, b2) 0.03(0.09) 0.03(0.09) 0.03(0.09) 0.03(0.09) 0.03(0.16) 0.03(0.21) 0.02(0.45)
Lagged change in private external debt to GDP
(logarithm, constant 2000; b3)
-0.01(0.65) -0.01(0.65) -0.01(0.65) -0.01(0.65) -0.02(0.80) -0.01(0.97) 0.29(0.25)
Lagged per capita GDP (logarithm, constant 2000; b4) 0.09(0.42) 0.09(0.42) 0.09(0.42) 0.09(0.42) 0.09(0.42) 0.07(0.56) -0.01(0.98)
Lagged GDP growth (logarithm, constant 2000; b5) 0.10(0.00) 0.10(0.00) 0.10(0.00) 0.10(0.00) 0.10(0.00) 0.10(0.00) 0.09(0.00)
Constant - - - - - - -
Voice and accountability dummy variable (VA; b6) - -
Political stability dummy variable (PS; b7) - -
Regulatory quality dummy variable (RQ; b8) - -
VA * Lagged self-financing domestic capital/ GDP (b9) -0.03(0.73)
VA * Lagged FDI inflows as a share of GDP (b10) 0.03(0.70)
VA * Lagged change in private external debt to GDP (b11) 0.01(0.87)
PS * Lagged self-financing domestic capital/ GDP (b12) 0.09(0.32)
PS * Lagged FDI inflows as a share of GDP (b13) 0.07(0.26)
PS * Lagged change in private external debt to GDP (b14) -0.01(0.94)
26
RQ * Lagged self-financing domestic capital/ GDP (b15) 0.10(0.02)
RQ * Lagged FDI inflows as a share of GDP (b16) 0.08(0.12)
RQ * Lagged change in private external debt to GDP (b17) -0.30(0.23)
Coefficients of determinations 0.87 0.87 0.87 0.87 0.87 0.86 0.93
Hausman test statistic (Chi-squared) 178.39(0.00) 408.44(0.00) 160.1(0.00) 306.09(0.00) 22.72(0.01) 26.58(0.01) 134.79(0.00)
LM test statistic 124.672 124.672 124.672 124.672 127.352 151.792 161.792
Koenker-Bassett test (t-test) statistic -0.02(0.00) -0.02(0.00) -0.02(0.00) -0.02(0.00) -0.02(0.00) -0.02(0.00) -0.03(0.03)
Number of observations 216 216 216 216 216 216 216
Remarks: 1. The figures are the coefficient estimates, and the quantities in parentheses are the P-values.
2. The 5% critical value of Chi-squared distribution with 1 degree of freedom is 3.84.
27
Table 5: Effects of domestic capital self-financing, FDI financing and private foreign debt financing on domestic investment in the
countries: 1996-2001
Dependent Variable: Gross capital formation as a share of GDP (logarithm, constant 2000)
PRAIS-WINSTEN ESTIMATION INDEPENDENT VARIABLES
(1) (2) (3) (4) (5) (6) (7)
Lagged self-financing domestic capital/ GDP (logarithm, b1) 0.15(0.01) 0.18(0.00) 0.15(0.01) 0.17(0.01) 0.17(0.02) 0.15(0.04) 0.10(0.04)
Lagged FDI inflows as a share of GDP (logarithm, b2) 0.06(0.04) 0.05(0.10) 0.05(0.08) 0.04(0.10) 0.06(0.04) 0.06(0.04) 0.03(0.08)
Lagged change in private external debt to GDP
(logarithm, constant 2000; b3)
-0.02(0.62) -0.02(0.66) -0.02(0.64) -0.02(0.61) 0.01(0.88) -0.01(0.93) 0.07(0.82)
Lagged per capita GDP (logarithm, constant 2000; b4) 0.03(0.13) 0.04(0.07) 0.02(0.45) 0.06(0.06) 0.03(0.22) 0.01(0.80) 0.04(0.20)
Lagged GDP growth (logarithm, constant 2000; b5) 0.06(0.01) 0.06(0.01) 0.06(0.01) 0.06(0.01) 0.06(0.01) 0.06(0.01) 0.06(0.01)
Constant 2.24(0.00) 2.12(0.00) 2.32(0.00) 2.01(0.00) 2.17(0.00) 2.44(0.00) 1.42(0.09)
Voice and accountability dummy variable (VA; b6) -0.04(0.34) -0.80(0.10)
Political stability dummy variable (PS; b7) -0.05(0.28) -0.84(0.09)
Regulatory quality dummy variable (RQ; b8) -0.08(0.23) -0.84(0.09)
VA * Lagged self-financing domestic capital/ GDP (b9) 0.17(0.10)
VA * Lagged FDI inflows as a share of GDP (b10) 0.26(0.00)
VA * Lagged change in private external debt to GDP (b11) -0.03(0.71)
PS * Lagged self-financing domestic capital/ GDP (b12) 0.17(0.08)
PS * Lagged FDI inflows as a share of GDP (b13) 0.27(0.00)
28
PS * Lagged change in private external debt to GDP (b14) -0.01(0.90)
RQ * Lagged self-financing domestic capital/ GDP (b15) 0.22(0.00)
RQ * Lagged FDI inflows as a share of GDP (b16) 0.30(0.00)
RQ * Lagged change in private external debt to GDP (b17) -0.09(0.77)
Coefficients of determinations 0.88 0.86 0.88 0.87 0.86 0.87 0.89
Number of observations 216 216 216 216 216 216 216
Chi-squared test: H0: b1= b2 5.01(0.03) 4.40(0.04) 14.89(0.01)
Chi-squared test: H0: b1+b9 = b2+b10 0.84(0.36)
Chi-squared test: H0: b1+b12 = b2+b13 0.40(0.53)
Chi-squared test: H0: b1+b15 = b2+b16 0.92(0.34)
Remark: The figures are the coefficient estimates, and the quantities in parentheses are the P-values.
29
6. Concluding Remarks
Using The World Bank’s governance indicators and annual aggregate data over the
period of 1995-2002 from 36 low or middle-income countries, our empirical investigation
based on a model of Dalmazzo and Marini (2000) provides evidence of relative significance of
three different sources of financing, namely domestic capital self-financing, FDI and foreign
debt financings, for domestic investment under politically unstable and stable regimes. Our
hypotheses are that external debt financing would be the major source of finance in both
regimes. Domestic capital self-financing would be least important source in unstable regime.
Yet, in stable regime, it would be of similar importance to FDI financing. FDI financing would
be between external debt financing and domestic capital self-financing in unstable regime.
Findings can be summarised as:
1. There is no evidence for a relationship of external debt financing and domestic
investment in both regimes. The results suggest that foreign debt financing has no
effect on the investment.
2. In unstable regime, domestic capitalist’s self-financing and FDI financing crowd in
domestic investment. However, the impact of domestic capital self-financing is
greater than FDI financing effect in such regime.
3. In stable regime, domestic capital self-financing and FDI financing encourage
domestic investment. Yet domestic capital self-financing is of similar importance to
FDI financing in such regime.
4. Domestic investment is at a higher level in stable regime. As a result a government’s
ability to provide a business-friendly environment for investors will create greater
amounts of domestic investment.
30
The results suggest some questions to be addressed for further research. This includes the
use of longer data and new governance measures to re-analyze the findings. An extended work
at the lower level of aggregation – industry and firm level analyses – may be challenged.
Another improvement of the investigation could consider the possibility in exploring the
relative importance of sources of financing for domestic investment in a more disaggregated
nature with time-series estimation. As suggested that, the impact may differ according to
individual country characteristics.
References
Agosin, R.M. and Mayer, R. (2000). ‘Foreign Investment in Developing Countries: ‘Does It Crowd in Domestic Investment?’, Discussion Paper No. 146, United Nations Conference on Trade and Development, February 2000.
Baltagi, B.H. and Li, Q. (1991). ‘A Transformation that will circumvent the problem of autocorrelation in an error-component model’, Journal of Econometrics, Vol. 48, pp. 385-393.
Beck, N. and Katz, J.N. (1995). ‘What to do (and not to do) with time-series cross-section data’, American Political Science Review, Vol. 89, pp. 634-637.
Borensztein, E. (1990). ‘Debt Overhang, Debt Reduction and Investment: The Case of the Philippines,’ IMF Working Paper No. 77, International Monetary Fund, September 1990.
Cohen, D. (1991). ‘Slow Growth and Large LDC Debt in the Eighties: An Empirical Analysis,’ CEPR Discussion Paper No. 461, The Center for Economic Policy Research, January 1991.
Dalmazzo, A. and Marini, G. (2000). ‘Foreign Debt, Sanctions and Investment: A Model with Politically Unstable Less Developed Countries’, International Journal of Finance and Economics, Vol. 5, pp. 141-53.
Egger, P. and Pfaffermayr, M. (2005). ‘Multinationals Have Higher Domestic Investments Than National Firms,’ Paper Presented at the Globalisation and Firm Level Adjustment Conference, the University of Nottingham, 24-25 June 2005.
Kaufmann, D., Kraay, A. and Mastruzzi, M. (2005). ‘Governance Matters IV: Governance Indicators for 1996-2004’, Report of the Governance and Anti-Corruption Study Group, The World Bank, May 2005.
Kim, D.D. and Seo, J. (2003). ‘Does FDI Inflow Crowd out Domestic Investment in Korea?’, Journal of Economic Studies, Vol. 30(6), pp. 605-22.
31
Kumar, R. (2003). ‘Changing Role of the Public Sector in the Promotion of Foreign Direct Investment’, Asia-Pacific Development Journal, Vol. 10(2), pp. 1-27.
International Monetary Fund (2005). Government Finance Statistics, available from http://www.esds.ac.uk.
Mišun, J. and Tomšík, V. (2002). ‘Does Foreign Direct Investment Crowd in or Crowd out Domestic Investment’, Eastern European Economics, Vol. 40(2), pp. 38-56.
Neumann, R.M. (2003). ‘International Capital Flows under Asymmetric Information and Costly Monitoring: Implications of Debt and Equity Financing’, Canadian Journal of Economics, Vol. 36(3), pp. 674-700.
Warner, A.M. (1992). ‘Did the Debt Crisis Cause the Investment Crisis’, The Quarterly Journal of Economics, Vol. 107(4), pp. 1161-86.
World Bank (2004). World Development Indicators 2004, available from http://devdata.worldbank.org/dataonline/.
32
APPENDICES
Appendix 1: A Summary of Previous Studies on FDI and Domestic Investment Interactions
Agosin and Mayer
(2000)
Mišun and
Tomšík (2002)
Egger and
Pfaffermayr
(2005)
Study Period 1970-96 1990-2000 1998-2001
Data Type annual annual annual
Country/Industry 32 countries; Africa,
Asia and Latin
America
Czech Republic,
Hungary and
Poland
Austrian
firm-level data
Econometric Method SUR OLS Probit
Dependent Variable total investment-GDP ratio (US$) √ √
domestic investment-sales ratio √
Independent FDI-GDP ratio (US$) positive negative -
Variables 1st lagged FDI-GDP ratio positive positive -
2nd lagged FDI-GDP ratio positive positive -
1st lagged domestic investment-GDP ratio
(US$)
positive positive -
2nd lagged domestic investment-GDP ratio positive positive -
3rd lagged domestic investment-GDP ratio - positive -
1st lagged GDP growth (US$) positive positive -
33
2nd lagged GDP growth positive positive -
3rd lagged GDP growth - negative -
foreign affiliate dummy - - positive
medium cash-flow to sales ratio - - positive
high cash-flow to sales ratio - - positive
34
Appendix 2: A Summary of Previous Studies on Foreign Debt and Domestic Investment Interactions
Borensztein (1990) Cohen (1991) Warner (1992)
Study Period 1970-90 1980-90 1982-89
Data Type annual annual annual
Country/Industry The Philippines 81 developing
countries
9 Latin
American
countries
Econometric Method OLS Fixed-effects Fixed-effects
Dependent Variable domestic investment-GDP ratio (US$)
Independent the relative price of investment goods negative - -
Variables real domestic interest rates negative - -
foreign debt (US$) negative - -
foreign debt-GDP ratio (US$) - negative
(insignificant)
-
population growth - positive -
inflation (% changes in PPI) - negative -
exports-GDP ratio (US$) - positive -
per capita income - negative -
share of population in primary school - positive -
host country’ s terms of trade - - positive
35
percentage change in an industrial
production index for developed countries
- - positive
1st lagged domestic investment-GDP ratio
(US$)
- - positive
real 10-year US T-bills rate - - negative
36
Appendix 3: Raw Data
Country Year GCF DCS FDI PFD GGDP PGDP
Argentina 1995 17.94 14.80 2.17 0.99 -2.85 7195.64
1996 18.08 14.42 2.55 1.12 5.53 7503.47
1997 19.37 15.25 3.13 1.47 8.11 8018.94
1998 19.93 16.45 2.44 1.40 3.85 8234.93
1999 18.01 8.61 8.46 -0.08 -3.39 7870.20
2000 16.19 11.61 3.67 -0.41 -0.79 7726.32
2001 14.18 12.56 0.81 2.23 -4.41 7311.67
Bolivia 1995 15.24 8.56 5.85 0.56 4.68 947.75
1996 16.24 8.30 6.41 0.26 4.36 967.09
1997 19.63 7.89 9.22 1.89 4.95 993.09
1998 23.61 10.45 11.17 2.41 5.03 1021.20
1999 18.77 5.23 12.20 -0.17 0.43 1004.75
2000 18.30 7.06 8.77 4.21 2.51 1009.72
2001 14.04 3.80 8.67 -0.73 1.68 1007.05
Bulgaria 1995 15.65 13.99 0.69 -37.76 2.86 1566.96
1996 8.12 12.08 1.10 -44.99 -9.40 1427.15
1997 9.88 5.22 4.87 -52.01 -5.60 1354.35
1998 16.88 7.26 4.22 0.49 3.86 1416.04
1999 17.91 6.71 6.32 1.13 2.30 1457.25
2000 18.29 4.82 7.94 0.05 5.40 1564.15
2001 20.67 9.28 5.98 -0.31 4.10 1659.15
Chile 1995 25.79 16.22 4.53 3.54 10.63 4295.10
1996 26.88 14.31 7.02 5.96 7.41 4546.84
1997 27.23 14.93 7.00 1.07 7.39 4814.95
1998 27.43 15.95 6.33 8.11 3.92 4936.62
1999 21.34 5.44 11.99 0.71 -1.14 4816.44
2000 22.49 11.40 6.44 0.08 4.40 4964.40
2001 20.69 11.64 6.32 1.21 2.80 5040.22
Costa Rica 1995 18.24 15.07 2.87 -0.97 3.92 3654.90
1996 15.96 12.45 3.61 -0.72 0.89 3610.97
1997 18.08 13.64 3.18 -0.48 5.58 3733.30
1998 20.54 15.09 4.35 0.38 8.40 3962.67
1999 17.13 13.26 3.92 -0.12 8.22 4199.10
2000 17.01 14.43 2.56 -0.22 1.80 4185.39
2001 20.38 15.09 2.77 -0.26 1.08 4161.71
Source: World Development Indicators 2004 and Government Finance Statistics, International Monetary Fund Note: For the definitions, please see the Data and Econometric Framework section in the main text.
37
Appendix 3: Raw Data (continued)
Country Year GCF DCS FDI PFD GGDP PGDP
Cote d'lvoire 1995 15.60 6.66 1.92 -4.18 7.13 644.68
1996 12.11 7.70 2.22 -10.23 7.73 674.39
1997 14.42 5.53 3.54 1.56 5.72 693.39
1998 13.34 6.97 2.97 -3.24 4.75 707.52
1999 13.12 7.27 2.58 -2.38 1.58 701.22
2000 10.60 6.00 2.21 -1.44 -2.27 669.66
2001 10.96 5.31 2.54 -1.57 0.12 655.95
Croatia 1995 17.60 12.08 0.61 1.08 6.83 3337.13
1996 21.94 13.31 2.57 -0.17 5.90 3675.69
1997 27.52 17.69 2.65 3.15 6.80 3856.92
1998 24.02 14.99 4.31 9.46 2.52 4015.79
1999 23.04 11.23 7.37 -2.43 -0.86 3937.46
2000 20.23 12.50 5.91 -1.43 2.86 4207.13
2001 23.94 12.19 7.85 -0.23 4.44 4334.68
Czech 1995 32.50 23.17 4.65 -0.85 5.95 5017.28
Republic 1996 33.01 25.59 2.35 0.86 4.16 5234.06
1997 30.54 24.87 2.28 -0.68 -0.73 5201.50
1998 28.54 19.62 6.09 0.86 -1.15 5146.40
1999 26.94 13.42 10.69 2.28 1.21 5214.57
2000 28.75 15.90 8.95 -0.98 3.89 5422.55
2001 28.87 16.17 9.27 2.01 2.64 5592.48
El Salvador 1995 20.32 16.57 0.40 -0.45 6.40 1993.08
1996 15.19 13.64 -0.05 -0.14 1.71 1984.29
1997 15.12 13.24 0.53 0.06 4.25 2028.06
1998 17.55 5.39 9.19 0.30 3.75 2066.17
1999 16.44 13.09 1.73 0.09 3.45 2102.17
2000 16.93 13.85 1.32 0.12 2.15 2115.33
2001 16.68 14.02 2.02 0.29 1.71 2117.31
Estonia 1995 26.63 18.60 4.69 -7.70 4.54 2902.44
1996 26.98 20.09 3.23 -22.68 4.51 3078.45
1997 30.54 19.79 5.41 -6.92 10.52 3441.30
1998 30.22 16.85 10.42 4.81 5.21 3656.51
1999 24.95 16.60 5.49 0.20 -0.08 3681.59
2000 27.86 16.77 7.09 -1.25 7.80 3986.89
2001 29.20 16.32 9.09 2.40 6.39 4258.90
Source: World Development Indicators 2004 and Government Finance Statistics, International Monetary Fund Note: For the definitions, please see the Data and Econometric Framework section in the main text.
38
Appendix 3: Raw Data (continued)
Country Year GCF DCS FDI PFD GGDP PGDP
Guatemala 1995 15.06 11.68 0.51 -0.29 4.95 1593.03
1996 12.69 11.05 0.49 -0.29 2.96 1597.38
1997 13.68 12.38 0.47 -0.20 4.36 1623.68
1998 17.40 12.02 3.47 0.45 4.99 1660.24
1999 17.40 16.88 0.84 -0.17 3.85 1679.11
2000 17.80 14.51 1.19 -0.15 3.61 1694.34
2001 17.69 12.32 2.17 -0.15 2.33 1689.70
Hungary 1995 22.60 9.12 10.92 5.39 1.49 3748.86
1996 25.50 16.17 5.23 -0.20 1.32 3811.76
1997 26.57 17.37 4.86 -0.02 4.57 4000.73
1998 28.85 19.21 4.43 2.93 4.86 4212.17
1999 28.72 19.71 4.20 2.21 4.15 4407.17
2000 30.91 19.89 3.63 1.34 5.20 4656.88
2001 26.81 18.51 5.01 1.81 3.85 4758.77
India 1995 26.53 22.04 0.60 -0.19 7.65 370.52
1996 21.77 20.63 0.63 0.06 7.39 390.97
1997 22.57 19.33 0.87 0.41 4.48 401.42
1998 21.38 19.44 0.64 -0.53 5.99 418.20
1999 23.66 19.87 0.49 -0.20 7.13 440.46
2000 22.67 20.12 0.55 0.92 3.94 450.21
2001 22.31 19.58 0.79 -0.25 5.15 465.81
Indonesia 1995 31.93 20.09 2.15 9.08 8.40 826.89
1996 30.69 21.60 2.72 1.14 7.64 877.75
1997 31.75 21.16 2.17 6.05 4.70 906.46
1998 16.77 19.69 -0.37 -12.58 -13.13 776.87
1999 11.37 17.08 -1.96 -12.32 0.79 772.63
2000 21.38 18.43 -2.76 -4.76 4.92 800.04
2001 22.27 17.90 -2.00 -6.02 3.83 819.75
Iran, 1995 14.51 13.38 0.09 -15.85 3.30 524.97
Islamic RB 1996 20.45 15.26 0.02 0.15 5.52 1400.27
1997 30.16 13.02 0.05 -0.13 3.39 1450.27
1998 25.76 12.22 0.02 -0.08 2.00 1455.91
1999 22.46 4.26 0.04 0.03 1.71 1459.89
2000 34.72 12.27 0.04 -0.03 5.05 1511.20
2001 35.58 15.76 0.05 -0.11 3.42 1542.03
Source: World Development Indicators 2004 and Government Finance Statistics, International Monetary Fund Note: For the definitions, please see the Data and Econometric Framework section in the main text.
39
Appendix 3: Raw Data (continued)
Country Year GCF DCS FDI PFD GGDP PGDP
Jamaica 1995 28.77 19.59 2.54 -0.35 4.59 3169.66
1996 29.13 15.60 2.82 -0.66 2.28 3215.95
1997 29.26 21.61 2.73 0.82 -0.04 3189.25
1998 27.80 19.23 5.13 -0.40 -9.30 2869.77
1999 26.26 15.59 7.27 -0.35 0.91 2873.23
2000 28.73 18.14 6.32 -1.04 0.79 2873.61
2001 31.17 20.26 8.09 -0.19 1.52 2894.80
Kyrgyz 1995 18.34 12.92 5.79 -15.63 -5.42 227.14
Republic 1996 25.20 18.56 2.58 -6.69 7.08 239.73
1997 21.68 6.37 4.74 -7.98 9.92 259.71
1998 15.44 4.77 6.64 -8.45 2.12 261.24
1999 18.03 10.45 3.56 -13.65 3.66 267.03
2000 20.01 16.25 -0.17 -9.19 5.44 278.68
2001 18.00 14.54 0.33 -4.97 5.33 291.15
Mauritius 1995 28.84 23.42 0.49 0.27 4.06 3050.47
1996 25.32 21.13 0.88 -1.36 5.19 3174.74
1997 27.16 21.73 1.26 -2.30 5.86 3320.49
1998 27.61 22.05 0.29 -1.24 6.00 3483.23
1999 26.03 21.15 1.18 -0.78 5.30 3622.03
2000 26.13 16.63 6.00 -0.53 4.00 3726.92
2001 23.32 20.31 -0.61 -0.77 6.70 3933.55
Mexico 1995 19.97 11.04 3.32 -2.55 -6.22 4891.62
1996 23.24 13.24 2.76 -1.64 5.14 5063.83
1997 25.98 14.46 3.20 0.84 6.78 5328.90
1998 24.36 16.29 2.93 1.33 4.91 5512.61
1999 23.57 16.86 2.74 1.33 3.87 5647.08
2000 23.81 16.96 2.85 -0.84 6.60 5935.00
2001 20.91 14.04 4.32 0.17 -0.16 5841.54
Moldova 1995 24.88 12.25 1.48 -3.42 -1.40 336.67
1996 24.25 11.54 1.40 1.34 -5.88 317.91
1997 23.82 11.85 4.08 0.34 1.65 324.12
1998 25.87 13.69 4.44 3.62 -6.54 303.83
1999 22.88 13.37 3.24 0.70 -3.37 294.35
2000 23.95 3.73 10.02 3.99 2.11 301.26
2001 20.04 6.08 6.76 3.37 6.07 320.15
Source: World Development Indicators 2004 and Government Finance Statistics, International Monetary Fund Note: For the definitions, please see the Data and Econometric Framework section in the main text.
40
Appendix 3: Raw Data (continued)
Country Year GCF DCS FDI PFD GGDP PGDP
Pakistan 1995 18.55 12.28 1.19 0.61 4.96 510.42
1996 19.00 12.62 1.46 0.46 4.85 522.21
1997 17.92 12.54 1.15 0.20 1.01 515.00
1998 17.71 11.74 0.81 0.19 2.55 515.59
1999 15.56 10.87 0.84 -0.72 3.66 521.74
2000 17.38 13.86 0.42 0.33 4.26 531.00
2001 17.19 14.05 0.54 -0.70 1.86 527.99
Panama 1995 30.28 21.67 2.82 -0.01 1.75 3521.97
1996 26.68 15.70 4.46 0.16 2.81 3562.73
1997 26.69 8.44 12.88 0.70 6.38 3727.28
1998 27.42 11.18 11.00 0.98 7.38 3937.25
1999 27.81 16.39 7.55 1.93 3.97 4025.05
2000 28.50 18.66 6.03 2.63 2.71 4071.65
2001 26.00 18.53 3.43 5.23 0.57 4034.25
Paraguay 1995 23.93 18.53 1.14 -0.93 4.71 1585.87
1996 23.40 18.29 1.55 0.69 1.27 1569.14
1997 23.55 16.46 2.45 0.76 2.59 1572.98
1998 22.93 15.14 3.98 -0.05 -0.42 1530.80
1999 23.05 16.11 1.22 -0.17 0.49 1503.46
2000 25.60 15.52 1.54 -2.05 -0.30 1465.22
2001 24.99 14.65 1.23 -1.44 2.66 1470.75
Peru 1995 24.83 16.17 4.77 -0.05 8.58 1972.01
1996 22.85 13.44 6.22 -0.15 2.49 1982.77
1997 24.04 17.49 3.62 3.13 6.84 2080.53
1998 23.66 17.85 2.90 1.96 -0.68 2031.69
1999 21.59 14.76 3.78 0.41 0.89 2017.67
2000 20.25 16.17 1.53 0.74 2.92 2046.56
2001 18.63 14.23 2.13 0.41 0.18 2018.58
The Philippines 1995 22.45 17.45 1.99 2.28 4.68 915.87
1996 24.02 19.38 1.83 6.41 5.85 947.66
1997 24.78 21.05 1.48 4.54 5.19 974.34
1998 20.34 16.01 3.51 4.94 -0.58 946.80
1999 18.75 14.77 2.27 -1.62 3.40 956.71
2000 21.17 17.61 1.77 3.62 5.97 990.68
2001 18.76 14.78 1.37 0.90 2.96 997.94
Source: World Development Indicators 2004 and Government Finance Statistics, International Monetary Fund Note: For the definitions, please see the Data and Econometric Framework section in the main text.
41
Appendix 3: Raw Data (continued)
Country Year GCF DCS FDI PFD GGDP PGDP
Poland 1995 18.42 13.66 2.69 -15.20 7.00 3361.83
1996 20.47 15.32 2.93 4.30 6.00 3560.73
1997 23.01 17.50 3.19 -3.88 6.80 3799.71
1998 24.60 18.44 3.77 -1.23 4.80 3980.43
1999 24.87 18.72 4.42 4.32 4.10 4144.50
2000 24.67 17.01 5.61 1.04 3.95 4309.37
2001 20.74 16.73 3.08 2.15 1.02 4398.36
Romania 1995 24.27 17.68 1.18 0.39 7.16 1767.63
1996 25.85 20.02 0.74 6.06 4.01 1844.43
1997 20.63 16.23 3.44 -7.82 -6.10 1736.02
1998 17.75 12.14 4.82 -2.27 -4.79 1656.62
1999 16.08 14.06 2.92 -3.58 -1.20 1640.02
2000 19.47 14.51 2.80 -1.00 0.60 1650.97
2001 22.58 15.98 2.88 -0.53 5.30 1762.90
Senegal 1995 16.71 6.35 0.71 -0.25 5.17 407.06
1996 18.53 8.42 0.18 -0.17 5.14 416.37
1997 17.85 5.34 4.03 0.43 5.04 424.98
1998 19.77 9.02 1.52 -0.87 5.74 436.65
1999 20.64 8.27 3.30 -0.20 5.00 446.00
2000 18.52 8.75 1.64 -0.04 5.58 458.90
2001 18.12 10.52 0.81 -0.05 5.57 472.65
Slovak 1995 24.83 20.33 1.22 -0.36 5.86 3146.15
Republic 1996 34.75 26.57 1.68 6.58 6.16 3333.61
1997 34.53 28.86 0.82 -4.30 4.61 3481.13
1998 34.00 30.44 2.53 6.49 4.19 3621.81
1999 27.57 25.36 1.74 1.63 1.50 3673.11
2000 26.14 13.68 9.52 -4.02 2.02 3750.29
2001 30.00 18.54 7.58 -4.98 3.78 3900.69
South Africa 1995 18.17 13.84 0.83 -0.91 3.12 2960.42
1996 17.29 14.91 0.57 -1.40 4.31 3019.97
1997 16.64 13.07 2.56 -1.83 2.65 3029.77
1998 17.00 15.00 0.41 0.06 0.52 2974.68
1999 16.38 13.12 1.13 1.77 2.36 2972.21
2000 15.91 13.54 0.73 0.78 4.15 3019.95
2001 15.29 7.88 6.14 0.82 2.74 3046.31
Source: World Development Indicators 2004 and Government Finance Statistics, International Monetary Fund Note: For the definitions, please see the Data and Econometric Framework section in the main text.
42
Appendix 3: Raw Data (continued)
Country Year GCF DCS FDI PFD GGDP PGDP
Sri Lanka 1995 25.73 18.91 0.43 0.01 5.50 739.29
1996 24.25 18.13 0.86 -0.16 3.80 758.17
1997 24.39 16.57 2.85 0.34 6.40 797.04
1998 25.14 18.57 1.22 0.36 4.70 823.65
1999 27.29 20.70 1.13 0.05 4.30 846.19
2000 28.04 21.59 1.06 0.87 6.00 884.38
2001 22.00 16.11 1.09 -0.17 -1.55 858.40
Thailand 1995 42.09 34.42 1.23 18.25 9.24 2047.53
1996 41.82 33.89 1.29 6.25 5.90 2154.90
1997 33.66 22.22 2.58 -3.33 -1.37 2111.23
1998 20.45 4.66 6.54 -6.37 -10.51 1875.98
1999 20.50 3.66 4.99 -4.72 4.45 1944.70
2000 22.84 14.60 2.74 -5.63 4.75 2020.90
2001 24.10 15.30 3.37 -5.32 2.17 2049.31
Tunisia 1995 24.76 17.16 1.47 -0.14 2.32 1655.29
1996 25.06 16.74 1.22 -0.09 7.06 1746.38
1997 26.45 18.01 1.79 -0.08 5.44 1816.27
1998 26.91 16.68 3.28 -0.05 4.78 1879.04
1999 26.29 18.57 1.68 -0.08 6.05 1966.97
2000 27.28 16.71 3.86 0.12 4.67 2035.70
2001 27.85 18.16 2.29 0.04 4.86 2110.38
Turkey 1995 25.47 22.18 0.52 -43.35 7.19 2659.75
1996 24.55 23.57 0.40 3.25 7.01 2794.67
1997 25.11 24.11 0.43 -15.57 7.53 2951.47
1998 24.18 22.99 0.47 -12.01 3.09 2989.40
1999 23.35 19.95 0.43 -12.23 -4.71 2799.82
2000 24.51 20.18 0.49 -7.12 7.36 2955.61
2001 16.78 14.21 2.25 -5.03 -7.49 2689.86
Source: World Development Indicators 2004 and Government Finance Statistics, International Monetary Fund Note: For the definitions, please see the Data and Econometric Framework section in the main text.
43
Appendix 3: Raw Data (continued)
Country Year GCF DCS FDI PFD GGDP PGDP
Uruguay 1995 15.41 10.87 0.81 0.40 -1.45 5622.23
1996 15.24 11.74 0.67 -0.22 5.58 5891.90
1997 15.22 12.49 0.58 0.38 5.05 6145.70
1998 15.87 13.02 0.73 0.22 4.54 6377.73
1999 15.14 11.75 1.13 0.70 -2.85 6170.30
2000 13.96 10.48 1.36 0.36 -1.44 6045.97
2001 13.77 9.82 1.46 -0.16 -3.39 5807.42
Venezuela 1995 18.11 12.33 1.27 -9.76 3.95 5119.60
Republic 1996 16.55 10.38 3.09 -5.74 -0.20 5005.25
1997 26.78 14.27 6.99 -0.38 6.37 5218.15
1998 30.66 18.63 5.46 3.08 0.29 5132.01
1999 26.52 17.48 2.95 -0.57 -5.97 4733.82
2000 24.17 13.18 4.01 -1.51 3.69 4818.71
2001 27.52 15.63 3.00 -0.65 3.39 4890.93
Source: World Development Indicators 2004 and Government Finance Statistics, International Monetary Fund Note: For the definitions, please see the Data and Econometric Framework section in the main text.
44
Appendix 4: A Summary of The World Bank’s Governance Indicators
According to Kaufman, Kray and Mastruzzi (2005), The World Bank’s governance
indicators measure the following six dimensions of governance:
1. Voice and Accountability – measuring the extent to which citizens of a country
are able to participate in the selection of governments, and the independence of
the media, which serves a significant role in monitoring those in authority and
holding them accountable for their actions
2. Political Instability and Violence – measuring perceptions of the likelihood that
the government in power will be destabilised or overthrown by possibly
unconstitutional and/or violent means, including domestic violence and terrorism
3. Government Effectiveness – focusing on inputs required for the government to
be able to produce and implement good policies (the quality of the bureaucracy,
the credibility of the government’s commitment to policies, for instance) and
deliver public goods
4. Regulatory Quality – measuring the incidence of market-friendly policies in areas
such as foreign trade and business development
5. Rule of Law – measuring the extent to which agents have confidence in and
abide by the rules of society and the extent to which property rights are protected.
These include perceptions of the incidence of crime, the effectiveness and
predictability of the judiciary, and the enforceability of contracts.
6. Control of Corruption – measuring the exercise of public power for private gain,
including both grand corruption and state capture
They cover 209 countries and territories for 1996, 1998, 2000, 2002 and 2004. They
are based on several hundred individual variables measuring perceptions of governance,
drawn from 37 separate data sources constructed by 31 different organisations such as
45
Freedom House, World Economic Forum, European Bank for Reconstruction and
Development and so forth.
An unobserved components model – providing estimates of governance for each
country and measures of the precision of these estimates for every country, indicator and
year – is employed to construct the six aggregate governance indicators in each period. The
governance estimates are normally distributed with a mean of zero and a standard deviation
of one in each period. This implies that virtually all scores lie between –2.5 and 2.5, with
higher scores corresponding to better outcomes (The World Bank also shows the point
estimates in percentile rank.). Table A and B present comparison within one country for all
six governance indicators and comparison for all six governance indicators across the 36
countries, respectively.
Based on table A, we can discuss the changes over time in the governance estimates
in individual countries. Changes in the estimates of governance in most countries are
relatively small over the seven-year period. However, the estimates of governance do
change substantially for some countries in some periods. For example, from 1996 to 2002,
countries like Croatia and Peru show substantial improvements in, among others, Voice and
Accountability measure, contrasting Cote d’lvoire deteriorates on it. Cote d’lvoire, Kyrgyz
Republic and Argentina largely decline on Political Stability and Absence of Violence
measure. Cote d’lvoire, Argentina and Indonesia considerably deteriorate on the
Government Effectiveness measure. The changes in the estimates also show significant
deterioration on the Regulatory Quality measure in Argentina, Paraguay, Indonesia and
Bolivia. Argentina and Cote d’lvoire show substantial declination on the Rule of Law and
the Control of Corruption, respectively (Kaufman, Kray and Mastruzzi (2005)).
46
Table A: Comparison within One Country for All Six Governance Indicators
Voice and Accountability
(percentile rank)
Political Stability and
Absence of Violence
(percentile rank) Country
2002 2000 1998 1996 2002 2000 1998 1996
Argentina 57.6 61.8 59.7 66.5 27.0 64.8 63.6 63.4
Bolivia 50.0 56.5 60.2 56.0 42.7 36.4 50.9 37.2
Bulgaria 66.2 63.4 61.3 57.6 63.8 60.6 63.0 51.8
Chile 81.8 64.9 67.0 74.9 83.2 77.0 71.5 71.3
Costa Rica 84.8 90.1 85.3 90.6 87.6 87.9 84.8 78.7
Cote d’lvoire 11.6 12.6 30.4 46.6 2.2 17.0 45.5 58.5
Croatia 64.1 59.7 39.3 34.0 62.2 65.5 64.2 56.7
Czech Republic 74.7 76.4 83.2 78.0 85.9 75.2 83.0 86.6
El Salvador 51.5 57.1 52.4 43.5 56.8 64.2 56.4 43.3
Estonia 80.3 72.3 71.2 70.2 82.2 75.2 80.6 75.6
Guatemala 34.8 42.9 37.2 28.8 31.4 17.6 18.2 14.6
Hungary 85.4 84.3 83.8 78.0 88.6 73.9 87.9 73.8
India 60.6 62.8 58.6 60.7 21.6 37.0 27.3 18.9
Indonesia 34.3 32.5 12.0 16.2 11.4 3.6 9.1 28.0
Iran, Islamic Republic 18.7 26.7 23.0 18.3 25.4 41.2 37.6 32.3
Jamaica 65.2 67.5 67.5 65.4 39.5 60.0 42.4 66.5
Kyrgyz Republic 20.7 27.2 35.6 34.6 17.8 44.2 73.3 72.6
Mauritius 71.7 86.4 74.9 72.3 88.6 86.1 89.7 89.6
Mexico 59.6 55.0 44.5 42.9 52.4 43.0 25.5 34.1
Moldova 39.9 51.8 49.7 44.5 42.7 44.2 51.5 40.9
Pakistan 16.2 6.8 31.9 21.5 13.0 25.5 18.8 10.4
Panama 64.6 67.0 63.9 61.8 59.5 68.5 60.0 59.1
Paraguay 32.3 30.9 40.8 38.7 17.8 18.8 33.3 44.5
Peru 52.5 51.8 25.7 27.2 23.8 32.1 26.1 17.7
Philippines 54.0 60.2 63.4 57.6 28.1 37.6 48.5 41.5
Poland 83.8 83.8 76.4 75.9 71.9 75.2 76.4 68.3
Romania 61.1 60.7 58.1 54.5 57.3 49.7 56.4 64.6
Senegal 53.0 47.1 35.1 47.1 34.1 23.0 15.2 21.3
Slovak Republic 75.8 72.8 62.8 62.3 81.1 70.9 80.6 65.9
South Africa 70.7 80.1 72.3 67.5 36.8 42.4 21.2 16.5
Sri Lanka 48.0 36.6 40.3 44.5 20.0 4.2 6.1 6.1
Thailand 57.1 57.6 55.0 52.9 61.1 57.6 59.4 53.0
Tunisia 22.7 26.2 20.4 33.0 54.6 70.9 65.5 56.7
Turkey 35.4 29.8 20.4 37.7 25.9 15.2 14.5 10.4
Uruguay 77.8 79.6 69.6 71.7 78.9 82.4 70.3 77.4
Venezuela, Republic 37.9 39.3 56.0 55.0 16.2 33.9 32.1 22.6
Source: The World Bank – http://www.worldbank.org/wbi/governance/govdata/
47
Table A: Comparison within One Country for All Six Governance Indicators (continued)
Government Effectiveness
(percentile rank)
Regulatory Quality
(percentile rank)
Country
2002 2000 1998 1996 2002 2000 1998 1996
Argentina 36.8 65.1 74.3 73.2 20.9 67.9 81.0 81.2
Bolivia 32.8 38.7 54.6 34.6 49.5 75.4 82.1 81.2
Bulgaria 59.7 48.9 13.7 33.5 69.9 54.5 65.2 44.8
Chile 87.1 88.2 87.4 84.4 90.8 90.4 92.4 93.9
Costa Rica 69.2 76.9 75.4 65.4 74.5 82.4 83.2 77.3
Cote d’lvoire 15.9 22.0 53.0 56.4 41.3 34.2 50.5 39.8
Croatia 66.2 60.2 69.9 54.2 62.2 59.9 60.9 44.8
Czech Republic 75.1 74.7 80.3 81.6 82.1 75.4 75.5 86.7
El Salvador 34.8 51.1 57.4 40.8 57.1 88.8 95.7 79.6
Estonia 78.1 82.3 73.8 76.0 87.2 89.8 84.8 90.1
Guatemala 30.8 34.9 44.8 29.1 51.5 69.0 79.9 51.9
Hungary 76.1 78.0 80.9 75.4 84.2 88.2 88.0 74.6
India 55.7 52.7 50.8 55.3 43.4 38.5 41.8 44.2
Indonesia 31.8 38.7 30.1 66.5 25.0 33.7 46.2 64.1
Iran, Islamic Republic 37.3 48.4 41.5 45.8 10.2 10.2 6.5 5.5
Jamaica 58.7 46.8 27.3 36.9 64.3 66.8 71.7 72.9
Kyrgyz Republic 24.9 28.5 44.3 35.2 38.8 33.2 20.1 38.7
Mauritius 70.6 78.5 71.6 79.3 68.4 78.1 64.1 59.1
Mexico 65.7 67.2 68.9 55.9 66.8 76.5 75.5 74.0
Moldova 28.9 13.4 32.8 31.8 48.0 12.3 30.4 53.6
Pakistan 32.8 32.8 22.4 38.5 21.4 24.1 39.7 24.3
Panama 55.7 54.8 59.6 29.6 67.3 85.6 92.9 76.2
Paraguay 7.5 9.1 8.7 21.2 30.6 17.6 35.3 73.5
Peru 37.3 45.7 69.9 53.1 62.2 73.3 81.5 76.2
Philippines 58.2 58.6 67.8 68.2 55.6 63.1 72.8 68.5
Poland 72.6 68.8 82.5 77.7 70.9 73.8 77.7 68.5
Romania 46.3 30.6 24.6 29.6 55.1 35.8 58.2 30.4
Senegal 55.7 64.5 63.9 38.5 45.4 45.5 35.3 29.3
Slovak Republic 68.2 65.1 61.7 69.3 73.5 64.7 57.1 64.1
South Africa 72.1 69.9 64.5 70.4 68.9 52.4 60.3 64.1
Sri Lanka 61.2 40.9 37.7 45.8 61.7 60.4 73.4 66.3
Thailand 66.7 62.9 62.8 74.3 64.8 77.5 56.0 69.6
Tunisia 73.6 86.6 82.5 74.9 53.6 64.7 67.4 52.5
Turkey 55.2 52.7 41.5 62.0 55.6 55.1 80.4 71.3
Uruguay 71.1 75.3 78.7 76.0 67.3 86.6 84.2 84.0
Venezuela, Republic 9.5 18.8 14.8 20.1 31.6 25.7 47.8 44.8
Source: The World Bank – http://www.worldbank.org/wbi/governance/govdata/
48
Table A: Comparison within One Country for All Six Governance Indicators (continued)
Rule of Law
(percentile rank)
Control of Corruption
(percentile rank) Country
2002 2000 1998 1996 2002 2000 1998 1996
Argentina 25.5 61.0 64.9 65.7 27.0 44.6 58.5 54.0
Bolivia 30.1 39.6 41.1 28.9 25.0 30.1 42.6 22.0
Bulgaria 54.6 55.1 52.4 56.0 53.1 54.8 39.3 29.3
Chile 85.2 86.6 85.4 86.7 90.8 90.3 86.3 86.0
Costa Rica 71.4 78.1 79.5 73.5 80.6 84.4 80.3 80.7
Cote d’lvoire 7.1 34.8 35.1 25.9 18.9 33.9 44.8 72.0
Croatia 57.7 60.4 59.5 32.5 63.8 59.7 45.9 34.7
Czech Republic 72.4 72.2 73.5 73.5 66.8 73.1 72.7 77.3
El Salvador 41.8 41.7 54.1 35.5 38.8 53.8 54.1 25.3
Estonia 74.0 75.4 70.8 66.9 74.5 80.1 76.5 60.0
Guatemala 21.4 23.0 27.6 30.1 32.1 30.6 27.3 18.0
Hungary 77.6 78.6 76.2 75.3 74.0 80.1 79.8 78.0
India 55.6 62.0 67.0 56.6 43.4 49.5 59.6 43.3
Indonesia 20.4 13.9 14.1 39.2 8.2 11.3 8.7 35.3
Iran, Islamic Republic 32.7 42.2 36.8 21.7 43.4 34.9 27.3 23.3
Jamaica 38.3 54.0 51.4 47.6 40.3 52.7 54.6 40.7
Kyrgyz Republic 26.0 16.6 29.2 25.9 23.5 18.3 25.7 24.7
Mauritius 78.1 79.1 81.1 76.5 71.9 76.3 70.5 75.3
Mexico 47.4 46.0 40.0 54.2 51.0 43.5 43.7 39.3
Moldova 34.2 36.4 54.6 48.2 21.9 19.9 38.3 50.7
Pakistan 28.6 32.6 25.9 38.0 25.5 22.6 19.7 12.7
Panama 54.1 58.3 60.5 65.1 49.5 46.2 53.0 33.3
Paraguay 11.2 19.8 23.8 34.3 5.6 10.8 7.7 33.3
Peru 36.7 38.5 37.3 40.4 50.5 57.5 59.6 56.0
Philippines 33.7 40.6 59.5 55.4 38.3 39.2 54.6 38.0
Poland 68.9 72.7 72.4 69.9 68.4 73.7 76.5 72.0
Romania 52.0 51.9 49.7 44.0 45.4 39.8 44.3 51.3
Senegal 50.0 47.6 49.2 53.0 51.5 42.5 41.0 38.7
Slovak Republic 64.8 64.7 63.8 61.4 64.8 68.3 62.8 72.0
South Africa 58.2 64.2 67.0 66.9 65.8 75.3 74.3 78.0
Sri Lanka 59.2 52.9 55.7 66.3 53.6 57.0 57.9 50.0
Thailand 60.2 69.0 69.2 71.1 47.4 47.8 54.6 42.0
Tunisia 61.2 69.5 69.7 59.6 70.4 78.0 68.9 58.0
Turkey 53.1 59.9 65.9 58.4 40.8 48.9 65.6 61.3
Uruguay 68.4 74.3 70.8 72.3 77.6 79.6 74.3 74.0
Venezuela, Republic 13.8 20.9 33.5 28.9 17.9 32.8 18.6 28.0
Source: The World Bank – http://www.worldbank.org/wbi/governance/govdata/
49
Table B: Comparison for Six Governance Indicators across 36 Developing Countries
Average Percentile Rank (1996-2002) Voice and Political Stability and Government Regulatory Rule of Control ofCountry
Accountability Absence of Violence Effectiveness Quality Law Corruption
Argentina 61.4 54.7 62.4 62.8 54.3 46.0
Bolivia 55.7 41.8 40.2 72.1 34.9 29.9
Bulgaria 62.1 59.8 39.0 58.6 54.5 44.1
Chile 72.2 75.8 86.8 91.9 86.0 88.4
Costa Rica 87.7 84.8 71.7 79.4 75.6 81.5
Cote d’lvoire 25.3 30.8 36.8 41.5 25.7 42.4
Croatia 49.3 62.2 62.6 57.0 52.5 51.0
Czech Republic 78.1 82.7 77.9 79.9 72.9 72.5
El Salvador 51.1 55.2 46.0 80.3 43.3 43.0
Estonia 73.5 78.4 77.6 88.0 71.8 72.8
Guatemala 35.9 20.5 34.9 63.1 25.5 27.0
Hungary 82.9 81.1 77.6 83.8 76.9 78.0
India 60.7 26.2 53.6 42.0 60.3 49.0
Indonesia 23.8 13.0 41.8 42.3 21.9 15.9
Iran, Islamic Republic 21.7 34.1 43.3 8.1 33.4 32.2
Jamaica 66.4 52.1 42.4 68.9 47.8 47.1
Kyrgyz Republic 29.5 52.0 33.2 32.7 24.4 23.1
Mauritius 76.3 88.5 75.0 67.4 78.7 73.5
Mexico 50.5 38.8 64.4 73.2 46.9 44.4
Moldova 46.5 44.8 26.7 36.1 43.4 32.7
Pakistan 19.1 16.9 31.6 27.4 31.3 20.1
Panama 64.3 61.8 49.9 80.5 59.5 45.5
Paraguay 35.7 28.6 11.6 39.3 22.3 14.4
Peru 39.3 24.9 51.5 73.3 38.2 55.9
Philippines 58.8 38.9 63.2 65.0 47.3 42.5
Poland 80.0 73.0 75.4 72.7 71.0 72.7
Romania 58.6 57.0 32.8 44.9 49.4 45.2
Senegal 45.6 23.4 55.7 38.9 50.0 43.4
Slovak Republic 68.4 74.6 66.1 64.9 63.7 67.0
South Africa 72.7 29.2 69.2 61.4 64.1 73.4
Sri Lanka 42.4 9.1 46.4 65.5 58.5 54.6
Thailand 55.7 57.8 66.7 67.0 67.4 48.0
Tunisia 25.6 61.9 79.4 59.6 65.0 68.8
Turkey 30.8 16.5 52.9 65.6 59.3 54.2
Uruguay 74.7 77.3 75.3 80.5 71.5 76.4
Venezuela, Republic 47.1 26.2 15.8 37.5 24.3 24.3
Source: The World Bank – http://www.worldbank.org/wbi/governance/govdata/
50
Appendix 5: Descriptive statistics and correlation matrix
Descriptive statistics
Sample: 36 countries and 1995-2001
Variable Mean Max Min S.D.
Gross capital formation as a share of GDP
(GCF: constant 2000, US dollar)
22.41 42.09 8.12 5.61
Self-financing domestic capital as a share of GDP
(SDC: constant, US dollar)
15.04 34.42 3.66 5.18
Net inflows of FDI to the countries as a share of GDP
(FDI: constant 2000, US dollar)
3.19 12.88 -2.76 2.83
Percentage change in private foreign debt to GDP
(PFD: constant 2000, US dollar)
-1.25 18.25 -52.01 7.01
GDP per capita
(PGDP: constant 2000, US dollar)
2,650.87 8,234.93 227.14 1,859.16
GDP growth
(GGDP: constant 2000, US dollar)
3.15 10.63 -13.12 3.71
Source: World Development Indicators 2004, International Monetary Fund, and author’s computation
Correlation matrix
Sample: 36 countries and 1995-2001
GCF SDC FDI PFD PGDP GGDP
GCF 1
SDC 0.70 1
FDI 0.15 -0.34 1
PFD 0.31 0.12 0.14 1
PGDP 0.15 0.16 0.25 0.14 1
GGDP 0.26 0.20 -0.05 0.22 -0.04 1
Source: World Development Indicators 2004, International Monetary Fund, and author’s computation.
51
Appendix 6: Prais-Winsten model with panel-corrected standard errors
According to Beck and Katz (1995), Prais-Winsten model with panel-corrected
standard errors is to estimate the panel model when disturbances are not assumed to be
independent and identically distributed (iid). The model can be rewritten as follow
yi,t = xi,t β + εi,t
where yi,t is the dependent variable, i = 1, …, m, i is the number of panels, t = 1, …, T,
T is the number of time periods in panel i, xi,t is a vector of one or more independent
variables and observations are indexed by both panel i and time t, and εi,t is a disturbance
that may be autocorrelated along time or heteroscedastic across panels,
tiiti ,, υεε +=
where iε ~ IIN(0, 2εσ ), tititi ,1,. ωρυυ += − , 1<ρ , and ti,ω ~ IIN (0, 2
ωσ ). The si 'ε are
independent of the .', stiω
For a model with heteroscedastic disturbances, the disturbance covariance matrix is
assumed to be
[ ]⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢
⎣
⎡
Ι
ΙΙ
=Ω=
mmm
E
2
2222
1121
'
00
0000
σ
σσ
εε
K
MOMM
K
K
where σi2 is the variance of the disturbances for panel i and I is a Ti by Ti identity matrix.
These could also be written as
[ ] TTmm **' Ι⊗∑=Ω=Ε εε
52
where mm*Σ is the panel-by-panel covariance matrix and I is an identity matrix.
If the errors do not follow the condition of ‘iid’, then Ordinary Least Squares (OLS)
estimates of β will be consistent but inefficient. They will be corrected and named as panel-
corrected standard errors. Any serial correlation of the errors, however, must be eliminated
before the panel-corrected standard errors are calculated.
First, it runs OLS on the first equation and calculates an estimate of ρ in AR(1)
process, say ρ̂ . Given ρ̂ , it performs Prais-Winsten transformation by premultiplying the
dependent and independent variables by (I ⊗C) where I is an identity matrix and
C =
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−−
−
⎟⎠⎞⎜
⎝⎛ −
1ˆ000001ˆ000
00001ˆ000001
2/12ˆ
ρρ
ρρ
K
K
MMMKMMM
MMMKMMM
K
K
.
This yields ( ) titi yCy ,*, ⊗Ι= and ( ) titi xCx ,
*, ⊗Ι= .
Calculate an estimate of residual ( )ti,ε , say *,tie , obtained from OLS regression of *
,tiy
and *,tix . Using the *
,tie ’s calculates consistent estimates of 2μσ , say 2ˆ μσ , and 2
εσ , say 2ˆ εσ ,
and αθ . In fact,
mcd i
m
i/ˆ 2
1
22 ∑=
=μσ and ( ) ( ) ( )1/ˆ1 1 2
2*,
2*1,
2 −⎥⎦
⎤⎢⎣
⎡−+−= ∑ ∑∑
= = =
Tmcecem
i
m
i
T
titiii ασ ε
where )1(22 −+= Td α , ρραˆ1ˆ1
−+
= , ( ) 22
*,
*1, / deec T
tiii ∑+= α . Finally, μ
εα σ
σθ
ˆˆ
1−= .
53
The final step to eliminate the first-order autocorrelation and obtain the estimate of β
is to premultiply the Prais-Winsten transformed model by 2/1ˆ −Ψεσ and run OLS to get
feasible generalised least squares estimate of β. The Prais-Winsten transformation
( ) titi yCy ,*, ⊗Ι= becomes *
,2/1**
, ˆ titi yy −Ψ= εσ and ( ) titi xCx ,*, ⊗Ι= becomes
*,
2/1**, ˆ titi xx −Ψ= εσ where ( )α
αε θσ TmTm J⊗Ι−Ι⊗Ι=Ψ− 2/1ˆ , 2/ dJ TTTααα ηη ′= ,
( )1, −= TT ηαηα , and Tη is a vector of ones of dimension T (Baltagi and Li (1991)).
The correct standard errors are calculated by square roots of the diagonal terms of
( ) ( ) [ ]( ) 11ˆ −− ′Ω′′= xxxxxxCov β . It is employed to provide accurate panel-corrected standard
errors in combination with the panel-by-panel covariance matrix. To estimate Ω , say Ω̂ ,
we need an estimate of mm*Σ , say mm*Σ̂ . Since the OLS estimates are consistent, we can use
the estimated OLS errors (from previous estimation) to provide mm*Σ̂ and Ω̂ . These finally
lead to the panel-corrected standard errors.
54
Appendix 7: Estimation Results by Random Effects Model
Table A: Effects of domestic capital self-financing, FDI financing and private foreign debt financing on domestic investment in the countries:
1996-2001
Dependent Variable: Gross capital formation as a share of GDP (logarithm, constant 2000)
RANDOM EFFECTS ESTIMATION INDEPENDENT VARIABLES
(1) (2) (3) (4) (5) (6) (7)
Lagged self-financing domestic capital/ GDP (logarithm, b1) 0.12(0.00) 0.13(0.00) 0.12(0.01) 0.13(0.00) 0.15(0.01) 0.12(0.06) 0.08(0.07)
Lagged FDI inflows as a share of GDP (logarithm, b2) 0.04(0.03) 0.04(0.02) 0.04(0.03) 0.04(0.03) 0.03(0.21) 0.02(0.29) 0.01(0.48)
Lagged change in private external debt to GDP
(logarithm, constant 2000; b3)
-0.01(0.91) -0.01(0.89) -0.01(0.93) -0.01(0.87) 0.01(0.99) 0.01(0.96) 0.27(0.22)
Lagged per capita GDP (logarithm, constant 2000; b4) 0.04(0.18) 0.05(0.12) 0.03(0.36) 0.06(0.12) 0.04(0.23) 0.02(0.54) 0.04(0.34)
Lagged GDP growth (logarithm, constant 2000; b5) 0.10(0.00) 0.10(0.00) 0.10(0.00) 0.10(0.00) 0.10(0.00) 0.10(0.00) 0.09(0.00)
Constant 2.12(0.00) 2.05(0.00) 2.17(0.00) 2.01(0.00) 2.10(0.00) 2.27(0.00) 1.00(0.23)
Voice and accountability dummy variable (VA; b6) -0.05(0.45) -0.48(0.27)
Political stability dummy variable (PS; b7) -0.04(0.58) -0.49(0.26)
Regulatory quality dummy variable (RQ; b8) -0.06(0.41) 0.61(0.49)
VA * Lagged self-financing domestic capital/ GDP (b9) 0.07(0.40)
VA * Lagged FDI inflows as a share of GDP (b10) 0.17(0.01)
VA * Lagged change in private external debt to GDP (b11) -0.01(0.93)
55
PS * Lagged self-financing domestic capital/ GDP (b12) 0.11(0.19)
PS * Lagged FDI inflows as a share of GDP (b13) 0.16(0.01)
PS * Lagged change in private external debt to GDP (b14) -0.01(0.90)
RQ * Lagged self-financing domestic capital/ GDP (b15) 0.18(0.00)
RQ * Lagged FDI inflows as a share of GDP (b16) 0.16(0.01)
RQ * Lagged change in private external debt to GDP (b17) -0.28(0.20)
Coefficients of determinations 0.78 0.75 0.79 0.79 0.87 0.88 0.90
Number of observations 216 216 216 216 216 216 216
Remark: The figures are the coefficient estimates, and the quantities in parentheses are the P-values.