Economic Modelling 37 (2014) 332–339

Contents lists available at ScienceDirect

Economic Modelling
journal homepage: www.elsevier.com/locate/ecmod

Growth and Foreign Direct Investment in the Pacific

Island countries☆,☆☆
Simon Feeny ⁎, Sasi Iamsiraroj, Mark McGillivray
Alfred Deakin Research Institute, Deakin University, Geelong, Australia

a r t i c l e i n f o a b s t r a c t

Article history: Achieving sustained high rates of economic growth in Pacific countries has proved incredibly challenging. Despite
Accepted 15 November 2013 many being rich in natural resources, receiving high levels of foreign aid and being open to external trade, the
economic growth rates of Pacific Island countries are the lowest and most volatile for all groups of developing
JEL classification:
countries. This paper examines the impact of Foreign Direct Investment (FDI) to the Pacific region. Results
F21
from the estimation of a number of empirical models suggest that the impact of FDI is lower in Pacific countries
O11
than it is in host countries on average. A 10% increase in the ratio of FDI to host Gross Domestic Product (GDP) is
Keywords: associated with higher growth of about 2% in all countries on average. The impact in Pacific countries falls to
Foreign Direct Investment between 0.1 and 0.4%. A number of explanations for this finding are provided including some empirical evidence
Economic growth that FDI displaces domestic investment in the region.
Pacific © 2013 Elsevier B.V. All rights reserved.

1. Introduction
Pacific Island countries (PICs) face many tremendous challenges.
These include small domestic markets, limited resource bases, great
distances from major markets, vulnerability to external shocks such
as hikes in the prices of key international commodities and natural
disasters, often inadequate governance and political instability. Western
governments seemingly recognise this to the extent that PICs receive
some of the highest levels of foreign development aid in the world
relative to the size of their economies. In 2009, Official Development
Assistance (ODA) accounted for as much as 34% of GDP in the
Solomon Islands and 68% of GDP in Tuvalu. On average ODA accounted
for over 25% of recipient GDP in PICs. Despite these aid levels, PICs have
experienced the lowest and most volatile economic growth of
any region of developing countries, including sub-Saharan Africa
(McGillivray et al., 2010). Partly due to these growth rates, poverty is
increasing in the region. Approximately 2.7 million people, one-third
of the region's population, live in poverty, without the income to satisfy
their basic human needs. More than 400,000 children are not enrolled in
☆ This paper is an output from an Australian Research Council funded linkage project
with Sustineo Pty Ltd entitled Supporting Pacific Development (LP110200746).
☆☆ The authors are grateful for helpful comments and suggestions from two anonymous
referees.
⁎ Corresponding author at: Alfred Deakin Research Institute, Deakin University,
Geelong, VIC 3220, Australia. Tel.: +61 3 5227 1428; fax: +61 3 9918 9001.
E-mail address: simon.feeny@deakin.edu.au (S. Feeny).
primary school and seven out of every 100 children die before their fifth
birthday. At least 80,000 adults have HIV and the rate of infection is
growing by more than 40% per year, the fastest of any region of the
world (AusAID, 2009).
Despite the immense challenges faced by the Pacific, the region
remains startlingly under-researched. There is a small literature
examining the economic growth impact of foreign aid and remittances
to the region. For example, Pavlov and Sugden (2006), Feeny (2007)
and Feeny and McGillivray (2010) provide positive assessments of
foreign aid to Pacific countries, while Connell and Brown (2005),
Browne and Mineshema (2007) and Brown (2008) examine the impact
of remittances. Yet, surprisingly, the impact of Foreign Direct Invest-
ment (FDI) has been neglected.1 This is despite FDI accounting for a
greater share of Gross Domestic Product (GDP) in Pacific countries
than for developing countries on average. The main objective of the
paper is therefore to help fill the void in development related research
for the PICs by providing the first study to comprehensively examine
the growth impact of FDI to the region. It is, to the knowledge of the
authors, the first study to look at the impact of FDI in these countries.
A fundamental premise of this paper is that the behavioural relationship
between growth and FDI in Pacific countries cannot be assumed to be
the same as that for other countries, observed either from cross country,
panel or time series datasets. It would therefore be inappropriate to rely
1
Jayaraman and Singh (2007) provide the only exception, finding that FDI contributes
to employment creation and economic growth in the case of Fiji. See Read (2007) for an
examination of the determinants of FDI flows to Small Island Developing States (SIDS).

0264-9993/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.econmod.2013.11.018
 S. Feeny et al. / Economic Modelling 37 (2014) 332–339
on the findings of previous studies to draw inferences about the impact
of FDI in PICs. The main reason for expecting that the incremental im-
pact of FDI will be different in the Pacific is the widely acknowledged
difficulty of doing business in these countries. As noted in AusAID
(2008) and World Bank (2011), doing business in these countries is be-
coming increasingly difficult. This is likely to retard the productivity of
FDI in the region. It is also recognised that FDI might displace domestic
investment in PICs, thereby limiting its effectiveness with respect to
growth. FDI is particularly likely to displace domestic investment if
foreign firms have superior managerial and technical expertise than
their domestic counterparts.
FDI can play a crucial role in contributing to growth and poverty
reduction in host countries, particularly in small countries located a
long way from major markets. These countries often lack the resources
to develop their own technology and suffer from technical and institu-
tional constraints to the accumulation of physical and human capital.
Domestic financing for investment projects can be limited and unpro-
tected property rights, corruption, and civil and political instability
may either hinder capital accumulation, or become obstacles for using
already existing resources. FDI should, therefore, be an attractive source
of development financing for these countries. A number of positive ex-
ternalities are associated with FDI inflows, such as advanced technology,
managerial expertise, R&D, employment, productivity and efficiency
gains in the domestic economy.2
Support for FDI is not universal despite its potential benefits. Critics
argue that the policies to attract FDI can distort domestic incentives and
as noted above, can displace domestic investment, crowding out
employment and domestic firms. The impact of FDI on the host country
is therefore an empirical issue and one that has been examined by a
voluminous (cross-country) empirical literature. The consensus of this
literature is that the impact of FDI has been favourable, by contributing
to the economic growth rates of host countries. It is also clear from a re-
view of this literature that the impact of FDI varies across host countries.
Little is known of the impact of FDI on Pacific Island countries. There
is anecdotal evidence to suggest that FDI has provided little benefit to
some of these countries. FDI to Papua New Guinea (PNG) and the
Solomon Islands, for example, has been characterised by large capital
intensive projects in the extractive industries (mining and logging) of
these economies. These industries have been plagued by corruption,
widespread concerns over environmental damage and the exploitation
of domestic landowners. FDI has focused on agriculture and tourism in
other Pacific countries, sectors which (directly and indirectly) provide
employment to a substantial proportion of their populations.
The paper finds that FDI is associated with higher rates of economic
growth in the Pacific. Yet it also finds that the impact of FDI is lower in
the region than it is for countries on average. Further evidence is pre-
sented which indicates that FDI has displaced domestic investment in
the Pacific which sets this region apart. One of the recommendations
emanating from this finding is for donor governments to shift the nature
of their assistance to the region. There are limits to what aid can achieve
to the region and instead of providing additional assistance, a change in
the focus of aid is appropriate. More specifically, donors should examine
ways of improving the growth (and other) impacts of FDI in the Pacific.
This is likely to require a greater focus on improvements in human
capital and private sector development.
2. An overview of FDI to the Pacific
Pacific countries suffer from a number of factors which might limit
the desire of international firms to invest in them. These include their
2
In the context of the neoclassical growth framework, FDI affects only the level of in-
come. Long-run economic growth will be unchanged unless FDI influences technological
progress (see de Mello, 1997; Solow, 1957). In endogenous growth models, however, eco-
nomic growth will be generated if FDI increases productivity via externalities and spillover
effects such as know-how and improvements in human capital (Lucas, 1988; Rebelo,
1991; Romer, 1986).
333
small size, remoteness, insecure land rights, high cost and low availabil-
ity of skilled labour and unfavourable business environments (Nathan
Associates Inc., 2007). Yet to varying extents, Pacific countries have
overcome these obstacles. Outside of the extractive industries, FDI in
the Pacific has flowed to agriculture (the palm oil, copra, sugar and live-
stock industries), forestry, fishing, banking and finance, real estate and
tourism. There has been less FDI to the manufacturing sector. A notable
exception is the Yazaki automobile wiring harness plant in Samoa.
While flows of FDI to the Pacific are small in absolute terms, levels
of FDI (relative to the size of host country GDP) are actually higher
(on average) in PICs than they are in other low and middle income
countries. This is demonstrated by Table 1 which provides net FDI
inflows relative to GDP for selected Pacific countries and country groups
during the last three decades.
Fiji and Vanuatu have received far higher levels of FDI (relative to
low and middle income countries on average) in each of the last three
decades. The same is true for Kiribati and the Solomon Islands in the
1990s and 2000s. Levels of FDI to Samoa and Tonga have been lower
and fluctuated widely. While PNG has received a lower level of FDI
in the 2000s, the level is starting to increase dramatically with the
construction of a large Liquid Nitrogen Gas (LNG) plant.
3. Data and methods
Empirical studies examining economic growth have specified
models to include measures of human capital, institutional factors, pol-
icy related factors, and conditional convergence in addition to domestic
and foreign investment (recent examples include Azman-Saini et al.,
2010; Basu and Guariglia, 2007; Li and Liu, 2005). The current study
follows this approach in examining the relationship between FDI
flows and economic growth in 209 countries covering the period 1971
to 2010.3 Cross country data are averaged over five-year periods, as is
standard practice. To examine whether the FDI–growth relationship is
different in Pacific countries, the paper includes a FDI–Pacific interaction
term. The Pacific variable is a dummy variable taking the value of 1 if
the country is located in the Pacific and zero otherwise.4 The model is
specified as follows:
0
g it ¼ α 0 þ β1 P i þ β2 FDI it þ β3 FDI it  P it þ β4 Z it þ μ it i ¼ 1; …; n ð1Þ
where gi is a real growth in GDP per capita, Pi is the Pacific dummy
variable, FDIi is the percentage of FDI relative to GDP, FDIi ∗ Pi is the
FDI–Pacific multiplicative interaction term and Zi is a vector of control
variables. Subscript i represents recipient country and t represents
time. The vector of additional variables (Zi) contains the initial level of
GDP per capita (to capture convergence), the secondary school enrol-
ment rate (as a measure of human capital), the ratios of imports and
exports relative to GDP, the rate of inflation (logged), domestic
investment, a measure of economic freedom and the coefficient of
variation for the FDI variable. It also includes interactions of FDI with
other region dummy variables. The data include 42 observations for
the seven Pacific countries provided in Table 1. Data are sourced from
the World Bank's World Development Indicators (WDI) online database.
The exception is the economic freedom variable which is obtained from
the Freedom House database.
The choice of the explanatory variables included in the model war-
rants some discussion. This is especially the case with the FDI variable
3
The specific choice of the empirical model is motivated by maximising the number of
observations for PICs to be included in the analysis.
4
An alternative approach is to estimate models for Pacific countries only. This option is
not pursued by the current paper on the grounds of data availability and sample size. If an-
nual data are employed, there are just 95 FDI observations for Pacific countries over the
period 1971 to 2010. Moreover, there is a greater chance of estimating spurious regres-
sions using annual data for just Pacific countries and therefore a greater need to address
the time-series properties of the data. Time-series properties are less of a concern with da-
ta averaged within countries over five-year periods.
334 S. Feeny et al. / Economic Modelling 37 (2014) 332–339

Table 1 The level of freedom is an institutional variable included in the

Flows of FDI to selected Pacific countries.
Pacific country/country group Net inflows of FDI to GDP (%)
1980–1989 1990–1999
Fiji 2.13 3.24
Kiribati 0.14 9.42
Papua New Guinea 3.98 3.65
Samoa 0.14 3.09
Solomon Islands 1.33 2.92
Tonga 0.13 0.55
Vanuatu 5.33 11.34
Pacific average 1.88 4.89
Low and middle income countries 0.58 1.94
World 0.67 1.49
Notes: Data are from the World Bank (2011).
as it can be specified in a number of different ways. Firstly, FDI can be
measured either as a ‘flow’ or a ‘stock’. While the flow of FDI is based
on current account inflows of foreign capital for a year, stock measures
estimate the total cumulated value of foreign-owned capital in a coun-
try (Bornschier et al., 1978). Secondly, a choice must be made between
using FDI inflows or net FDI inflows (with FDI outflows subtracted).
The analysis in this paper focuses on net FDI inflows, as a percentage
of host country GDP. This approach is preferred over one examining the
stock of FDI on the grounds that the paper is concerned with the current
impacts of annual FDI inflows to the host country.5 Net FDI inflows
provide a measure of the additional capital available to the production
process of the host economy in a particular year (Alfaro et al., 2004;
Bornschier et al., 1978). Net FDI inflows include new investment from
foreign investors less disinvestment. Net FDI flows can therefore be neg-
ative if disinvestment exceeds new foreign investment in a country.
However, the regression sample includes just 41 observations that are
negative. The initial level of GDP per capita is included in the model to
capture convergence effects. This is standard practice in an empirical
growth model given the predictions of the Solow model that poorer
countries will tend to grow faster than richer ones (Barro, 1991;
Mankiw et al., 1992). Initial GDP per capita is measured in constant US
dollars and the variable is logged. If convergence holds, the coefficient
on the initial GDP per capita variable will be negative.
The importance of human capital to economic growth is highlighted
by Barro and Sala-i-Martin (1995) and Mankiw et al. (1992). In addition
to a direct impact on economic growth, human capital can be critical for
absorbing foreign knowledge and is an important determinant of
whether positive FDI spillovers will be realised. Human capital has
been measured in various ways by the literature. Average years of
schooling and school completion rates from Barro and Lee (2001) are
often favoured. The current study uses the secondary school enrolment
rate to capture human capital since these data are widely available for
the Pacific countries under consideration.
Trade variables are also included in the empirical model. Both
the ratio of imports and exports to GDP are included in the model.6 In
addition, the inflation rate is included as an indicator of macroeconomic
policy and stability. The variable is logged to capture the non-linearity in
the inflation–growth relationship.
model. Indices are constructed by Freedom House to measure freedom
in terms of two dimensions: political rights and civil liberties. The sur-
vey reports a numerical rating on a scale of 1 to 7 for each index, with
2000–2009
1 representing the highest and 7 the lowest level of freedom. In this
5.42
study, the variable Free is a dummy variable taking the value of 1 if
11.41
1.71 the country has an average rating of between 1.0 and 2.5 and zero
0.87 otherwise.
5.72 Pacific countries experience volatile levels of FDI in addition to vola-
2.77 tile growth rates, potentially hampering FDI's impact on economic
6.09
4.86
growth. To capture this volatility, the coefficient of variation (CV) of
2.79 FDI is employed as an explanatory variable. The CV is a statistical mea-
2.86 sure of the dispersion in a series around the mean. It is calculated as
the ratio of the standard deviation to the arithmetic mean of the vari-
able. Finally, while the initial level of per capita income is included as
an explanatory variable to capture convergence, a lagged dependent
variable (lagged growth) is included in some specifications to capture
any persistence and mean-reverting dynamics in economic growth
rates.
Data are averaged within countries over five-year periods. The esti-
mation sample includes all countries (developed and developing) for
which requisite data are available. This maximises sample size and min-
imises the chances of sample selection bias as well as allowing for an ex-
amination of whether the FDI–growth relationship differs for the Pacific
(as well as other regions). Full details of data sources and definitions are
reported in Table A1 of Appendix A. Summary statistics for the variables
are provided in Table A2, disaggregated into Pacific countries and other
countries. They reveal that average growth is lower in Pacific countries
relative to other countries although the level of FDI in these countries is
higher.7 The statistics also reveal the greater openness to trade of Pacific
countries but lower levels of human capital relative to other countries.
Eq. (1) is initially estimated using Ordinary Least Squares (OLS). Re-
sults from these estimations can be validly questioned due to concerns
over the endogeneity of the FDI (and other) variables. For example,
FDI will be endogenous if the economic growth rate of a host country
is an important factor for international firms when deciding where to
invest. In other words, FDI can determine and be determined by host
country growth rates. Studies, in response to these concerns, have
adopted different instrumentation strategies.
Some studies have sought to find valid external instruments for
the FDI variable. Unfortunately suitable instruments have been difficult
to identify.8 The favoured approach of the recent FDI literature
has therefore been to control for the potential endogeneity of the
FDI (and other) variables using internal instruments (see for example,
Basu and Guariglia, 2007; Bengoa and Sanchez-Robles, 2003;
Berthélemy and Démurger, 2000; De Vita and Kyaw, 2009). Dynamic
panel estimators such as ‘Difference’ and ‘System’ General Method of
Moments (GMM) have become popular estimation methods for panel
data with small T and large N. They are able to handle lagged dependent
variables; unobserved fixed effects; endogenous independent regres-
sors as well as heteroskedasticity and autocorrelation across and within
individuals or countries (Roodman, 2009).9
System GMM is preferred over difference GMM due to gains in
efficiency. In difference GMM all regressors are transformed through
differencing. Lagged levels of the regressors are used as instruments
for the first differenced regressors. However, for variables that closely

7
While there are many values of net FDI flows (relative to GDP) that are close to zero,
there are no observations which are actually zero implying that censoring/truncation is
5
Moreover, poor data availability for Pacific countries prevents the calculation of appro- not a pertinent issue for the analysis.
8
priate FDI ‘stock’ variables. Previous studies have used the distance that host countries are from major investing
6
The adoption of the Sachs–Warner trade openness variable often used in the literature countries (Egger, 2008; Guerin, 2006).
9
is not possible due to the limited observations for Pacific countries. See also Harris et al. (2008) for an overview of different dynamic panel models.
 S. Feeny et al. / Economic Modelling 37 (2014) 332–339
follow a random walk (such as economic growth), lagged levels can be
poor instruments for first differences. In response, Arellano and Bover
(1995) and Blundell and Bond (1998) develop system GMM by includ-
ing an equation in levels in addition to the differenced one. The equa-
tion specified in differences uses lagged levels of the regressors as
instruments and the equation in levels uses lagged differences of the
regressors as instruments. Efficiency is gained through the use of the
additional instruments. System GMM makes the assumption that
the differenced variables used as instruments are uncorrelated
with country fixed effects. Two-step robust estimation is employed
ensuring that standard errors are adjusted for autocorrelation and
heteroskedasticity.10
4. Results and interpretation
Results from OLS estimation are provided in columns (1) to (3) of
Table 2. The second column presents results when the model is
augmented with period (time) dummy variables. These period fixed-
effects capture time specific factors that determine growth, which are
not captured by the other explanatory variables. Results from column
(1) indicate that FDI has a positive association with host country
growth. The coefficient attached to the FDI variable suggests that a
10% increase in FDI is associated with (on average) about a 1% increase
in per capita income growth. In contrast, the coefficient attached to the
FDI–Pacific interaction variable is negative, large and statistically signif-
icant. Since the coefficient on the FDI–Pacific interaction is in absolute
terms larger than the coefficient on the FDI variable, the results imply
that FDI has a negative association with growth in the Pacific.
Other results from column (1) indicate that, as expected, the level of
exports and domestic investment have positive associations with
growth while imports, inflation and the volatility of FDI have negative
associations. The coefficients attached to these variables are all statisti-
cally significant (at the 10% level or greater). The coefficient attached
to the initial GDP per capita variable provides evidence of growth
convergence.11
Similar results are provided from OLS estimation of the model with
period effects in column (2). The period effects are (jointly) statistically
significant but the coefficients attached to the FDI and FDI–Pacific
interaction variables are very similar to those presented in column
(1). The coefficient attached to the human capital variable is also posi-
tive and statistically significant in this specification.
As discussed in Section 3 above, there are valid concerns over these
results due to the endogeneity of the FDI and other variables. Coefficient
estimates will be biased and inaccurate due to simultaneity bias. One
way to try and overcome this problem is to lag the potentially endoge-
nous variables. Results from this exercise are presented in column (3).
Interestingly, results remain largely unchanged. The coefficients
attached to the imports, exports and inflation variables are no longer
statistically significant. Results regarding FDI and its impact in the
Pacific are comparable to those reported earlier.
The results from OLS estimation capture relationships using both the
cross-country and time-series variation in the data. This paper proceeds
10
Note that country fixed effects are not explicitly included in system GMM estimation.
The system includes equations in levels and differences. They are purged in the differenced
equations and are not introduced in the level equations since they would introduce bias
(Roodman, 2009).
11
A model was estimated with a Pacific region dummy included and the FDI–Pacific re-
gion interaction term omitted. However the coefficient attached to the region dummy var-
iable was not statistically significant, implying that growth rates have not been different in
the Pacific than they have elsewhere. There is a high level of collinearity between the re-
gion dummy and the interaction term and including them in the model simultaneously
makes interpreting results very difficult.
335
with GMM estimation as a preferred method to account for the
endogeneity of the variables. System GMM purges country specific
effects and results are largely driven by the time-series variation of
the data within countries. However, system GMM includes equations
specified in levels as well as first differences and in this way a cross-
country dimension of the data is preserved.
Results from system GMM estimation of Eq. (1) are provided in
columns (4) to (7) of Table 2. In these specifications FDI, human capital,
imports, exports, inflation, domestic inflation and the FDI–Pacific
interaction variable are all treated as endogenous. Instrument lags
begin at t-2 and end at t-4. Further lags are unable to be employed
since the number of instruments should be less than the number of
country groups (Roodman, 2006). All GMM specifications pass
the Hansen J test for the validity of instruments (overidentifying restric-
tions) and the Arellano–Bond AR(2) test for autocorrelation.12 The
Arrellano–Bond test for autocorrelation is based on the difference resid-
uals. First order autocorrelation is therefore expected since differenced
error terms will share a common error term.13
In GMM estimation, coefficient estimates will be unbiased and
greater emphasis should be provided to these results. The results
provided in column (4) are broadly consistent with those from OLS.
Important differences, however, are the size of the coefficients attached
to the key variables. The coefficient attached to the FDI variable remains
positive but becomes larger and the coefficient attached to the FDI–
Pacific interaction variable remains negative but becomes smaller.
Given the size of the coefficients on these variables, these results
suggest that the impact of FDI in the Pacific is actually positive — but
is much smaller than for countries on average.
Since the coefficient attached to the freedom variable is not found to
be statistically significant in any of the models it is dropped in order to
obtain more observations. Similar results are reported in column (5).
The impact of FDI to the Pacific remains small and statistically signifi-
cant. Column (6) reports results from a model which examines whether
the impact of domestic investment on growth is lower in Pacific
countries. The coefficient attached to the domestic investment–Pacific
interaction variable is not statistically significant suggesting that it is
only the FDI–growth relationship which is different in this group of
countries. Finally a dynamic specification of the model is estimated
and results are presented in column (7). Unfortunately, the inclusion
of a lagged dependent variable reduces the number of observations
(including observations for Pacific countries). While results are broadly
consistent with those presented earlier, they suggest that the impact of
FDI on growth in the Pacific is negative but small.
In summary, results indicate that higher levels of FDI are associated
with higher growth rates. However, the impact of FDI is much lower in
the case of PICs.14
While the primary focus of this paper is the Pacific, it is also useful
to examine whether the FDI–growth relationship is different for
other regions of the world. Countries are categorised into the following
regions: Africa; Asia; Central and Eastern Europe; Latin America and the
Caribbean; and North America and Europe. Results from system GMM
12
The Hansen J statistic for over identifying restrictions is used and reported rather than
the Sargan statistic. The latter is not robust to heteroskedasticity or autocorrelation. The
Hansen J statistic indicates whether the instruments (as a group) are exogenous
(Roodman, 2009).
13
For example, D.ei,t = ei,t − ei,t−1 which will correlate with D.ei,t−1 = ei,t−1 − ei,t−2
since they share the common error term ei,t−1. Therefore to check for autocorrelation in
levels we test for second order autocorrelation in differences. The detection of autocorre-
lation would indicate that the variables used as instruments are bad instruments since
they are endogenous (Roodman, 2009).
14
The empirical models were re-estimated using data averaged within countries over
three and four-year periods. While some results changed, the core finding of the paper:
that FDI has a lower impact on growth in the Pacific relative to other countries prevailed.
Results are available from the authors upon request.
336 S. Feeny et al. / Economic Modelling 37 (2014) 332–339

Table 2
Econometric results.

Dependent variable: Growth in real GDP per capita (1) (2) (3) (4) (5) (6) (7)

OLS GMM

Constant 7.090⁎⁎⁎ 6.229⁎⁎⁎ 2.456⁎⁎ 11.984⁎⁎⁎ 12.062⁎⁎⁎ 11.922⁎⁎⁎ 11.461⁎⁎⁎

(1.133) (1.199) (1.212) (2.911) (2.985) (2.864) (3.238)
FDI 0.099⁎⁎ 0.094⁎⁎ 0.189⁎ 0.194⁎ 0.177⁎⁎ 0.155⁎⁎⁎
(0.043) (0.042) (0.102) (0.102) (0.088) (0.089)
Initial −0.576⁎⁎⁎ −0.569⁎⁎⁎ −0.439⁎⁎⁎ −1.324⁎⁎⁎ −1.419⁎⁎⁎ −1.450⁎⁎⁎ −1.352⁎⁎⁎
(0.135) (0.144) (0.147) (0.354) (0.361) (0.327) (0.363)
Human capital 0.010 0.013⁎ 0.037⁎ 0.041⁎⁎ 0.046⁎⁎ 0.039⁎
(0.007) (0.008) (0.021) (0.019) (0.018) (0.021)
Imports −0.052⁎⁎⁎ −0.046⁎⁎⁎ −0.069⁎⁎ −0.075⁎⁎ −0.075⁎⁎ −0.054⁎⁎
(0.013) (0.012) (0.033) (0.034) (0.030) (0.025)
Exports 0.047⁎⁎⁎ 0.044⁎⁎⁎ 0.098⁎⁎ 0.105⁎⁎⁎ 0.088⁎⁎ 0.081⁎⁎
(0.013) (0.013) (0.041) (0.040) (0.039) (0.033)
Inflation −1.419⁎⁎⁎ −1.284⁎⁎⁎ −1.868⁎⁎⁎ −1.903⁎⁎⁎ −1.918⁎⁎⁎ −1.613⁎⁎⁎
(0.220) (0.201) (0.430) (0.459) (0.434) (0.520)
Investment 0.155⁎⁎⁎ 0.151⁎⁎⁎ 0.104⁎ 0.116⁎⁎ 0.144⁎⁎⁎ 0.089⁎
(0.027) (0.028) (0.058) (0.059) (0.046) (0.053)
Freedom 0.183 0.139 0.203 0.128 0.063 0.270
(0.252) (0.248) (0.283) (0.331) (0.303) (0.339)
CV −0.005⁎⁎ −0.004⁎⁎ −0.006⁎⁎⁎ −0.010 −0.008 −0.007 −0.010
(0.002) (0.002) (0.002) (0.008) (0.008) (0.008) (0.008)
FDI ∗ Pacific −0.182⁎⁎ −0.174⁎⁎ −0.179⁎⁎ −0.157⁎ −0.182⁎⁎
(0.078) (0.081) (0.089) (0.089) (0.075)
Investment ∗ Pacific −0.043
(0.033)
Lagged growth 0.108⁎⁎
(0.052)
Lagged schooling 0.019⁎⁎⁎
(0.007)
Lagged imports −0.012
(0.010)
Lagged exports 0.010
(0.011)
Lagged inflation 0.216
(0.156)
Lagged investment 0.083⁎⁎⁎
(0.024)
Lagged FDI 0.171⁎⁎
(0.067)
Lagged FDI ∗ Pacific −0.380⁎⁎⁎
(0.083)
Period dummies No Yes Yes Yes Yes Yes Yes
R-squared 0.25 0.29 0.20
AR (1) (p-value) 0.00 0.00 0.00 0.00
AR (2) (p-value) 0.39 0.43 0.46 0.77
Hansen test (p-value) 0.37 0.31 0.50 0.61
N 967 967 831 967 980 957 881

Notes: Numbers in parentheses are robust standard errors.

⁎ Denotes significantly different from zero at the 90% confidence level.
⁎⁎ Denotes significantly different from zero at the 95% confidence level.
⁎⁎⁎ Denotes significantly different from zero at the 99% confidence level.

estimation, our preferred technique, are presented in Table A3 of
Appendix A. The approach necessitates the employment of interaction
variables in the model on an individual basis to avoid problems of collin-
earity between the key variables and the proliferation of instruments in
system GMM estimation. FDI is positively associated with growth in all
specifications expect when a FDI–Central and Eastern Europe region
interaction term is introduced. In this case the coefficients attached to
the FDI variables are no longer statistically significant. Other results sug-
gest that FDI has a larger impact on growth in Africa, relative to other
countries and a lower than average impact in North America and
Europe (although still positive).
There are number of explanations as to why the impact of FDI to the
Pacific is lower (on average) than in other parts of the world. Some of
these explanations might also apply to North America and Europe.
One explanation is that a lot of FDI has been concentrated in the
extractive industries with such investments being plagued with allega-
tions of corruption, disputes with local landowners and environmental
damage. Profits from these investments may also have been repatriated
overseas rather than re-invested in the Pacific. Another explanation is
that the returns to FDI in the Pacific are lower due to factors that reduce
the productivity and profitability of foreign investments in Pacific coun-
tries. These factors include a high cost and low availability of skilled la-
bour, a great distance to major markets as well as unfavourable business
environments.
The World Bank's ‘Doing Business’ project lends some support to this
latter assertion. The project provides objective measures of business
regulations and their enforcement in 183 economies. Each economy is
ranked according to their ease of doing business. A high ranking
means the regulatory environment is more conducive to the starting
and operation of a local firm. The best performing Pacific countries in
 S. Feeny et al. / Economic Modelling 37 (2014) 332–339 337

Table 3 indicate that FDI crowds-out domestic investment in the Pacific.

Econometric results of FDI effect on domestic investment. This helps explain the lower impact of FDI on growth in this region.
(1) (2) (3) Results from system GMM for other regions are reported in
Table A4 of Appendix A. They suggest that FDI ‘crowds in’ domestic
OLS FE GMM
investment (more than average) in Asia and in Latin America and
Constant 14.866⁎⁎⁎ 16.556⁎⁎⁎ 8.754⁎⁎⁎
the Caribbean but there is evidence of a crowding out effect in
(1.433) (2.411) (2.072)
FDI 0.389⁎⁎⁎ 0.494⁎⁎⁎ 0.309⁎⁎ North America and Europe. Taken together, these results suggest
(0.111) (0.114) (0.151) that it is important to account for heterogeneity across regions and
Exports 0.016 −0.017 −0.023 countries when exploring the impacts of FDI.
(0.011) (0.033) (0.026)
Government Expenditure 0.189⁎⁎⁎ 0.083 0.125
(0.039) (0.054) (0.083)
Inflation 0.604⁎ 0.502 0.115 5. Conclusion
(0.321) (0.359) (0.493)
Growth rate of GDP/cap 0.778⁎⁎⁎ 0.443⁎⁎⁎ 0.463⁎⁎⁎ The economic growth record of the Pacific region has been very
(0.087) (0.092) (0.108)
disappointing and partially as a result of this, poverty in the region is
Financial market 0.017⁎⁎⁎ 0.017 0.002
(0.005) (0.012) (0.005) increasing and progress towards the MDGs has been slow and in some
FDI ∗ Pacific −0.669⁎⁎⁎ −0.637⁎⁎ −0.328⁎ cases non-existent. With their limited resource bases, small domestic
(0.209) (0.281) (0.193) markets and lack of economies of scale, Pacific countries are heavily
Lagged investment 0.485⁎⁎⁎ reliant on assistance from other countries in the forms of foreign aid,
(0.100)
AR (1) (p-value) 0.003
trade, access to labour markets and FDI. This paper finds that the growth
AR (2) (p-value) 0.475 return from FDI to the region has (on average) been very small
Hansen test (p-value) 0.655 (although positive). A 10% increase in the ratio of FDI to host GDP is
R-squared 0.29 0.26 associated with higher growth of about 2% in host countries on average
Period dummies Yes Yes Yes
but just 0.1 to 0.4% in the case of the Pacific.
N 1036 1036 935
Given the (often) more positive assessments of aid and remittances
Notes: Numbers in parentheses are robust standard errors. to the region it might be tempting to conclude that donors should focus
⁎ Denotes significantly different from zero at the 90% confidence level.
⁎⁎ Denotes significantly different from zero at the 95% confidence level. their assistance to PICs through higher levels of Official Development
⁎⁎⁎ Denotes significantly different from zero at the 99% confidence level. Assistance (ODA) and greater access to their labour markets. While
the latter policy is certainly likely to provide high development returns,
calls for additional aid should be met with a great amount of caution.
There is a substantial literature which demonstrates that there are
diminishing returns to foreign aid owing mainly to limited absorptive
capacities within recipient countries. Past a certain level or threshold
2011 include Vanuatu, Samoa and Fiji with country ranks of 60, 61 and of aid, its growth impact starts to diminish. A number of Pacific coun-
62 respectively. Other Pacific countries perform less well including tries already receive some of the highest levels of aid in the world rela-
Kiribati, the Solomon Islands and Papua New Guinea, ranked 93, 96 tive to the size of their economies, and levels at which diminishing
and 103 respectively (World Bank, 2011). Unfortunately, a paucity of returns will have set in (Feeny and McGillivray, 2008). A key implication
this data for countries over time prevents an empirical analysis into is NOT for higher levels of aid to be provided to PICs to compensate for
this explanation. the relatively small impact of FDI on growth, unless absorptive capaci-
Another explanation is that FDI might have crowded out domestic in- ties can be quickly increased.
vestment and employment in the Pacific region. Data availability allows The international community must find other ways to assist PICs.
for such an investigation and is undertaken in the current paper. A With regard to FDI, two policy recommendations emerge from the find-
model of domestic investment is estimated with FDI being included as ings of this paper, based on the principle of attacking the development
an explanator together with a number of other variables including ex- problems of the region: (i) for developed (source) country govern-
ports, government spending, the rate of inflation, economic growth and ments to encourage firms to invest more in Pacific Island countries
financial market depth. The model is kept relatively parsimonious in without introducing market distortions; and (ii) for both host and
order to maximise the number of observations for Pacific countries. source country governments to find ways of increasing the impact of
Again, the FDI variable is interacted with a Pacific dummy variable to FDI on economic growth. Concerning the first recommendation,
examine whether any crowding in or crowding out effect is different Moran (2010) points to three areas in which developed country policies
for this region than it is in others. Results are presented in Table 3. are important for facilitating the flow of FDI to developing countries: the
Table 3 presents results from OLS (column 1), fixed effects provision of national or multilateral political risk insurance; the avoid-
(column 2) and system GMM estimation (column 3) which allows ance of double taxation of profits earned abroad; and finally, regulation
for the inclusion of a lagged dependent variable.15 Results are consis- to combat bribery and to prevent diversion of public revenues to private
tent across these estimation techniques.16 FDI is found to crowd-in pockets.
domestic investment in host countries (on average) but the sign Concerning the second recommendation, some general lessons
and size of the coefficient attached to the interaction variable from the FDI–growth literature are likely to be particularly applica-
ble for PICs. Limited absorptive capabilities and poor business envi-
ronments hamper the positive impacts of FDI on growth. They also
imply that domestic firms are less able to compete with foreign
ones and that FDI is more likely to displace domestic investment. A
greater donor focus on improving human capital and private sector
development in Pacific Island countries should therefore improve
15
The finding of FDI displacing domestic investment in the Pacific remains without the the productivity of investment and lead to FDI having a greater im-
inclusion of a lagged dependent variable. However, results from the Arellano–Bond test
provide evidence of second order autocorrelation in this specification.
pact on economic growth. The research community could also assist
16
In the system GMM estimations, all explanatory variables are treated as endogenous in this regard, by investigating ways of increasing the growth impact
with the exception of the financial market variable. Lags start at t-2 and end at t-4. of FDI in the Pacific.
338 S. Feeny et al. / Economic Modelling 37 (2014) 332–339

Appendix A

Table A1
Data description and sources.

Variable Description Source

Growth Growth in real GDP per capita (%) World Bank (2011)
FDI Net FDI inflows as a percentage of host country GDP World Bank (2011)
Initial GDP per capita Log level of GDP per capita (constant US$) World Bank (2011)
Human capital Secondary school enrolment rate (net) (%) World Bank (2011)
Imports Imports of goods and services as a percentage of GDP World Bank (2011)
Exports Exports of goods and services as a percentage of GDP World Bank (2011)
Inflation Annual growth rate of the GDP deflator World Bank (2011)
Investment Gross domestic investment as a percentage of GDP World Bank (2011)
Free Variable taking the value of 1 for countries with an average index score of between 1 and 2.5 for political rights and civil liberties Freedom House (2011)
CV The coefficient of variation for FDI defined as its standard deviation divided by its mean Authors' calculations
Financial market Domestic credit to the private sector as a percentage to GDP. Domestic credit to private sector refers to financial resources provided to World Bank (2011)
the private sector, such as through loans, purchases of no equity securities, and trade credits and other accounts receivable, which
establish a claim for repayment. For some countries these claims include credit to public enterprises.
Government expenditure General government final consumption expenditure as a percentage to GDP. It includes all government current expenditures for World Bank (2011)
purchases of goods and services (including compensation of employees) as well as most expenditure on national defense and security,
but excludes government military expenditures that are part of government capital formation.

Table A2
Summary statistics.

Variable Pacific countries Other countries

Mean SD Max Min Mean SD Max Min

Growth 1.135 2.540 6.032 −5.221 1.899 3.580 31.236 −18.645

FDI 4.174 4.948 24.798 −1.016 3.050 5.304 61.976 −20.921
Initial GDP capita 7.333 0.469 8.216 6.575 7.580 1.567 10.626 4.390
Imports 60.889 15.342 92.971 31.450 41.076 21.367 135.657 5.081
Exports 37.007 13.677 59.514 12.555 35.416 20.467 111.698 4.241
Human capital 56.115 33.675 105.18 9.022 61.238 33.300 155.579 0.374
Inflation 2.763 0.193 3.338 2.464 3.059 0.787 8.860 1.184
Investment 25.629 15.576 78.833 9.200 22.012 7.462 83.043 3.184
Free 0.595 0.497 1.000 0.000 0.427 0.495 1.000 0.000
CV 94.793 44.837 167.579 46.788 101.282 57.492 391.223 −139.618
Financial market 27.981 10.995 50.438 11.743 41.772 38.510 290.352 0.997
Government expenditure 23.242 7.857 39.906 11.087 16.413 6.469 50.862 2.727

Table A3
Econometric results of FDI on growth by region.

Dependent variable: Growth in real GDP per capita (1) (2) (3) (4) (5)

GMM

Africa Asia Central and Eastern Europe Latin America and the Caribbean North America and Europe

Constant 12.183⁎⁎⁎ 11.749⁎⁎⁎ 11.980⁎⁎ 12.113⁎⁎⁎ 10.993⁎⁎⁎

(2.433) (2.877) (3.370) (3.039) (2.725)
FDI 0.103⁎⁎ 0.170⁎ 0.118 0.172⁎⁎ 0.353⁎⁎⁎
(0.049) (0.092) (0.083) (0.086) (0.107)
Initial −1.339⁎⁎⁎ −1.352⁎⁎⁎ −1.243⁎⁎⁎ −1.436⁎⁎⁎ −1.100⁎⁎⁎
(0.330) (0.361) (0.432) (0.333) (0.345)
Human capital 0.051⁎⁎⁎ 0.039⁎ 0.029 0.041⁎⁎ 0.032⁎
(0.018) (0.022) (0.024) (0.020) (0.017)
Imports −0.069⁎⁎ −0.075⁎⁎ −0.072⁎⁎ −0.087⁎⁎ −0.072⁎⁎
(0.029) (0.033) (0.032) (0.034) (0.035)
Exports 0.077⁎⁎ 0.097⁎⁎ 0.104⁎⁎ 0.098⁎⁎ 0.101⁎⁎⁎
(0.030) (0.042) (0.042) (0.038) (0.038)
Inflation −1.931⁎⁎⁎ −1.818⁎⁎⁎ −1.959⁎⁎⁎ −1.814⁎⁎⁎ −1.987⁎⁎⁎
(0.416) (0.420) (0.473) (0.459) (0.421)
Investment 0.107⁎⁎ 0.127⁎⁎ 0.132⁎⁎ 0.154⁎⁎ 0.076⁎
(0.053) (0.060) (0.057) (0.063) (0.043)
Freedom 0.113 0.090 0.110 0.192 −0.097
(0.327) (0.326) (0.351) (0.346) (0.346)
CV −0.011 −0.010 −0.013 −0.009 −0.011
(0.009) (0.010) (0.011) (0.008) (0.009)
FDI ∗ regional variable 0.327⁎⁎⁎ 0.672 0.067 −0.142 −0.321⁎⁎⁎
(0.094) (0.707) (0.106) (0.089) (0.107)
Period dummies Yes Yes Yes Yes Yes
AR (1) (p-value) 0.00 0.00 0.00 0.00 0.00
AR (2) (p-value) 0.22 0.42 0.48 0.39 0.62
Hansen test (p-value) 0.25 0.41 0.20 0.28 0.22
N 967 967 967 967 967

Notes: Numbers in parentheses are robust standard errors.

⁎ Denotes significantly different from zero at the 90% confidence level.
⁎⁎ Denotes significantly different from zero at the 95% confidence level.
⁎⁎⁎ Denotes significantly different from zero at the 99% confidence level.
 S. Feeny et al. / Economic Modelling 37 (2014) 332–339 339

Table A4
Econometric results of FDI on domestic investment by region.

Dependent variable: Investment to GDP (1) (2) (3) (4) (5)

GMM

Africa Asia Central and Eastern Europe Latin America and the Caribbean North America and Europe

Constant 9.124⁎⁎⁎ 7.755⁎⁎⁎ 8.981⁎⁎⁎ 4.738⁎ 9.128⁎⁎⁎

(2.331) (2.222) (2.198) (2.452) (1.931)
FDI 0.269⁎ 0.309⁎ 0.313⁎ 0.188 0.628⁎⁎⁎
(0.160) (0.172) (0.171) (0.153) (0.144)
Exports −0.039 −0.016 −0.023 −0.022 −0.014
(0.027) (0.027) (0.028) (0.025) (0.026)
Government expenditure 0.161⁎⁎ 0.193⁎⁎ 0.175⁎⁎ 0.227⁎⁎ 0.181⁎⁎
(0.080) (0.082) (0.089) (0.089) (0.075)
Inflation 0.066 0.161 −0.108 0.662 −0.277
(0.475) (0.499) (0.472) (0.527) (0.404)
Growth rate of GDP/cap 0.521⁎⁎⁎ 0.511⁎⁎⁎ 0.444⁎⁎⁎ 0.628⁎⁎⁎ 0.352⁎⁎⁎
(0.117) (0.113) (0.124) (0.131) (0.119)
Financial market 0.007 0.001 0.0001 0.0001 0.013⁎
(0.007) (0.006) (0.006) (0.006) (0.008)
FDI ∗ regional variable 0.186 2.070⁎⁎ 0.033 0.353⁎⁎ −0.636⁎⁎⁎
(0.180) (0.937) (0.149) (0.147) (0.150)
Lagged investment 0.461⁎⁎⁎ 0.451⁎⁎⁎ 0.471⁎⁎⁎ 0.523⁎⁎⁎ 0.372⁎⁎⁎
(0.109) (0.104) (0.108) (0.092) (0.090)
AR (1) (p-value) 0.002 0.005 0.000 0.001 0.003
AR (2) (p-value) 0.402 0.415 0.264 0.473 0.429
Hansen test (p-value) 0.436 0.587 0.264 0.271 0.340
Time dummies Yes Yes Yes Yes Yes
N 935 935 935 935 935

Notes: Numbers in parentheses are robust standard errors.

⁎ Denotes significantly different from zero at the 90% confidence level.
⁎⁎ Denotes significantly different from zero at the 95% confidence level.
⁎⁎⁎ Denotes significantly different from zero at the 99% confidence level.

References
Alfaro, L., Chanda, A., Kalemli-Ozcan, S., Sayek, S., 2004. FDI and economic growth: the
role of local financial markets. J. Int. Econ. 64 (1), 89–112.
Arellano, M., Bover, O., 1995. Another look at the instrumental variable estimation of
error-components models. J. Econ. 68, 29–51.
AusAID, 2008. Pacific Economic Survey 2008. Australian Agency for International Devel-
opment, Canberra.
AusAID, 2009. Tracking Development and Governance in the Pacific. Australian Agency
for International Development, Canberra.
Azman-Saini, W.N.W., Baharumshah, A.Z., Law, S.H., 2010. Foreign direct investment, eco-
nomic freedom and economic growth: international evidence. Econ. Model. 27 (5),
1079–1089.
Barro, R., 1991. Economic growth in across section of countries. Q. J. Econ. 106 (2),
407–443.
Barro, R.J., Lee, J.W., 2001. International data on educational attainment: Updates and im-
plications. Oxf. Econ. Pap. 53 (3), 541–563.
Barro, R.J., Sala-i-Martin, X., 1995. Economic Growth. McGraw-Hill, Cambridge, MA.
Basu, P., Guariglia, A., 2007. Foreign direct investment, inequality, and growth.
J. Macroecon. 29, 824–839.
Bengoa, M., Sanchez-Robles, B., 2003. Foreign direct investment, economic freedom and
growth: new evidence from Latin America. Eur. J. Polit. Econ. 19, 529–545.
Berthélemy, J.-C., Démurger, S., 2000. Foreign direct investment and economic growth:
theory and application to China. Rev. Dev. Econ. 4 (2), 140–155.
Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel
data model. J. Econ. 87 (1), 115–143.
Bornschier, V., Chase-Dunn, C., Rubinson, R., 1978. Cross-national evidence of the effects
of foreign investment and aid on economic growth and inequality: a survey of find-
ings and reanalysis. Am. J. Sociol. 84 (3), 651–683.
Brown, R.P.C., 2008. Remittances and Development in the Pacific: Effects on Human De-
velopment in Fiji and Tonga, United Nations Expert Group Meeting On International
Migration and Development in Asia and the Pacific. United Nations Economic and So-
cial Commission for Asia and the Pacific, Bangkok, Thailand.
Browne, C., Mineshema, A., 2007. Remittances in the Pacific region. IMF Working Paper
WP/07/35.International Monetary Fund, Washington.
Connell, J., Brown, R.P.C., 2005. Remittances in the Pacific: An Overview. Asian Develop-
ment Bank, Manila.
de Mello, L.R., 1997. Foreign direct investment in developing countries and growth: a se-
lective survey. J. Dev. Stud. 34 (1), 1–34.
De Vita, G., Kyaw, S., 2009. Growth effects of FDI and portfolio investment flows to devel-
oping countries: a disaggregated analysis by income levels. Appl. Econ. Lett. 16 (3),
277–283.
Egger, P., 2008. On the role of distance of outward FDI. Ann. Reg. Sci. 42, 375–389.
Feeny, S., 2007. Impacts of foreign aid to Melanesia. J. Asia Pac. Econ. 12 (1), 34–60.
Feeny, S., McGillivray, M., 2008. Do Pacific countries receive too much foreign aid? Pac.
Econ. Bull. 23 (2), 166–178.
Feeny, S., McGillivray, M., 2010. Aid and growth in Small Island Developing States. J. Dev.
Stud. 46 (5), 897–917.
Freedom House, 2011. Freedom in the World: 1972–2010. Freedom House.
Guerin, S.S., 2006. The role of geography in financial and economic integration: a compar-
ative analysis of foreign direct investment, trade and portfolio investment flows.
World Econ. 29 (2), 189–209.
Harris, M.N., Mátyás, L., Sevestre, P., 2008. In: Mátyás, L., Sevestre, P. (Eds.), The Econo-
metrics of Panel Data. Springer, Netherlands.
Jayaraman, T.K., Singh, B., 2007. Impact of foreign direct investment on employment in
Pacific Island countries: an empirical study of Fiji. Econ. Int. 60 (1), 57–74.
Li, X., Liu, X., 2005. Foreign direct investment and economic growth: an increasingly en-
dogenous relationship. World Dev. 33 (3), 393–407.
Lucas, R.E., 1988. On the mechanics of economic development. J. Monet. Econ. 22 (1),
3–42.
Mankiw, G.N., Romer, D., Weil, D.N., 1992. A contribution to the empirics of economic
growth. Q. J. Econ. 107 (2), 407–437.
McGillivray, M., Naudé, W., Santos-Paulino, A.U., 2010. Vulnerability, trade, financial flows
and state failure in small developing states. J. Dev. Stud. 46 (5), 815–827.
Moran, T., 2010. Assessing Developed Country Efforts to Support Developing Country
Growth via Foreign Direct Investment, Background Paper and Explanation of Scoring
System for the 2010 Commitment to Development Index. Center for Global Develop-
ment, Washington.
Nathan Associates Inc., 2007. Pacific Regional Trade and Economic Cooperation Joint
Baseline and Gap Analysis, Report Prepared for the Pacific Islands Forum
Secretariat.
Pavlov, V., Sugden, C., 2006. Aid and growth in the Pacific Islands. Asian Pac. Econ. Lit. 20
(2), 38–55.
Read, R., 2007. Foreign Direct Investment in Small Island Developing States, UNU-WIDER
Research Paper No. 2007/28. United Nations University World Institute for Develop-
ment Economics Research, Helsinki.
Rebelo, S., 1991. Long-run policy analysis and long-run growth. J. Polit. Econ. 99 (3),
500–521.
Romer, P.M., 1986. Increasing returns and long run growth. J. Polit. Econ. 94,
1002–1037.
Roodman, D., 2006. How to do xtabond2: an introduction to “difference” and “system”
GMM in Stata. Center for Global Development Working Paper Number 103.
Roodman, D., 2009. How to do xtabond2: an introduction to “difference” and “system”
GMM in Stata. Stata J. 9 (1), 86–136.
Solow, R.M., 1957. Technical change and the aggregate production. Rev. Econ. Stat. 39,
312–320.
World Bank, 2011. World Development Indicators Online Database. World Bank,
Washington.