Policy Research Working Paper 6750 Economic Growth in Ghana Determinants and Prospect Anna K. Raggl e World Bank Africa Region Poverty Reduction and Economic Management Department January 2014 WPS6750 Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized
26
Embed
Economic Growth in Ghana - World Bank · Policy Research Working Paper 6750 Economic Growth in Ghana Determinants and Prospect Anna K. Raggl The World Bank Africa Region Poverty Reduction
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Policy Research Working Paper 6750
Economic Growth in Ghana
Determinants and Prospect
Anna K. Raggl
The World BankAfrica RegionPoverty Reduction and Economic Management DepartmentJanuary 2014
WPS6750P
ublic
Dis
clos
ure
Aut
horiz
edP
ublic
Dis
clos
ure
Aut
horiz
edP
ublic
Dis
clos
ure
Aut
horiz
edP
ublic
Dis
clos
ure
Aut
horiz
edP
ublic
Dis
clos
ure
Aut
horiz
edP
ublic
Dis
clos
ure
Aut
horiz
edP
ublic
Dis
clos
ure
Aut
horiz
edP
ublic
Dis
clos
ure
Aut
horiz
ed
Produced by the Research Support Team
Abstract
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.
Policy Research Working Paper 6750
This paper employs a simple cross-country panel framework to assess the determinants of growth in Ghana’s gross domestic product over the past four decades. A set of standard covariates is used to explain growth rates. Natural resource variables are included because the effects of natural resource rents in gross domestic products are of particular interest for Ghana. Using the preferred specification, Ghana’s growth potential is predicted for the upcoming decades under different scenarios. The results indicate that under the most pessimistic scenario of no improvements in the
This paper is a product of the Poverty Reduction and Economic Management Department, Africa Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at [email protected].
determinants of growth compared with the period 2005-09, Ghana’s gross domestic product per capita growth rates will stagnate at approximately 4.5 percent during the next decade and decrease thereafter. If the policy measures and country characteristics improve in the way they did in the past three decades, average per capita growth rates of roughly 5.5 percent could be reached during 2015–34. Taking into account the expected oil production until 2034 adds 0.6 percentage points to projected gross domestic product growth rates on average.
Economic Growth in Ghana: Determinants and Prospects∗
∗This report serves as a background study for the World Bank’s Policy Note on Long Run Growth in Ghana(2014)†WiC - Wittgenstein Centre for Demography and Global Human Capital, WU - Vienna University of Economics
and Business. Address: Welthandelsplatz 1, 1020 Vienna, Austria. Email: [email protected]. The authorgratefully acknowledges valuable contributions from Jesus Crespo Cuaresma, Leonardo Garrido, Santiago Herreraand Mathias Moser.
1 Introduction
Ghana’s GDP per capita growth rates increased steadily in the past two decades and varied around
5-7% between 2007 and 2010 (see Figure 1). Over the same period, poverty rates decreased and
other, non-monetary, welfare indicators improved. In order to assess the sustainability of these
positive developments, their determinants need to be detected and quantified. This investigation
of historic growth rates can further be used for predicting future GDP growth rates under different
scenarios.
This piece of research uses panel growth regressions to address the question about the drivers
of Ghana’s historic growth rates, emphasizing in particular the role of natural resources. Cocoa,
timber, minerals and gold account for a large share of total exports of the country (see Lejarraga
(2010), for example), and the recent discovery of oil raises the question about the expected impact
of natural resources on growth. The estimates obtained from the cross-country panel analysis are
used to predict the growth potential for the country under different scenarios.
-.05
0.0
5.1
5-ye
ar-a
vera
ge G
DP
per
cap
ita g
row
th r
ate
1970 1980 1990 2000 2010Year
Figure 1 – Growth rate of per capita GDP in Ghana, averages over 5 years: Data from 1970-2010 are used.Values are 5-year averages for the years 1970-2006, and averages of 1-4 years for values thatcorrespond to the years 2007 and after.
The choice of variables that are used in the empirical analysis is based to a large extent on the
findings in Barro (2003), but additionally considers the aim of assessing the growth performance
of the country of Ghana by estimating various specifications that tackle the impact of natural
resources.
2
The dataset covers the period 1970-2009 and contains economic, political and institutional
characteristics of 151 countries, thereby extending the analysis in Barro (2003) by roughly 15 years
and 80 countries. We estimate different specifications of growth regressions controlling for fixed
country and period effects, trying to explain a large share of within-country variations in growth
rates.
The findings suggest that education significantly increases GDP per capita growth rates, an
effect that is larger in countries with a relatively low initial GDP per capita. Furthermore, invest-
ment, openness to trade as well as natural resources are found to be robust determinants of income
growth, all related positively to growth. Government consumption seems to decrease GDP growth
rates in the long run. Relating institutional characteristics of countries to the estimated country-
specific fixed effects shows that there exists a strong positive correlation between them. Although
no causal interpretations can be made in this context, the finding supports the supposition that
fixed effects capture the persistent part of the institutional environment and an improvement of
the quality of institutions is favorable for a country and translates into higher GDP per capita
growth.
Based on the performed growth regressions, we predict Ghana’s future growth potential for
different scenarios. In a baseline scenario, where investment, natural resources, government expen-
ditures and other covariates are assumed to remain at the level of 2005-09, we find that average
growth rates are slightly above 4% in the upcoming two decades. In a more favorable scenario,
where it is assumed that the covariates behave as they did since the 1980s, predicted growth rates
increase by about 1%-point as compared to the baseline scenario. Finally, when taking into account
the expected oil production in Ghana, GDP growth is likely to reach 6.2-6.5% in the next 15-20
years.
The paper is organized as follows. Before the results are discussed in Section 5, the model
specification is described in Section 2 and the econometric methodology and the data are described
in Sections 3 and 4. Section 6 applies the results to predict GDP per capita growth rates for Ghana
until 2030 and Section 7 summarizes and concludes.
2 Model Specification
The specifications of the growth regressions are based on the findings in Barro (2003), and aug-
mented by variables that are of particular interest for Ghana.
The dependent variable is the arithmetic average of the real GDP per capita growth rate of a
country over each 5-year period during 1970-2009. The basic specification includes country fixed
3
effects and time dummy variables and a set of variables that is explained in the following. In order
to control for (conditional) convergence effects, the (natural logarithm of the) level of GDP per
capita at the beginning of each 5-year period is included in the regression. The parameter estimate
associated with this variable is expected to be negative, as relatively low levels of GDP per capita
increase growth rates (holding all other factors constant). Its magnitude can be interpreted as the
speed of (conditional) convergence to a country-specific equilibrium growth rate. Human capital
enters as the share of the (female and male) working age population with tertiary education. In
order to alleviate a potential endogeneity bias due to simultaneity, the value at the start of each
period is used.
A further set of basic covariates are policy related variables and national characteristics. A vari-
able for international openness, measured as imports plus exports as shares of GDP, is included
to assess the effects of trade and globalization. As trade shares are highly correlated with country
sizes, Barro (2003) suggests adjusting the openness measure for this relation. More specifically, the
openness variable is regressed on the logarithms of the population and area (in square-kilometers)
of the countries, and only the remaining part of the measure is used in the growth regression. The
adjusted openness to trade of the countries enters each regression as an average over the 5-year
periods for each country. Additionally, 5-year averages of inflation, government consumption1 and
investment, the latter two as shares in total GDP, are added to the basic regression model. Finally,
as a proxy for institutional quality, a democracy index is included.
The oil production expected in Ghana during the upcoming decades and the country’s depen-
dency on cocoa, timber, gold and minerals raise the question about the growth effects of natural
resources in general, and oil production in particular. We address this issue by including different
variations of natural resource variables in the panel growth regressions. First, the 5-year averages of
natural resource rents as shares of GDP are used as a regressor. This variable combines rents from
oil, natural gas, coal, minerals and forest. As this variable does not allow to isolate an effect of oil
production per se, the measure is split into rents obtained from oil production and non-oil natural
resource rents in a subsequent specification. The coefficient of the variable that measures oil rents
as shares in GDP might still hide some heterogeneities. For that reason we further decompose the
variable by interacting the oil rents with dummy variables indicating the relative importance of oil
production as compared to other economies. This allows the effects of oil rents on GDP growth to
differ across countries with varying oil intensities. The results of this final specification are used
to perform in- and out-of-sample predictions of Ghana’s growth rates.
1Barro (2003) suggests subtracting expenditures for defense and education from government consumption, asthese should be considered as investments. Data limitations do not allow us to perform a similar adjustment.
4
The data sources of the variables described above can be found in Section 4, and Table 1 briefly
summarizes the variables and their respective sources.
In the context of economic growth and especially in interaction with the role of natural resources,
it would be of particular interest to study the impact of the quality of institutions of the countries.
Most institutional indicators, such as those proxying the rule of law, government effectiveness or
political stability, are not available for a sufficient number of countries and years to include them
in the growth regressions. Additionally, institutional characteristics are very persistent, which
complicates the identification of the parameters in a fixed effects framework. To the extent that
institutional variables are constant over the period 1970-2009, they are captured by the fixed
country effects and only significant yearly deviations from the long-run country means would allow
including them in the estimation. The advantage of the persistent nature of institutions is that fixed
country effects control for general differences in the quality of institutions across countries, even
when the indicators are not available for most of the years considered. This approach, however,
does not allow us to isolate the effect of institutions on GDP growth, but the risk for a potential
omitted variable bias is reduced. In an extension of the analysis below, we assess the correlation of
the estimated fixed effects and a number of institutional quality indicators (measured in a recent
year, to maximize the sample for the analysis) in an attempt to explain parts of the variation of
fixed effects across countries.
3 Econometric Methodology
We estimate the specifications discussed above controlling for fixed country effects and including
period dummy variables. The period dummies control for effects that are specific for a certain
period and have an impact on all countries in a given period. The country specific fixed effects
allow controlling for unobserved heterogeneity as they capture effects that are characteristic for a
country and that do not vary over time, such as most geographical factors, colonial linkages, or
cultural factors inherent to a country. The conclusions we can draw from this analysis are within-
country effects, and do not reflect between country effects.
In particular, we estimate a model of the following form
yit = α+
K∑k=1
βkxkit + µi + δt + εit (1)
where yit is the GDP per capita growth rate of country i in year t, α is a constant, xit are
the K explanatory variables corresponding to country i and period t, µi are country-specific fixed
effects, δt are time fixed effects and εit is an error term.
5
The explanatory variables vary by specification, but time and country fixed effects are in-
cluded in every specification, so are the log of the initial GDP, education, investment, government
consumption, inflation, democracy, trade openness and some measure of natural resource rents.
4 Data
Table 1 – Description of variables
Variable Description Source
Dependent variable Growth in per capita GDP, 5-year averages PWT
Log of initial GDP Log of initial GDP per capita measured at start of each period PWT
Education Share of tertiary educated in working age population IIASA
Log initial GDP * Education Interaction between education variable and initial GDP
Openness Share of exports plus imports in GDP, filtered for its relation tolog(area) and log(population), 5-year average
PWT
Investment Share of investment in GDP, 5-year average PWT
Government consumption Government consumption as share of GDP, 5-year average PWT
Democracy Democracy index (0 least, 10 most democratic), lagged by 5 years Polity IV
Inflation Inflation rates, 5-year average WDI
Natural resource rents Natural resource rents as share of GDP, 5-year averages WDI
Oil rents Oil rents as share of GDP, 5-year averages WDI
Mineral rents Mineral rents as share of GDP, 5-year averages WDI
Forest rents Forest rents as share of GDP, 5-year averages WDI
Non-oil rents Natural resource rents excluding oil as share of GDP, 5-year averages WDI
Oil rents, 1st quartile Oil rents if oil rents belong to the lowest quartile in correspondingperiod, 0 otherwise (i.e. interaction of the oil rents variable with adummy variable indicating the first quartile)
Oil rents, 2nd quartile Oil rents if oil rents belong to the 2nd quartile in corresponding period,0 otherwise
Oil rents, 3rd quartile Oil rents if oil rents belong to the 3rd quartile in corresponding period,0 otherwise
Oil rents, 4th quartile Oil rents if oil rents belong to the highest quartile in correspondingperiod, 0 otherwiseNote: The Variables oilrent 0025, oilrent 2550, oilrent 5075 andoilrent 7500 add up to the variable oilrent
PWT corresponds to the Penn World Tables (Heston, Summers, and Aten, 2012)
IIASA corresponds to the IIASA-VID dateset on educational attainment (Lutz, Goujon, K.C., and Sanderson, 2007)
Polity IV corresponds to the Polity IV dataset obtained from Teorell, Charron, Samanni, Holmberg, and Rothstein (2011)
WDI corresponds to the World Development Indicators (2013)
6
We use a dataset of 117-151 countries spanning the period 1970-2009. The four decades are
split into 8 time periods, each covering a 5-year window. Data availability does not allow to observe
each country in each period, the resulting database has the structure of an unbalanced panel and
the numbers of countries considered depends of the empirical specification. Most of the variables
enter as averages over the 5 years or initial values measured at the start of each period are used. Ta-
ble A.11 in the Appendix provides an overview of the countries and periods included in the analyses.
The dataset combines data from various sources. Per capita GDP (PPP converted, at constant
2005 prices) is taken from the Penn World Table (PWT) Version 7.1 (Heston, Summers, and Aten,
2012). Growth rates of per capita GDP are computed as log-differences of that variable. The value
of the logarithm of GDP per capita at the start of each period is used as a right hand side variable.
Government consumption, investment as well as the measure for openness and inflation are ob-
tained from the same source. Data on educational attainment are taken from an updated version of
the IIASA/VID Dataset on Educational Attainment (Lutz, Goujon, K.C., and Sanderson, 2007).
To control for the level of democracy of a country, an index from the Quality of Government Stan-
dard Data collection (Teorell, Charron, Samanni, Holmberg, and Rothstein, 2011) is used, and in
order to extend the time coverage we additionally use data on democracy provided by Bollen (1990).
A summary and brief description of the variables used can be found in Table 1.
5 Results
Column 1 of Table 2 shows the first baseline regression including the initial GDP per capita, hu-
man capital as well as an index for democracy, the openness ratio, government consumption and
investment as shares in GDP, inflation and 7 period dummies, where the first period (1970-74)
constitutes the reference category. The negative and significant coefficient of the lagged GDP
per capita variable indicates that countries with a relatively low initial income grow faster towards
their country specific equilibrium. The institutional variable, an index of democracy, does not show
significant effects in this specification. Likewise, education is not identified as a significant driver of
growth rates. This is a common problem in cross country growth regressions and often attributed
to differences in the quality of education across countries and over time (Hanushek and Wossmann,
2008), the lack of including the demographic structure of educational attainment (Lutz, Cuaresma,
and Sanderson, 2008), omitting (in)equality measures of schooling within a country (Sauer and Za-
gler, 2012) or heterogeneous effects across countries or over time. The specification in Column 1
assumes a linear relationship between education and growth and it does not allow the impact of
human capital to differ across countries. In the following, an interaction of the initial GDP per
7
capita and the education measure is included to address this issue.
Increases in government consumption lower growth rates (in long run perspective of this anal-
ysis), so do increases in inflation. Investment and the openness ratio of a country foster growth
significantly. These results are in line with Barro (2003).
As an addition to the set of variables suggested by Barro (2003), an aggregate measure of natural
resource rents is included in specification 1 and shows a positive impact of GDP growth rates of
the countries included. A disaggregation of this measure will shed further light on the impact of
natural resources on growth rates.
Additionally to the variables in Column 1, Column 2 of Table 2 includes an interaction of
education and the initial level of GDP. The findings show that the effect of human capital on
economic growth depends on the level of development of the country. The lower the initial income
level of a country, the higher the gains from a better educated working age population. For the
case of Ghana, the effects of the historic increases in education on per capita GDP growth rates are
plotted in Figure 22. The steady increase in the tertiary education of the working age population
between 1970 and 2009 contributes by roughly 0.5-0.9%-points to yearly GDP per capita growth
rates.
In the specifications 3-5 we are focusing on a decomposition of the natural capital variable.
The variable combines rents obtained from oil, natural gas, minerals, coal and forests. Splitting
the variable into rents from oil production and other natural capital rents (column 3 of Table 2)
suggests that while non-oil natural capital rents increase GDP growth, oil rents show no significant
impact on output growth. This however could be due to the heterogeneous nature of oil exporting
countries. While many countries obtain oil rents worth less than 1% of their GDP, some countries’
economy is more dependent on oil production and these might react to changes in oil rents in a
different way, an issue that is further addressed below.
Splitting the non-oil natural resource variable further into rents from minerals and forests (as
these two resources are of particular interest for Ghana), we find that minerals, as opposed to rents
from forests, increase GDP growth rates (Column 4). It should be noted, that the coefficients of
the other covariates do not change considerably when altering the natural resource variable.
Column 5 allows the impact of oil rents on GDP growth to depend on the relative size of the
oil rents in GDP. In order to implement that, the 25th, 50th and 75th percentile of the oil rent
2As specification 2 in Table 2 is not the preferred specification, the estimates shown in the figure are based onthe final specification in Column 5.
8
Table 2 – Fixed effects estimations
(1) (2) (3) (4) (5)
Log of initial GDP -0.0410*** -0.0386*** -0.0415*** -0.0442*** -0.0444***[0.000] [0.000] [0.000] [0.000] [0.000]
Each specification includes country fixed effects and period dummy variables
9
.03
.04
.05
.06
.07
Sha
re te
rtia
ry e
duca
tion
in W
AP
.006
.007
.008
.009
.01
Con
trib
utio
n of
edu
catio
n to
gro
wth
1970 1980 1990 2000 2010Year
Contribution of education to growthShare tertiary education in WAP
Figure 2 – Share of tertiary education since 1970 and its contribution to GDP per capita growth rates inGhana (based on Specification (5) in Table 2)
variable for each period is determined. In each period, the countries are then assigned to one of
the following four groups: oil production below the 25th percentile, between the 25th and the 50th
percentile, between the 50th and the 75th percentile and above the 75th percentile. The countries’
allocation to a group must not be constant across periods, i.e. a country can belong for example
to the top quarter in one period, and to the 3rd in another. Including these four oil rent variables,
each one containing the rents of the relevant quartile, we allow the impact of oil rents to change
with the level of oil production. The results show that the positive impact seems to decrease with
an increase in the share of oil rents in GDP.3 Countries whose oil production is relatively low
compared to other oil producing countries tend to gain more from it than those whose oil rents in
GDP are comparably high, keeping everything else constant. An explanation of this result cannot
be assessed in this regression analysis, but a possible reason could be that the negative effects
often associated with natural resources (see for instance Sachs and Warner 1995, 2001) are less
accentuated when the natural resources account only for a small part of total GDP. Crowding-
out effects, strong price-dependence and low competitiveness in other sectors that suppress exports
3In another specification that is not included in the table, the oil rents variables is interacted with initial incomelevels. Interactions of the oil variables with variables that capture the quality of the institutions of a country wouldbe preferable, but we lack sufficient data on that. The results indicate that countries with low initial income levelstend to benefit more from oil production. However, the effects are insignificant, except for the 4th quartile.
10
are less likely to happen when oil production is relatively low and not a central part of an economy.
Correlation coefficients between estimated fixed effects and indicators of institutional quality (measured in 2000) as well
as among institution indicators; 119 countries included.
The correlation coefficients vary between 0.79 (voice and accountability and government ef-
fectiveness) and 0.97 (rule of law and government effectiveness), showing the strong positive re-
lationship between those indicators. More importantly, as shown in column 2 (Fixed Effects) in
Table 4, all institutional indicators are highly positively correlated with the country-fixed effects.
The graphs in Figure 4 further confirm this finding. A high value of the rule of law indicator is
associated with high country-specific fixed effects. A similar correlation is found for government
effectiveness, and for the indicators for political stability and corruption, where the latter two in-
dicators are set in relation to the fixed effects in Figure A.8 in the Appendix. The strong positive
relationship between the institutional quality indicators and the fixed effects thus strengthens the
supposition that a considerable part of the country effects is determined by the quality of institu-
tions.
It is apparent in both graphs that Ghana lies below the regression line that represents the av-
erage relationship between the institutional indicators and the fixed effects for the countries under
consideration. Based solely on the rule of law index, for example, one could expect a fixed effect for
Ghana that is similar to the ones estimated for Egypt, Tunisia or Argentina (based on the values
corresponding to the year 2000). This can be a hint for other negative country characteristics that
13
partially offset the positive impact of institutions. The same finding can be observed when looking
at the government effectiveness indicator (Figure 4, right part), corruption and political stability
(Figure A.8 in the Appendix). It also appears that Ghana is not the only country in Sub-Saharan
Africa that is positioned below the regression line. In fact, most of the Sub-Saharan African coun-
tries exhibit lower fixed effects than one would expect looking at the average correlation between
institutional quality and fixed effects.
CMR
MOZTGO
ETH
KEN
COG NGA
BEN
TZA
GHA
GNQ
GAB
ZMBZWE
SENCIV
TCD
GHA
-.1
-.05
0.0
5.1
.15
Fix
ed E
ffect
s
-2 -1 0 1 2Rule of Law
CMR
MOZTGO
ETH
KEN
COG NGA
BEN
TZA
GHA
GNQ
GAB
ZMBZWE
SENCIV
TCD
GHA
-.1
-.05
0.0
5.1
.15
Fix
ed E
ffect
s
-2 -1 0 1 2Government Effectiveness
Figure 4 – Correlation between fixed effects and institutional characteristics
It is of crucial importance, however, not to oversimplify the relationship between institutional
characteristics and the fixed country effects and not to over-interpret the findings above. The
numbers in Table 4, as well as Figures 4 and A.8, show simple correlations, and must not be
interpreted as causal effects.
6 Implications for Ghana
Based on the specification shown in Column 5, in Table 2 we assess Ghana’s growth potential for
the upcoming two decades. The high degree of uncertainty we accommodate by assuming different
scenarios which will be discussed in detail below. Figure 5 shows how the model predicts Ghana’s
past GDP per capita growth rates. Comparing 5-year averages of the actual growth rates with the
ones predicted by the model (denoted as fitted growth rates in the figure) shows that the preferred
specification predicts the large decline in growth in the 1970s and early 1980s, as well as the rise
in the 1990s and 2000s. As we assess long term growth rates, we use 5 year averages in order
to average out short term fluctuations, and for that reason our sample ends in 2009. The last
period we use actual data is thus 2005-09, the growth rates corresponding to the period 2010-14
are already projections.
14
-.04
-.02
0.0
2.0
4.0
6
1960 1980 2000 2020 2040year
GDP per capita growthFitted GDP per capita growthScenario 1Scenario 2Scenario 3
Figure 5 – Actual and fitted GDP per capita growth rates (1970-2009) and predictions under different sce-narios (2010-2034)
For the projections we assume that the fixed effect specific to Ghana is converging towards
the world mean of fixed effects until 2050, with our projections ending in 2030. This assump-
tion reflects a slow equalization of fixed effects across countries over time. If Ghana follows this
trend, it reaches a fixed effect of -0.015 by 2030, which would correspond to the fixed effect es-
timated for Ecuador, South Africa, Egypt or Azerbaijan in the estimation for the period 1970-2009.
Scenario 1: In a first scenario, we assume that all covariates remain at the level of the last
observation, which corresponds to 2005-09, and only the initial income level is updated (and with
it the interaction term of education and initial GDP). This implies that the values of investment,
government consumption, openness, education, inflation, democracy and natural resource rents do
not improve as compared to the period 2005-09 and the impact of oil production is not considered.
In particular, the covariates remain at the values displayed in Table 5.
Figure 5 shows how growth rates could develop under these assumptions (Scenario 1). As the
initial income level increases, growth rates are predicted to slow down, but for the upcoming decade
growth rates above 4% are predicted.
15
Table 5 – Values of covariates in Scenario 1
Variable Values in 2005-09
Education: share of tertiary education in working age population 6.3%Openness: imports plus exports as share of GDP (net of country size and population) 5.5%Government consumption as share of GDP 6.7%Investment as share of GDP 22.4%Inflation 13.9Democracy index 9.2Oil rents as share of GDP 0%Non-oil natural resource rents as share of GDP 6.9%
Without further improvements in investment, trade, education and government consumption,
Ghana’s GDP per capita growth rates are expected to stagnate at roughly 4.5% for the upcoming
decade and begin to decline after that.
Scenario 2: In a less pessimistic scenario, education, openness, investment, government con-
sumption and the non-oil natural resource rents are assumed to grow as they did during 1980-2009,
while inflation and democracy stay constant at 2005-09 values. It is assumed that there is no oil
production. More precisely, we use period growth rates of the explanatory variables as listed in
Table 6.
Table 6 – Values of covariates in Scenario 2
Variable Growth per 5-yearperiod
Education: share of tertiary education in working age population 7.2%Openness: imports plus exports as share of GDP (net of country size and population) 39.9%Government consumption as share of GDP -0.04%Investment as share of GDP 0.075%Inflation 0% (value of 2005-09:
13.9)Democracy index 0% (value of 2005-09:
9.2)Oil rents as share of GDP 0%Non-oil natural resource rents as share of GDP 16.8%
Under this scenario, GDP per capita growth rates are predicted to steadily increase until 2025,
with an average value of approximately 5.5% GDP growth until 2030. It should be noted, that the
average growth rate of the investment share in GDP since 1980 was as low as 0.075%, nevertheless,
a continuation of this increase accounts for more than 2%-points of the predicted GDP growth rates.
Table 9 tabulates the contribution of different explanatory variables to predicted GDP growth
rates under Scenario 3. The numbers show the %-point contribution to GDP growth rates for the
given specification. If investment increases by the rate it did on average since the 1980s (see Table
6), it contributes to GDP growth by 2%-points. The numbers have to be interpreted with care,
18
however, as due to the dynamic nature of the model they do not imply that growth rates decline by
2%-points if investment fell to zero. A similarly high positive impact on growth rates is attributed
to the non-oil natural resource rents, while government consumption appears to be the variable
that reduces GDP growth rates by the largest extent. It should be noted, however, that the results
should be interpreted as long run effects and no predictions can be done for a short run perspective.
The democracy measure does have a very small and even negative impact on GDP growth and
it does not appear as a significant driver of growth. This fact can partially be explained by the
inclusion of fixed effects: To the extent that they are constant over time, institutional character-
istics, political stability and business environment are comprised in the fixed effects. Only the
part that changes over time can be addressed in the estimation. The persistent nature of institu-
tional characteristics usually leads to a low variation in the variables and therefore identification
is difficult.
19
References
Barro, R. J. (2003): “Determinants of Economic Growth in a Panel of Countries,” Annals of
Economics and Finance, 4(2), 231–274.
Bollen, K. (1990): “Political Democracy: Conceptual and Measurement Traps,” Studies In
Comparative International Development, 25, 7–24.
Hanushek, E., and L. Wossmann (2008): “The Role of Congnitive Skills in Economic Devel-
opment,” Journal of Economic Literature, 8(1), 1–14.
Heston, A., R. Summers, and B. Aten (2012): Penn World Table Version 7.1 Center for
International Comparisons of Production, Income and Prices at the University of Pennsylvania.
Lejarraga, I. (2010): “Roaring Tiger of Purring Pussycat: A Growth Diagnostics Study of
Ghana,” Development Research Department Chief Economist Complex, African Development
Bank.
Lutz, W., J. C. Cuaresma, and W. Sanderson (2008): “The demography of educational
attainment and economic growth,” Science, 319, 1047–1048.
Lutz, W., A. Goujon, S. K.C., and W. Sanderson (2007): “Reconstruction of population by
age, sex and level of educational attainment of 120 countries for 1970-2000,” Vienna Yearbook
of Population Research, 2007, 193–235.
Sauer, P., and M. Zagler (2012): “(In)equality in Education and Economic Development,”
Paper Prepared for the 32nd General Conference of The International Association for Research
of Income and Wealth Session 2D, [http://www.iariw.org/papers/2012/SauerPaper.pdf].
Teorell, J., N. Charron, M. Samanni, S. Holmberg, and B. Rothstein (2011): The
Quality of Government Dataset Version, 6Apr11,University of Gothenburg: The Quality of
Government Institute [http://www.qog.pol.gu.se].
World Bank (2013): Energizing Economic Growth in Ghana: Making the Power and Petroleum
Sectors Rise to the Challenge,Energy Group, Africa Region, The World Bank.
20
A Appendix
Table A.10 – Description of variables
Variable Description Source
Rule of Law A composition of various indicators measuring the extent to which agentshave confidence in and abide by the rules of society; it includes perceptionsof the incidence of crime, the effectiveness and predictability of judiciary,and the enforcement of contracts.
WGI
Government Effectiveness A composition of various indicators measuring the quality of public servicepositions, the quality of bureaucracy, the competence of civil servants, theindependence of the civil service from political pressures, and the credibilityof the governments commitment to policies.
WGI
Political Stability A composition of various indicators measuring the likelihood that thegovernment in power will be destabilized or overthrown by unconstitutionalor violent means, including domestic violence of terrorism.
WGI
Corruption A composition of various indicators measuring the perception of corruption,defined as the exercise of public power for private gain.
WGI
Indicators used to assess the institutional quality.
WGI is short for the World Bank Worldwide Governance Indicators
For a more detailed description see Teorell, Charron, Samanni, Holmberg, and Rothstein (2011)
21
ETHCOGMOZNGA
TGOMNGKEN
NPLCIV
BGDPHL
KGZZMB
JORBENMDASEN
ZWETJK
KHMMAR
CMRINDVNM
PRYPAKTZAUKRLKA
NICHTIGEO
IDNBOLNAMBHRTHAMYSTUNPER
DZAHND
SYRIRQKAZ
AZEECUZAFEGYGAB
RUSALB
SDNMKD
SGPCOL
GHA
ESTIRNGTM
PANBGRSAUDOMARM
HRVCHNTURJAMBRAMEXSLVBLRLTU
HUNARGURYSVK
SURPOL
LVASVNCZE
BELTCDPRTLUXTTOMLTCYPIRLKOR
CRICANAUS
NZLGNQFINESP
NLDDNKDEUGRCITA
CHEJPNAUTSWEFRA
USAGBR
ISRISLNOR
QAT
-.1 -.05 0 .05 .1 .15Estimated Fixed Effects
Figure A.7 – Fixed effects based on specification (5) in Table 2
22
Table A.11 – Inclusion of countries and periods in the different specifications of Table 2
Country (1),(2) (3),(5) (4) Country (1),(2) (3),(5) (4)
The difference between the specifications (1), (2) and (3), and (5) arises from the impossibility of splitting the natural resource variableinto oil and non-oil components for some countries. Likewise, specification (4) includes only countries for which the data allow a furtherdecomposition of non-oil natural resources.
23
CIV ZMB
COG
TZA
ETH
CMRGHA
BEN
GAB
ZWE
MOZ
TCD
GNQ
NGA TGOKEN
SENGHA
-.1
-.05
0.0
5.1
.15
Fix
ed E
ffect
s
-3 -2 -1 0 1 2Polititcal Stability
CIVZMB
COG
TZA
ETH
CMRGHA
BEN
GAB
ZWE
MOZ
TCD
GNQ
NGA TGOKEN
SENGHA
-.1
-.05
0.0
5.1
.15
Fix
ed E
ffect
s
-2 -1 0 1 2Corruption
Figure A.8 – Correlation between fixed effects and institutional characteristics
Calculation of Expected Oil Rents in GDP
Based on oil production forecasts5 for Ghana (World Bank, 2013), expected oil rents in GDP, mea-
sured as the share of GDP that results from oil production, are estimated. Using IMF projections
of GDP for Ghana with and without oil production, we approximate the expected oil rents by the
difference of the two GDP scenarios. We divide the difference by GDP to obtain expected shares
of oil rents in GDP that enter the growth rate projections.
The difference between the projected oil- and non-oil GDPs is the inclusion of revenues, cal-
culated as the product of oil quantity forecasts (taken from World Bank, 2013) and expected oil
prices (IMF, World Economic Outlook), and an estimate of oil-related investment in the former.6
5Base case oil production forecasts are used.6As a robustness check of the approximation of the oil rents variable, we projected GDP growth rates based on
oil-rents that do not include oil-related investment. The results do not change qualitatively, and average expectedGDP growth rates for 2010-30 differ by 0.2%-points.