WPS8153 Policy Research Working Paper 8153 Natural Resources, Institutions, and Economic Growth The Case of Nigeria Anna K. Raggl Macroeconomics and Fiscal Management Global Practice Group July 2017 Policy Research Working Paper 8153 Abstract Using growth regressions with panel data, this study iden- growth-enhancing effect of natural resources is tied to a tifies the determinants of economic growth, highlighting sound institutional environment and low levels of corrup- in particular the role of natural resources and institutional tion. Accumulation of human as well as physical capital, quality. The overarching aim of this exercise is to learn but also the quality of institutions and natural resource about the drivers of growth in Nigeria, and to predict rents are estimated to be particular important ingredi- growth rates of gross domestic product per capita for the ents for a prosperous economic development in Nigeria. country under different scenarios. This study finds that a This paper is a product of the Macroeconomics and Fiscal Management Global Practice Group and is a background paper for the Nigeria Growth and Competitiveness Report, entitled Towards Sustainable Growth in Nigeria: Empirical Analysis and Policy Options. Vols. 1 and 2. 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 anna.raggl@wu.ac.at. 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. Produced by the Research Support Team Natural Resources, Institutions, and Economic Growth: The Case of Nigeria Anna K. Raggl∗ Keywords: Economic growth, natural resources, institutions, Nigeria. JEL Classification Codes: O13, O47, Q32, P48, O55. ∗ Foreign Research Division, Oesterreichische Nationalbank (OeNB). Email: anna.raggl@oenb.at. The opinions are strictly those of the author and in no way commit the OeNB. This study was written in the course of a consultancy to the World Bank while the author held a position at the University of Salzburg. The author is grateful for valuable comments from Dilek Aykut, Santiago Herrera, Carolina Lennon, and an anonymous referee. 1 Introduction After the discovery of oil in Nigeria in 1956, the country started oil production in 1958, and soon after became the main oil exporting nation on the African continent. In spite of the oil boom in the 1970s, the expected prosperity was failing to appear. GDP per capita (constant) was stagnating. It did not improve significantly and sustainably until the mid 2000s, and only caught up with the Sub-Saharan African average in 2010. Other indicators of economic development have been following a similar pattern—the share of people living below $1 per day increased from 36% in 1970 to a staggering 70% in 2000, and at the same time the share of extremely wealthy individuals grew, such that the income distribution widened considerably1 . An additional characteristic of Nigeria’s development is the volatility of its growth rates. While its average GDP per capita growth rate was just over 1% between 1980 and 2014, the standard deviation was close to 7.5, which is higher than in other Sub-Saharan African countries, and in other oil-producing nations (see Table A.1 in the Appendix). Ever since the important contribution by Sachs and Warner (1995), in which a negative influ- ence of natural resource wealth on economic development was shown empirically, Nigeria’s lack of development was attributed to its oil-abundance, and accompanying Dutch disease effects. van der Ploeg (2011), however, remarks that ”[i]t is hard to maintain that the standard Dutch disease story of worsening competitiveness of the non-oil-export sector fully explains [Nigeria’s] miserable economic performance”. Empirical research by Sala-i-Martin and Subramanian (2003, 2013) backs up that presumption: natural resources have a deteriorating impact on the quality of institutions, and through that channel natural resources harm economic development, even in the absence of Dutch-disease effects. This data-based analysis aims at re-investigating the causes for Nigeria’s growth performance. Using panel data of close to 150 countries during 1970 and 2014, we assess the determinants of GDP per capita growth, highlighting in particular the role of natural resources and institutional quality. Long-term, cross-sectional analyses are performed in addition, in order to carefully take into account that institutional quality measures are prone to endogeneity and measurement errors by using Two-Stage-Least-Squares estimators. The ultimate aim of the analysis is to learn about the drivers of growth in Nigeria, and to assess the country’s future growth potential. Therefore, various interaction terms in the panel setting allow a deviating impact of several factors in Nigeria as compared to the rest of the sample. These estimations are used to project GDP per capita growth rates for Nigeria under different scenarios. The remainder of this paper is organized as follows. Section 2 summarizes existing literature on the resource-growth nexus, and the relevance of the institutional environment of countries in this context. Before the estimation results are presented in Section 5, Sections 3 and 4 outline the 1 Data on poverty rates are taken from van der Ploeg (2011), and developments in income inequality from Sala- i-Martin and Subramanian (2003). 2 estimation strategies and describe the data that are used in the analysis. Drawing on the results of the growth regressions, Nigeria’s growth rates are predicted assuming different scenarios in Section 6. Section 7 concludes. 2 Natural Resources, Institutional Quality, and Growth Starting with the observation that resource-abundant countries are often performing worse than resource-poor countries, Sachs and Warner (1995) study empirically the impact of natural resources on economic growth using worldwide cross country data. In their seminal contribution, they find evidence for a detrimental effect of natural resources on growth, and in subsequent studies they further confirm the robustness of this result2 —a controversial finding that triggered an ongoing scholarly discussion about the curse and the blessing of natural resources. In response, various analyses were dedicated to deepen the understanding of this result, and in particular, to identify the mechanisms through which natural resources result in low economic growth rates. One of the most common arguments is that high shares of natural resources can lead to overvaluations of the real exchange rates. A consequential contraction of the tradable sector can weaken economies’ development prospects, especially if this sector exerts economies of scale, by learning-by-doing, for example (Torvik, 2001; Atkinson and Hamilton, 2003). Sachs and Warner (1995, 1997, 2001) mainly attribute their empirical findings to these so-called Dutch-disease effects. Another strand of the literature argues that natural resources may lead to a ”crowding-out” of investment in human capital. Strong primary sectors lower the incentives to dedicate sufficient resources to other, more education-intensive sectors. Gylfason, Herbertsson, and Zoega (1999) highlight that school enrollment rates are lower in countries with a high share of the labor force engaged in the primary sector. Low human capital accumulation can translate into poor growth rates—directly via the channel of productivity, or through indirect effects on political stability, health, or democracy. Empirical evidence suggests, that a considerable part of the negative impact of natural resources on growth can be attributed to lower educational attainment in resource- abundant countries (Gylfason, 2001). Rampant rent-seeking behavior is a further often-investigated transmission mechanism. The rents generated by natural resources cause an increased number of agents engaged in rent-seeking, as opposed to pursuing productive activities, and this voracity effect destroys the rents generated by natural resources. Lane and Tornell (1996), for example, develop a model that shows that in countries with powerful groups and low institutional quality, growth rates reduce due to natural resource windfalls, as higher productivity increases the demand for transfers, and these redistribu- tional effects may outweigh the growth-enhancing effects of resource endowments. Hodler (2006) provides theoretical as well as empirical evidence that increased rent-seeking behavior weakens 2 See Sachs and Warner (1997) and Sachs and Warner (2001). 3 property rights, and that in turn further reduces the attractiveness of productive activities. The more rivaling groups, i.e. the higher the fractionalization in the countries, the more likely it is, that natural resources are a curse, as opposed to a blessing.3 While the rent-seeking and the Dutch disease hypotheses claim an unconditional negative influ- ence of natural resources on growth (Mehlum, Moene, and Torvik, 2006), there is a large literature that tied the negative relationship to certain conditions, most importantly the institutional environ- ment and corruption in the countries. The contribution by Bulte, Damania, and Deacon (2005) is among the first studies investigating the inter-relationship between natural resources, institutions, and economic and human development. While they find only limited evidence for a direct effect of natural resources on human development, evidence for an indirect link via institutional quality is presented. Similarly, Mehlum, Moene, and Torvik (2006) confirm a negative direct effect of natural resources on growth, but conclude that the combination of ”grabber friendly” institutions and natural resources harms growth, whereas ”producer friendly” institutions help materializing the full benefits of resources.4 A final important literature, related to the discussed influence of institutional quality, and the rent-seeking hypothesis, is the role played by corruption. Leite and Weidmann (1999), and more recently Badinger and Nindl (2014), show, that natural resources facilitate corruption. Again, the strength of this link has been tied to the countries’ institutional environments. Resource rents appear to mainly increase corruption levels, if countries have a comparably low polity -score, a measure of the degree of democracy in a country (Bhattacharyya and Hodler, 2010). Stable democracies—Norway, Australia, and Canada are named as examples—do not suffer from these adverse effects, as their institutions prohibit rent-seeking behavior to a large degree. CHN CHN 8 8 GDP per capita growth, 1980-2014 GDP per capita growth, 1980-2014 6 6 BTN KOR CPV BWA LAO THA IND IRQ IND THA IRQ LKA 4 4 POL MUS IDN POL IDN IRL CHL MYS CHL MYS MNG MOZ DOM PANBGD TCD BGR MAR ROM TUR SWZ EGY UGA BGR MAR TUR ROM TCD EGY ISR RWA LSO ETH NPL TUN ISR TUN URYPAK CYP TZA BFA COL PAK COL NOR 2 2 JPN PRT CRI GBR FIN GHA NOR AUS GHA JPN GBR AUS AUT SWE DEU ESP USAPRY TTO QAT AUT SWE DEU ESP USA TTO QAT JOR DNK FRA NLD NZL PHL BRA ARGCAN SYR PER COG NLD NZL BRA CAN DNK PHL FRA PER ARG ECUSYR COG CHE ITA MLI BEN HND ECU NGA ITA NGA FJI SLV NAM SLE BOL MEX ZMB BOL MEX BRB GRC KEN GTM MWI JAMMRT GTM ZAFSUR IRN IRN DJI SEN GMB CMR GIN GNB CMR 0 0 TGO AGO VEN GAB AGO VEN GAB COMZWE SAU SAU BDI NER MDG CIV CIV CAF LBR -2 -2 ZAR ZAR 0 20 40 60 80 0 20 40 60 80 Natural resources, share in GDP Oil rents, share in GDP Figure 1 – Correlation between natural resource shares in GDP (left) and oil shares in GDP (right, only if oil rents positive) and GDP per capita growth 3 See also Baland and Francois (2000); Torvik (2002) for more rent-seeking models. 4 Brunnschweiler (2008) offers similar findings, using an alternative measure of resource-abundance. 4 100 DNK PRT FIN NOR ZAR NGA FRA CHE DEU JPN SWE IRL ESP NLD CRI NZL AUS CAN USA TCD SYR CMR IDN COG GAB AUT GRC ITA GBR KEN CAF LBREGY BRB CPV TTO MDG NPL GIN PRY 80 GTM COMBGD SLE MUS CYP HND VEN IRQ KORPOL BWA JAM DOM GNB UGA URYSEN THA TGOMRT DOM ARG ECU DJI SLV LAO ISR PAK BOL AGO 80 VEN CIV MLI BRA GHA TZA SUR MEX NER FJI GHAETH TUN PAN IND BEN CHL BOL ZWEROM LSO NAM BTN MAR 60 BGR JOR GAB SEN PHLBEN MLI CIV ZMB NER BFA THA MNG BFA MEX COL MYS GMBPHL BGD ROM PANBRARWA ECU IRN 60 PRY MYS QAT ZAF PER PER Rule of law Corruption TGOHND MUS IND LKACOL NPL TUN GMB MOZ BDI TUR CMR NGA LKA SWZ ZMB SLV COM SLE UGA BGR TZA CHN DJI CAF TUR GRC MWI LSO ARG ZAF MDG BDI MNG 40 MAR ZWE LBR NAM SAU MOZ RWA MRT GNB 40 GTM KEN ITA JAM MWI GIN IDN QAT KOR SUR PAK SWZ CYPFJI ETH COG IRN EGY JORCRI TCD CHNAGO BRB BWA ISR POL 20 ZAR IRQ SAU TTO 20 CPV IRL CHL BTN SYR PRT AUT JPN FRA LAO ESP URY NLD USA GBRAUS CAN DEU CHENZLFIN SWE DNK NOR 0 0 0 20 40 60 80 0 20 40 60 80 Natural resources, share in GDP Natural resources, share in GDP Figure 2 – Correlation between natural resource abundance and institutional indicators: rule of law (left) and corruption (right) Figure 1 provides a graphical representation of long-term GDP per capita growth rates and natural resource rents as well as oil rents (both measured as shares in GDP). It is apparent, that no correlation between long-term growth and natural resources can be detected. There are high-growth countries—such as the Asian tigers—that are poorly endowed with natural resources, whereas ublica Bolivariana countries rich in resources—Nigeria, Liberia, the Republic of Congo, or Rep´ de Venezuela are examples—lack long-term economic progress. Recent research stresses the role of institutions in the growth-resource-nexus. The graphs in Figure 2 attempt to find descriptive evidence for a correlation between resource endowments and institutional quality. The correlation is remarkable: Resource-rich countries, among them Nigeria, are associated with lower average institutional quality and higher levels of corruption (based on long-term averages of the variables), supporting the presumption that resources can deteriorate institutional quality and promote rent- seeking and corruption. Based on the current state of the literature, and on the important work by Sala-i-Martin and Subramanian (2003, 2013), this study addresses the determinants of GDP per capita growth, focusing in particular on the role played by natural resources, institutions, and corruption, and highlights the growth prospects for Nigeria in a global setting. 3 Empirical Setting Panel fixed effects estimations The main results of this analysis and the ingredients for the out-of-sample predictions of Nigeria’s growth rates are based on cross country panel growth regressions. The real per capita growth rate in country i and period t, git , is regressed on a variable expressing natural resource abundance, institutional quality (Iit ), and a set of covariates X, and the basic specification can be characterized as 5 GDPi,t−1 N ATit git = α + β log + γf + δIit + Xη + µi + νt + it (1) P OPi,t−1 GDPit where α is a constant, µi and νt are country and period fixed effects, and it is the remaining error term. In particular the country fixed effects that control for time-invariant characteristics inherent to the countries are important components in this setting, because they limit potential unobserved heterogeneity biases.5 In order to limit the chance of business cycles and short term GDP fluctuations distorting the results, t corresponds to five-year periods and the variables enter as five-year averages or initial values of the respective period.6 Important control variables that are included in the matrix X are human capital, investment, government consumption, the openness of the countries, as well as a measure for the undervaluation of the currency. In all specifications, a measure of natural resources is included. In a simple setting, we control for the share of total natural resource rents in GDP. In order to allow for heterogeneous effects, we decompose that variable into rents from oil and non-oil natural resources. As the effects of oil rents on GDP per capita growth may be non-linear, the oil rents variable is split into four quartiles in a third setting. The underlying hypothesis is that the degree of oil-dependence might affect the contribution of oil rents to GDP growth rates. As the major aim of this study is to carve out the main determinants of growth in Nigeria, several variables are interacted with a Nigeria dummy in order to learn about possible deviating effects. Endogeneity The coefficient estimates of the institutional measures (incl. the corruption index) might be biased in an estimation framework that does not account for endogeneity. Endogeneity problems could come from various sources. First, both the institutional characteristics and the growth rates of countries could respond simultaneously to omitted factors. Such factors could endez and be cultural dispositions, legal frameworks or historical conventions (see for example M´ ulveda, 2006). As a panel setting allows to control for country-specific fixed effects, country- Sep´ inherent factors that are constant over time are controlled for, and persistent country characteristics will not cause biased estimates. Second, the estimates could suffer from reversed causality—a problem that arises when not only corruption influences GDP growth, but also the reverse is true. Third, the rule of law as well as the corruption indicator are indices, and not precise measurements. Measurement errors cause biased estimates, when they are correlated with the observed (and potentially) mis-measured values. Biased estimates that should not be interpreted causally are the consequences of all described sources of endogeneity. Several attempts to instrument the institutional variable in a panel setting were not fruitful due to a lack of credible instruments that vary over time. Instrumental variables 5 See for instance M´ ulveda (2006). endez and Sep´ 6 See Table A.2 in the Appendix for detailed description of the variables and the form they enter the estimations. 6 that have proved helpful are constant within countries (see below for a discussion of those instru- ments), and therefore they cannot be used in a panel framework that controls for country fixed effects. As these country fixed effects help overcoming another important (heterogeneity) bias, the cost of omitting them is too high. Inspired by recent literature (Werker, Ahmed, and Cohen, 2009; Nunn and Qian, 2014; Dreher and Langlotz, 2015) on the causal link between aid and growth, attempts were made to interact presumably excludable, but constant instruments with a variable that varies over time. The special feature of this strategy is that the second variable that the excludable instrument is interacted with must not necessarily be exogenous. Although promising at first, further considerations revealed that the instruments for natural resources proposed in the literature are not appropriate for such an instrumentation strategy. For this reason, a bias reduc- tion with two-stage least squares methods seems not feasible in the panel setting and we rely on a cross-sectional analysis for an attempt to establish causality. Instrumentation of institutional quality in a non-panel, cross country setting Very much in the line with the literature on institutional quality and economic growth, the possible endogeneity of the institutional variable is instrumented in a cross-country, long term growth setting. The time dimension exploited in the main results, that are based on the panel setting outlined above, needs to be neglected, and long term averages and initial values of the variables are used. In this setting, one can draw on the literature for successful instrumentation strategies. Hall and Jones (1999), for instance, use the fraction of the population speaking English or another major European language as an instrument for institutional quality, arguing that the language shares are approximating the exposure to Europe. Using similar arguments, the seminal contribution by Acemoglu, Johnson, and Robinson (2001) exploits the variation in mortality rates of early European settlers in the colonial countries to approximate the foundations of current institutional quality that have been established in the past by European settlers. Together with the presumption of high persistence of institutional quality, their instrumentation rests on the assumption that bad living conditions increased the likelihood of ”‘extractive”’ institutions, whereas a favorable environment caused settlers to build ”Neo-Europes”. Easterly and Levine (2003) instrument the institutional variable with ”endowments”, and use settler mortality, latitude, crops/minerals dummies, and a landlocked dummy. Using settler mortality as an instrument drastically reduces the number of observations, and for that reason we are following Hall and Jones (1999) and Sala-i-Martin and Subramanian (2003, 2013) and use English and other European language shares as instrumental variables for the two indicators for the quality of institutions—the rule of law and a corruption index. Formally, the growth equations that are estimated have the following form GDPi N ATi gi = α + β log + γf + δIi + Xη + i (2) P OPi GDPi 7 Institutional quality and the level of corruption is instrumented with the fraction of people speaking English and another European language using Two-Stage-Least-Squares (2SLS) estimation. The corresponding first stage is GDPi N ATi Ii = θ0 + θ1 Engi + θ2 Euri + θ3 + θ4 f + Xκ + ui (3) P OPi GDPi where Engi and Euri denote the proportion of people in country i, that speak English or another European language, respectively. 4 Data Panel analysis The sample used for the panel regressions contains 1,000 observations that rep- 7 resent 150 countries during 1970 (or later) and 2014 The time dimension is reduced to five-year averages (for log GDP per capita, the initial value is used), in order to net out business cycles and high short term volatility. One observation indexed (i, t) thus corresponds to a country i and the average during years t, t + 4 or the initial value in t. Table A.4 in the Appendix lists the countries and periods included in detail. The sources and precise definitions of the variables used are summarized in Table A.2 in the Appendix. A measure of the real undervaluation of currencies is constructed by the price level of an economy adjusted by the Balassa-Samuelson effect8 following Rodrik (2008). More specif- ically, the real effective exchange rate, calculated as the exchange rate over the PPP conversion factor, is regressed on (the logarithm of) per capita GDP and a set of time dummies. The differ- ence of the real exchange rate and the predictions from this regression is used as a proxy for real undervaluation. The resulting index is centered around zero. A positive value indicates under- valuation, and a negative value indicates overvaluation. Figure 3 shows the natural logarithm of the undervaluation index for Nigeria, as compared to other Sub-Saharan African oil-exporting and non-oil-exporting countries. This comparison shows descriptively, that Nigeria’s exchange rates have been overvalued—considerably so compared to other Sub-Saharan African countries—during the 1980s, 1990s, and early 2000s. Only in the mid-2000s, the degree of undervaluation returns to the Sub-Saharan African average. In order to measure institutional quality and the level of corruption in the countries, we rely on the recently established Varieties of Democracy Data Base (Coppedge, Gerring, Lindberg, Skaaning, Teorell, Tzelgov, Wang, Altman, Bernhard, Fish, Glynn, Hicken, Knutsen, McMann, Staton, and Zimmermann, 2015), that provides a wide array of indicators related to various aspects 7 All figures presented here are lower bounds, and refer to the sample of 1,000 observations in columns (3) to (5) in Table 1 and columns (3) to (5) in Table 2. 8 The adjustment accounts for the finding, that increases in income levels lead to a relative price increase of non-tradeable goods due to productivity improvements. 8 3 3 2 2 Log(undervaluation) Log(undervaluation) 1 1 0 0 -1 -1 1970 1980 1990 2000 2010 1970 1980 1990 2000 2010 Year Year Sub-Sahara Africa, oil Nigeria Sub-Sahara Africa, no oil Nigeria Figure 3 – Log(undervaluation) in Nigeria and other Sub-Saharan-African countries that produce oil (left) and that do not produce oil (right). A positive value indicates an undervaluation, a negative value an overvaluation of the currency. of democracy. The indicators chosen for this analysis are a rule of law index (v2xcl rol) and a corruption index (v2x corr). The former measures the equality before the law and individual liberty, the transparency and the enforceability of laws, and to what extend citizen have access to justice, secure property rights, freedom from forced labor, freedom of movement, physical integrity rights as well as freedom of religion. The corruption index includes measures of distinct types of corruption, thereby distinguishing between bribery and embezzlement, as well as between the levels at which corruption takes place—the highest levels as opposed to the public sector at large. It is calculated as a weighted average of public sector, executive, legislative, judicial corruption indices.9 Both indices in their original forms are normalized between 0 and 1, but had been rescaled for the analysis to run from 0 to 100 for interpretation purposes. In the panel data application, the indices enter the regressions as deviations from country-specific means (as country fixed effects are included, the standardization does not alter the coefficient estimates in any way). Cross-sectional analysis In order to account for a possible endogeneity of the institutional variable and the corruption indicator that is not controlled for by country fixed effects in the panel setting, we estimate long-run growth regressions using only long-term averages and/or initial values of the variables. The estimation sample consists of 113 countries. The country coverage is lower in this setting as compared to the panel regressions, because for some countries the time dimension of the data is not long enough to calculate long-term growth rates. While in the panel setting it is possible to control for period effects, this cannot be done in this framework, and the data need to be comparable in order to obtain reliable results. Table A.3 in the Appendix provides an overview 9 Please refer to the code book of that data set for a more detailed description (Coppedge, Gerring, Lindberg, Skaaning, Teorell, Tzelgov, Wang, Altman, Bernhard, Fish, Glynn, Hicken, Knutsen, McMann, Staton, and Zim- mermann, 2015). 9 over of data used, how they are measured, and which source they are from. 5 Growth Regressions 5.1 Panel Data Analysis Table 1 presents the elementary specification of the growth regressions using panel data and various extensions, all including country as well as period fixed effects. The first column allows human and physical capital, government expenditures, inflation, as well as openness to trade and institutional quality, measured by the rule of law, to influence GDP per capita growth rates. The inclusion of initial GDP per capita controls for convergence within countries, i.e. for convergence to a country- specific long-term equilibrium growth rate. The negative coefficient of initial GDP per capita confirms that classical convergence hypothesis, as increases in income per capita are followed by lower growth rates of GDP. Higher human capital, investment shares and trade openness, as well as good macroeconomic management (low inflation) and low levels of government expenditures are increasing medium-term growth rates of GDP per capita. Although the coefficient of institutional quality cannot be interpreted causally due to a potential endogeneity bias, the estimates suggest a positive relationship between an improved institutional environment and economic development. The estimated coefficients are not only statistically significant, but also their magnitude shows non-negligible impacts on economic growth. An increase in the share of upper secondary and tertiary educated in the labor force by 10%-points is estimated to increase GDP per capita growth by roughly 0.7%-points. Similarly, based on the results, an increase in the share of investment in GDP by 10%-points has the potential to raise growth by 1.6%-points. In columns (2) and (3), natural resource rents are added to the equation. Neither in the aggregate form, nor when disaggregated into oil and non-oil resource rents a significant impact can be detected.10 It is likely, that the impact of natural resources is too heterogeneous and/or depends on additional factors, and only a further refinement can shed light on the role played by natural resources. As an extension of column (2), Figure 4 presents the estimates of the impact of natural resources on economic growth when interacted with period dummies. The variation over time is considerable and might offer an explanation for the insignificance of the natural resource variable in the previous columns. There is evidence for a heterogeneous effect over time, and it appears that during the last decade, natural resource rents had a significant (at the 10% level) and positive impact on economic growth.11 10 When splitting the natural resource variable further into oil, natural gas, mineral and forest rents (not displayed in the table), no further insights are gained. While all coefficients but the one for forest rents are positive, none of them can be distinguished from zero at a statistically significant level of confidence. 11 The exceptionally large effect in the first half of the 1970s could be due to the vast increase in oil prices, but also to a smaller country-coverage in this early period of the sample. 10 Table 1 – Determinants of growth per capita: Panel fixed effects estimations (1) (2) (3) (4) (5) Log(GDP per capita) -4.558∗∗∗ -3.407∗∗∗ -3.442∗∗∗ -3.571∗∗∗ -3.538∗∗∗ (0.000) (0.000) (0.000) (0.000) (0.000) Education 0.0683∗∗ 0.0510∗∗∗ 0.0451 0.0577∗∗ 0.0492∗ (0.029) (0.009) (0.115) (0.046) (0.087) Governm. cons. -0.102∗∗ -0.143∗∗∗ -0.0991∗∗∗ -0.0947∗∗∗ -0.0923∗∗ (0.018) (0.000) (0.009) (0.010) (0.015) Investment 0.162∗∗∗ 0.144∗∗∗ 0.150∗∗∗ 0.153∗∗∗ 0.153∗∗∗ (0.000) (0.000) (0.000) (0.000) (0.000) Openness 0.0147∗∗ 0.0176∗∗∗ 0.0134∗∗ 0.0107∗ 0.0133∗∗ (0.037) (0.005) (0.029) (0.082) (0.033) Inflation -0.00271∗∗∗ -0.00185∗∗∗ -0.00159∗∗∗ -0.00154∗∗∗ -0.00154∗∗∗ (0.008) (0.000) (0.000) (0.000) (0.000) Rule of law 0.0159∗∗ 0.0193∗∗∗ 0.0168∗∗ 0.00479 0.0145∗∗ (0.020) (0.007) (0.014) (0.571) (0.033) Natural res. 0.0272 (0.316) Oil rents 0.0627 0.106∗∗ (0.144) (0.038) Non-oil rents 0.0267 -0.121∗∗ 0.0189 (0.437) (0.022) (0.544) Rule of law x oil rents -0.000843 (0.325) Rule of law x non-oil rents 0.00236∗∗∗ (0.000) Oil rents, 1st qu. 6.753∗∗∗ (0.000) Oil rents, 2nd qu. 0.757∗∗∗ (0.002) Oil rents, 3rd qu. 0.151∗∗ (0.012) Oil rents, 4th qu. 0.0766∗ (0.075) Observations 1031 1004 1000 1000 1000 Countries 150 150 150 150 150 R2 0.384 0.210 0.293 0.304 0.308 ∗ ∗∗ Each specification includes country and period fixed effects, as well as a constant. p -values in parentheses. p < 0.1, p < 0.05, ∗∗∗ p < 0.01. Another reason for the lack of a clear relationship for natural resources and growth could be a missing link to institutional quality. For natural resources to improve growth rates, a stable institutional environment might be a prerequisite that guarantees a fair distribution of revenues associated with the rents. When allowing for such an interaction in column (4), it appears that on 11 .4 .2 0 -.2 70 75 80 85 90 95 00 05 10 99% CI 95% CI 90% CI Figure 4 – Impact of natural resources on per capita growth rates, by period average, oil rents improve growth rates when institutional quality is held constant, whereas rents from other natural resources require solid institutions for them to be beneficial. It is important to note again, that as the endogeneity of the institutional variable is not accounted for, causal interpretations cannot be made.12 In the final column of Table 1, heterogeneities with respect to the countries’ dependence on oil are allowed for (see also Raggl, 2014, for a similar setting). Therefore, oil rents are interacted with dummies indicating the quartile of oil rents in GDP that are calculated using the full sample for each period. The results in column (5) show, that on all levels of oil dependence, oil rents increase GDP per capita growth rates on average. However, the magnitude of a 1%-point increase in oil differs considerably across the four levels of oil dependence, and is highest for countries with low oil dependence. Nigeria is highly dependent on oil, and is allocated to the fourth quartile in all but 12 Several attempts to instrument the institution indicators in the panel setting were not fruitful. As common in the literature, 2SLS estimations are carried out on the cross-sectional, long-term level, that abstracts from the variation over time in the data (Hall and Jones, 1999; Acemoglu, Johnson, and Robinson, 2001; Sala-i-Martin and Subramanian, 2003). However, possible endogeneity in the panel framework will not negatively affect the quality of the out-of-sample predictions in Section 6, if it is safe to assume that the patterns of endogeneity prevail during the course of the prediction period. 12 one period, and belongs to the group of countries that exhibit the lowest benefits on average. A 10%-point increase in the share of oil rents in GDP—for Nigeria this would imply an incline from approximately 16% in GDP in 2010-14 to 26%—is estimated to increase GDP per capita growth by roughly 0.8-1.5%-points in the third and fourth quartile, respectively. The following paragraph reviews and discusses results particular for Nigeria in more detail. A Focus on Nigeria Table 2 displays panel fixed effects estimations that highlight the role of natural resources and institutional quality in Nigeria. The first column suggests that the impact of natural resources on growth are significantly lower in Nigeria than in the rest of the sample. In fact, increases in natural resource rents appear detrimental to growth in the country. In the second column, an index for the rule of law as a measure of institutional quality is added, and as a result, the negative impact of natural resources declines in magnitude for Nigeria (from -0.042 to -0.029). These findings are in line with Sala-i-Martin and Subramanian (2003), and provide evidence for a negative relationship between natural resources and institutional quality. When omitting institutional quality, and natural resources weaken institutions, then the impact of lower institutional quality is wrongly attributed to natural resources. Controlling for institutions in column (2) thus increases the coefficient of resources for Nigeria. This interdependence between resources and institutions is further confirmed in column (3), where the addition of another measure of institutional quality, political corruption, leads to a further (small) increase in the impact of natural resources. The positive impact of corruption in the full sample has previously been found in the literature, it is summarized as the ”‘greasing the wheels”’-effect of corruption (Egger and Winner, 2005; Vial and Hanoteau, 2010; Campos, Dimova, and Saleh, 2010). As in this setting no measures have been undertaken to limit a possible endogeneity bias, causal interpretations are not justified, however. In columns (4) and (5) the corruption variable is interacted with a Sub-Saharan Africa dummy and with a Nigeria dummy, respectively. Apart from the finding that corruption is associated with lower growth in Sub-Saharan Africa, and even more so in Nigeria13 , the effect of natural resources on growth in Nigeria further increases (column 4) and becomes positive in column (5). This latter finding suggests, that in particular in Nigeria, there is a strong connection between natural resources and corruption, and when controlling for this link, resource rents do no longer harm, but foster economic growth. The negative impact of natural resources in Nigeria found in the previous specifications appears to be driven by the high levels of corruption associated with the country’s resource wealth, and in order to enable natural resources to be beneficial for growth a stable institutional environment is crucial. Distinguishing between the oil and non-oil rents in GDP in column (6) reveals, that both oil and non-oil resource rents are stimulating economic growth in Nigeria. Although Nigeria is an 13 The coefficient for Nigeria appears particularly large. However, when interpreting the effect in terms of standard deviations, the effect is still considerable, but less pronounced. A one standard deviation increase in the corruption indicator reduces growth by 0.8 standard deviations. 13 Table 2 – Determinants of growth per capita: Panel fixed effects estimations, focus on Nigeria (1) (2) (3) (4) (5) (6) (7) ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ Log(GDP per capita) -3.157 -3.376 -3.313 -3.416 -3.235 -3.312 -3.403∗∗∗ (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Education 0.0518∗ 0.0431 0.0521∗ 0.0454∗ 0.0502∗ 0.0540∗ 0.0584∗∗ (0.067) (0.122) (0.054) (0.078) (0.060) (0.051) (0.036) Governm. cons. -0.139∗∗ -0.0969∗∗ -0.0898∗∗ -0.0821∗∗ -0.0849∗∗ -0.0848∗∗ -0.0785∗∗ (0.023) (0.011) (0.025) (0.031) (0.033) (0.031) (0.048) Investment 0.191∗∗∗ 0.145∗∗∗ 0.147∗∗∗ 0.150∗∗∗ 0.151∗∗∗ 0.158∗∗∗ 0.161∗∗∗ (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Openness 0.0162∗∗∗ 0.0141∗∗ 0.0137∗∗ 0.0127∗∗ 0.0127∗∗ 0.0119∗ 0.0117∗ (0.009) (0.021) (0.029) (0.045) (0.039) (0.057) (0.064) Inflation -0.0015∗∗∗ -0.0016∗∗∗ -0.0016∗∗∗ -0.0016∗∗∗ -0.0016∗∗∗ -0.0016∗∗∗ -0.0016∗∗∗ (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Rule of law 0.0164∗∗ 0.0204∗∗∗ 0.0224∗∗∗ 0.0205∗∗∗ 0.0212∗∗∗ 0.0190∗∗∗ (0.017) (0.002) (0.002) (0.002) (0.002) (0.004) Rule of law x NGA 0.0406∗∗∗ 0.0731∗∗∗ 0.0932∗∗∗ (0.003) (0.000) (0.000) Corruption 0.0284∗∗ 0.0517∗∗∗ 0.0295∗∗ 0.0334∗∗ 0.0338∗∗ (0.039) (0.001) (0.032) (0.015) (0.011) Corruption x SSA -0.0659∗∗ (0.044) Corruption x NGA -4.578∗∗∗ -3.593∗∗∗ -3.203∗∗∗ (0.000) (0.000) (0.000) Natural res. 0.0721∗ 0.0439 0.0413 0.0449∗ 0.0425 (0.082) (0.108) (0.123) (0.083) (0.112) Nat. res. x NGA -0.114∗∗∗ -0.0730∗∗ -0.0675∗∗ -0.0665∗∗ 0.0967∗∗∗ (0.007) (0.020) (0.029) (0.026) (0.001) Oil rents 0.0758∗ (0.087) Oil rents x NGA -0.0504 -0.0400 (0.323) (0.444) Non-oil rents 0.0193 0.0114 (0.584) (0.720) Non-oil rents x NGA 0.675∗∗∗ 0.780∗∗∗ (0.000) (0.001) Oil rents, 1st qu. 6.815∗∗∗ (0.000) Oil rents, 2nd qu. 0.737∗∗∗ (0.002) Oil rents, 3rd qu. 0.161∗∗∗ (0.007) Oil rents, 4th qu. 0.0880∗∗ (0.048) Observations 1088 1004 1004 1004 1004 1000 1000 Countries 166 150 150 150 150 150 150 R2 0.352 0.292 0.299 0.307 0.310 0.314 0.328 ∗ ∗∗ Each specification includes country and period fixed effects, as well as a constant. p -values in parentheses. p < 0.1, p < 0.05, ∗∗∗ p < 0.01. 14 economy highly dependent on oil, after controlling for both institutional quality indicators, oil rents are contributing to growth in the country. The impact of the rule of law indicator does not change when interactions between resources and Nigeria dummies are included, nor when the corruption variable is added. Although no causal interpretation can be made in this setting, improvements of institutional quality within countries are associated with higher GDP per capita growth rates. Columns (5) to (7) of Table 2 allow for a deviating coefficient of the rule of law measures in Nigeria as compared to the rest of the sample. The results suggest, that the connection between the rule of law and GDP growth is particularly relevant in the country, even when an additional measure of institutional quality, political corruption, is controlled for. Therefore, both dimensions of institutional quality should thus be addressed in order to smooth the way for sustainable growth, and a growth-enhancing effect of Nigeria’s natural resources. Further results Table A.5 in the Appendix provides additional results related to the growth determinants in Nigeria. First, the impact of human capital on economic development is stronger when the level of GDP per capita is comparably small. In other words, the growth-enhancing effect of educational expansions is higher in developing countries. This result is underlined by the findings in column (2), where the effect of education is allowed to differ between Nigeria and the rest of the sample. It appears, that human capital accumulation is of particular importance for growth in the country. The variable is measured as the share of people with upper secondary education or more in the age group 20 to 64. An expansion of secondary and tertiary education is estimated to be a fruitful strategy to enhance economic progress in the long run. Second, a measure of the undervaluation of a currency, as suggested by Rodrik (2008) and discussed in more detail in the Data section, is included in the specifications (3) to (5) of Table A.5. Rodrik (2008) finds a significant and positive relationship between the degree of undervaluation of a currency and economic growth, and he argues and shows empirically that the impact is especially high in countries with low incomes per capita. When interacting the measure of undervaluation with a Nigeria dummy variable, the effect for the country is positive and significant. If, to a certain extent, the impact of natural resources on growth is channeled through overvaluations of the real exchange rates (Dutch disease effects), then an inclusion of a measure of undervaluation should improve the growth impact of natural resources. In other worlds, if the degree of undervaluation of a currency is held constant, the impact of natural resources should be more positive (less negative), as the negative effect via the exchange rate is controlled for. This can be observed for Nigeria. The impact of natural resources on growth turns positive, once controlled for undervaluation. For oil rents, the impact is still negative (columns 4 and 5), but considerably closer to zero compared to estimations without the undervaluation measure. This implies that there is evidence for Dutch disease effects in Nigeria, and policies that 15 are directed towards the prevention of strong currency overvaluation could improve the growth effects of natural resources. The magnitude of the effect of undervaluation on growth is non- negligible in Nigeria: an increase in the undervaluation index by 1% is estimated to raise growth rates by 0.06%-points (all relevant columns 3–5 in Table A.5), that could correspond to an increase from a 4% to a 4.06% growth rate. Similarly, an increase in the undervaluation index by 10%—not an unrealistic change given the large fluctuations—could improve GDP per capita growth rates by 0.6%-points. This finding is robust across different specifications, that include various alternating measures of natural resources. In order to highlight the potential of a less overvalued currency, Figure A.1 in the Appendix presents the estimated growth impact of a steady improvement of the undervaluation index towards the values estimated for the mid-1960s between 2010-14 and 2040-44. Nigeria had long periods of a severely overvalued currency from the 1970s until the early 2000s. Currently, the undervaluation is no longer negative, the index suggests even a slight undervaluation of the currency. Assuming that the undervaluation index continuously improves in the future, until by the period 2040-44 it reaches the level of 1965-69 (0.38), GDP per capita growth rates could improve by more than 1%-point on average in the long run. 5.2 Cross-sectional Analysis The results of the long-term growth analysis that instrument the institutional variable and the corruption indicator are presented in Table 3. The top panel displays the second stage of the 2SLS estimations, in which the rule of law indicator is instrumented, whereas the bottom panel displays the results for the corruption variable14 . The first two columns refer to OLS results, and columns (3) to (7) to 2SLS results. In the first specification (column 1), the institutional variables are omitted, and natural resource rents are not significant in the growth regressions. Controlling for the rule of law and the corruption index, respectively, results in a positive and significant coefficient of the resource variable. This finding is in line with the results of the panel analysis, and suggests a link between natural resources and institutional quality. Natural resources are associated with lower institutional quality, and that negatively affects growth rates. As soon as institutions are held constant, the impact of natural resources turns positive and significant. Instrumenting the rule of law and corruption does not change the significance nor the sign of the variables. If anything, their impact becomes more pronounced. This result holds for various model specifications, the inclusion of variables related to macroeconomic management, a currency undervaluation index, life expectancy and geographical characteristics. Further confirming the panel data results, the coefficient of the oil rents for highly oil-dependent countries (oil rents > median) is lower as compared to the coefficient for low oil-dependence. 14 Due to the high collinearity and the increased standard errors resulting from the 2SLS estimation, both indicators are treated in two separate regressions. 16 Table 3 – Determinants of long-term growth: 2SLS estimations using cross-sectional data (1) (2) (3) (4) (5) (6) (7) OLS OLS 2SLS 2SLS 2SLS 2SLS 2SLS Rule of law 0.0217∗∗ 0.0857∗∗∗ 0.0871∗∗ 0.0717∗∗ 0.0850∗∗ 0.0844∗∗ (2.40) (2.65) (2.53) (2.15) (2.21) (2.21) Log(GDP per capita) -0.684∗∗∗ -0.800∗∗∗ -1.142∗∗∗ -1.100∗∗∗ -1.348∗∗∗ -1.596∗∗∗ -1.593∗∗∗ (-4.18) (-4.91) (-3.83) (-3.54) (-4.37) (-3.60) (-3.61) Natural res. 0.0132 0.0210∗∗ 0.0443∗∗∗ 0.0419∗∗∗ 0.0418∗∗∗ (1.28) (2.23) (2.89) (2.61) (2.85) Oil rents 0.0586∗∗ (2.43) Oil rents < median 0.145 (0.64) Oil rents > median 0.0586∗∗ (2.44) Non-oil rents 0.0180 0.0179 (1.22) (1.21) Observations 113 113 113 113 113 113 113 Hansen J 1.879 1.932 1.802 1.248 1.274 p-value 0.170 0.165 0.180 0.264 0.259 Kleibergen-Paap 8.592 7.157 7.404 6.758 6.865 p-value 0.014 0.028 0.025 0.034 0.032 Corruption -0.0280∗∗∗ -0.0496∗∗∗ -0.0469∗∗∗ -0.0380∗∗∗ -0.0369∗∗∗ -0.0363∗∗∗ (-4.17) (-2.80) (-3.81) (-3.19) (-2.91) (-2.93) Log(GDP per capita) -0.684∗∗∗ -0.822∗∗∗ -0.929∗∗∗ -0.852∗∗∗ -1.156∗∗∗ -1.231∗∗∗ -1.228∗∗∗ (-4.18) (-5.14) (-5.06) (-5.09) (-5.92) (-5.63) (-5.63) Natural res. 0.0132 0.0185∗ 0.0227∗∗ 0.0197∗∗ 0.0234∗∗∗ (1.28) (1.83) (2.28) (2.00) (2.96) Oil rents 0.0286∗∗∗ (2.77) Oil rents < median 0.114 (1.00) Oil rents > median 0.0286∗∗∗ (2.77) Non-oil rents 0.0109 0.0109 (0.87) (0.87) Observations 113 113 113 113 113 113 113 Hansen J 2.243 4.016 5.465 5.719 5.855 p-value 0.134 0.260 0.141 0.126 0.119 Kleibergen-Paap 10.112 11.076 11.084 11.185 11.494 p-value 0.006 0.026 0.026 0.025 0.022 p -values in parentheses. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01. All specifications include education, the logarithm of total population, an openness measure, investment, and a measure of ethnic fractionalization. In Column (4), inflation and its standard deviation, as well as a measure of undervaluation, in Column (5) life expectancy and latitude are added. Overall, the results of the long-term cross-sectional analysis confirm the results drawn from the panel setting, and most importantly, even after instrumentation, the rule of law measure and the 17 corruption index are significantly related to GDP per capita growth. 6 Predictions In-sample fit The in-sample predictions of the growth rates of Nigeria based on the different specifications in Tables 2 and A.5 are graphically represented in Figures A.2 and A.3 in the Ap- pendix. Relying on this graphical analysis, specifications (5) to (7) in Table 2 provide the best in-sample fit of Nigeria’s growth rates. This seems to be mainly driven by the prediction of the spike in growth in the period 2000-200415 . Specification (7) is chosen to serve as the base for the out-of-sample growth predictions, because it contains the finest decomposition of the natural resource variable. General assumptions GDP per capita growth rates are predicted for six five-year periods starting in 2015-19, such that the last prediction period is 2040-44. The high degree of uncertainty about the future development of covariates such as investment, trade or government expenditures, is accommodated by the definition of different scenarios. Common to all scenarios is the assumption that the fixed effect of Nigeria is moderately improving over time.16 Figure A.4 in the Appendix shows the magnitude of the fixed effects of all countries in the sample. Fixed effects can be understood as country-specific deviations from the overall constant of the regression. Nigeria’s fixed effect is strikingly low, and among all the countries in the sample, only three countries have lower fixed effects, among them Liberia as another Sub-Saharan African country. Assumptions concerning changes of fixed effects over time reflect beliefs about the development of countries relative to each other. In other words, if income convergence is assumed, fixed effects are modeled to converge to each other. We assume, that Nigeria’s fixed effect will increase to -3 until 2050, which is a level that is close to Burkina Faso (-3.7), the Senegal (-3.2), Indonesia (-2.9) or Cote d’Ivoire (-2.4) in the current estimation. In addition, in each scenario GDP per capita is updated, i.e. the GDP per capita growth rate in (t − 1) is used in combination with the level of GDP per capita in (t − 1) to calculate GDP per capita in period t. The future development of the remaining covariates differs across scenarios, and the underlying assumptions are explained in detail below. Scenario 0: This baseline scenario is rather pessimistic, and assumes that oil rents in GDP remain at the low level that is predicted for 2015-19 until the end of the prediction period, 2040- 44. Expected oil rents for 2015-19 are based on the World Bank’s Commodity Price Forecasts for 15 The high average growth rate in this period comes from a growth rate of close to 30% in 2004, as reported in the World Bank’s World Development Indicators, but also by the United Nations (official data) and by the National Bureau of Statistics in Nigeria. 16 See for example IIASA (2015). 18 10 5 0 -5 1980 2000 2020 2040 Year Actual Predicted Scenario 0 Scenario 1 Scenario 2 Scenario 3 Figure 5 – Predicted growth rates of per capita GDP for Nigeria: Scenarios 0–3 19 this period. All other covariates remain at the levels of the (last observed) period 2010-14 and are summarized in Table 4. Table 4 – Values of covariates in Scenario 0 Variable Value assumed for all periods between 2015-2044 Share of upper secondary and tertiary educated in the labour force 34.0% Government consumption as share of GDP 9.0% Investment as share of GDP 15% Openness indicator -12.8 Inflation 26.6 Oil rents as share of GDP 8.2% Non-oil natural resource rents as share of GDP 2.8% Rule of law 17.4 Corruption -0.3 The resulting forecasts of GDP per capita growth rates are graphically displayed in Figure 5.17 The dotted line corresponds to Scenario 0. Under this scenario, growth rates are projected to rise only moderately as compared to the current levels. Low oil prices, which are assumed to remain at the low level over the next decades, and stagnating human capital stocks, investment, and institutional quality keep growth prospects below 3% until 2040-44. The top left graph in Figure A.7 in the Appendix shows the contributions of the main explanatory variables to the growth predictions in Scenario 0, and to what extent they compensate for the significantly negative fixed effect of Nigeria. Inherent to the underlying assumptions is that the contributions remain constant over time, and the variation in the expected growth rate results from income convergence alone. The factors that contribute most to Nigeria’s growth rates are human and physical capital, oil and non-oil natural resource rents, institutional quality, and corruption.18 Scenario 1: That scenario differs from the baseline scenario with respect to the underlying expected oil rents in GDP. The future development of oil rents is tied to the oil price forecasts of the World Bank’s Commodity Price Forecasts (see Figure A.5 in the Appendix). All other covariates remain at their 2010-14 level as displayed in Table 4, including the non-oil natural resource rents. 17 Growth rates used in the regressions as well as for the predictions are growth rates of per capita GDP, that correspond to yearly averages over five-year periods. With an annual population growth of roughly 2.7%, GDP growth rates are considerably higher. 18 It might seem puzzling that the institutional environment in Nigeria positively contributes to growth. This finding is explained by the construction of the underlying indices: they are measured as deviations from country- specific means. As Nigeria’s institutional quality as well as corruption levels improved since the 1980s and 1990s, the current deviation from the mean is positive (negative) for the rule of law (corruption) index. If this improvement did not happen, growth rates were predicted to be considerably lower. 20 4 4 3 3 Change in contribution Change in contribution 2 2 1 1 0 0 -1 -1 2015 2020 2025 2030 2035 2040 2015 2020 2025 2030 2035 2040 4 Change in contribution 1 0 -12 3 2015 2020 2025 2030 2035 2040 Education Gov. cons. Investment Openness Oil rents Non-oil rents Rule of law Corruption Figure 6 – Change in the contributions of covariates to predicted GDP per capita growth rates: Scenario 0 vs. 1 (top left), Scenario 1 vs. 2 (top right), and Scenario 2 vs. 3 (bottom left) Projected growth rates of GDP per capita based on this scenario are moderately higher than in Scenario 0, and by 2040-44, they are expected to be close to 4%. The better part of the difference between Scenarios 0 and 1 is materialized not before the end of the prediction period, however. The top left graph in Figure 6 shows that inclining oil rents in GDP drive the difference to the baseline projections. The expected recovery of the oil prices gradually raises the growth prospects of the country, assuming that other factors remain constant. Most importantly, this improvement can only be materialized, if institutional quality and corruption are not worsening simultaneously. This scenario certainly suggests that recovering oil prices alone are not sufficient for noteworthy and sustainable improvements of GDP per capita growth rates. Scenario 2: Growth predictions based on Scenario 2 are more optimistic, and the underlying assumptions can be found in Table 5. The share of upper secondary and tertiary educated indi- viduals in the working age population follows the medium scenario of the IIASA/VID Education 21 projections (Lutz and Butz, 2014). Government consumption is assumed to fall by 5.6% per period, as during 2000-14, investment is assumed to increase by 10% per period such that it reaches a share of 26.7% by 2040-44. The openness index is expected to gradually improve back to the level of 1990-2000, and inflation is assumed to remain at a comparable high level of 26.6. The projections of oil rents in GDP follow the oil price forecasts of the World Bank’s Commodity Price Forecasts as in Scenario 1, but in this scenario also non-oil resource rents develop in line with the respective price forecasts. This implies a further decline in 2015-19, and a gradual increase to present values until 2040-44, and altogether this is less favorable for the country than a continuation of the current trend, but at the same time a more realistic assumption. The rule of law and corruption indicators remain at the 2010-14 level. Table 5 – Values of covariates in Scenario 2 Year Educ. Gov. Invest- Openness Inflation Oil, 4th Non-oil Rule of Corruption cons. ment quart. rents law 2015 40 7.53 16.61 -12.10 26.63 8.18 1.56 17.41 -0.29 2020 46 7.10 18.27 -10.91 26.63 10.31 1.67 17.41 -0.29 2025 52 6.70 20.10 -8.88 26.63 11.79 1.80 17.41 -0.29 2030 58 6.32 22.11 -5.43 26.63 13.48 1.95 17.41 -0.29 2035 63 5.96 24.32 0.44 26.63 15.41 2.11 17.41 -0.29 2040 69 5.62 26.75 10.43 26.63 17.61 2.29 17.41 -0.29 Growth forecasts under these assumptions are significantly revised upwards as compared to Scenarios 0 and 1, and are projected to cross the 5% threshold in 2030. Particular to this scenario is that as opposed to the first two scenarios, the contributing factors can—at least to a certain degree—be influenced by policies. The top right graph in Figure 6 displays how the contribution of various factors to GDP per capita growth differs between Scenarios 1 and 2. In particular the influence of human and physical capital accumulation is striking. Also the assumed slight reduction in government consumption and the gradual improvement of the openness measure contribute positively to growth. Merely the tying of non-oil natural resource rents to commodity price forecasts causes a lower contribution of that factor to growth. This latter assumption, however, is important, and more realistic than the presumption of constant non-oil natural resource rents in the upcoming decades. Scenario 3: This scenario differs from Scenario 2 with respect to the adopted institutional char- acteristics. While in Scenario 2, the rule of law and corruption were assumed to remain constant at the current levels, Scenario 3 shows predicted growth rates that could be materialized given moderate improvements of the institutional quality and the corruption indicators (for a graphical representation of the expected development, please see Figure A.6 in the Appendix). 22 Table 6 – Predicted GDP per capita growth rates for different scenarios Period Scenario 0 Scenario 1 Scenario 2 Scenario 3 2010-14 2.94 2.94 2.94 2.94 2015-19 1.60 1.60 1.67 1.86 2020-24 2.15 2.41 3.22 3.59 2025-29 2.54 2.94 4.44 4.99 2030-34 2.80 3.35 5.51 6.22 2035-39 2.95 3.67 6.42 7.30 2040-44 3.03 3.92 7.34 8.38 Corresponding growth projections further improve, and highlight the potential inherent to enhancements of the quality of institutions and a reduction in corruption. Summary For the projections of Nigeria’s GDP per capita growth rates, we rely on the panel growth regression results presented in column (7) Table 2, as this specification exhibits the best in-sample fit and at the same time contains decomposed natural resource variables. Predicted growth rates are yearly growth rates of per capita GDP, and correspond to a time window of five years. In order to accommodate the high degree of uncertainty concerning the future developments of the explanatory variables, four different scenarios are defined. Table 6 summarizes the predicted growth rates between the 2015-19 and 2040-44 for the four scenarios, and below the key implications are highlighted. 1. Scenario 0 effectively simulates an extrapolation of the status-quo in combination with continuously low oil prices. Predicted GDP per capita growth rates increase moderately over time due to income convergence, but reach merely 3% by 2040-44. 2. In Scenario 1, expected oil rents in GDP follow the path of oil price projections, which suggest a moderate recovery after the sharp drop between 2015 and 2019. Growth rates are predicted to improve as compared to the baseline scenario, but a rise above 4% appears not to be feasible until 2040-44. Based on the forecasts, a recovery of oil prices alone is not sufficient for obtaining sustainable growth rates of per capita GDP above 4%, even if they are not accompanied by a reduction in the quality of institutions. 3. Growth predictions based on Scenario 2 exceed 5% from the period 2030-34 onwards. The underlying assumptions are moderate improvements of human and physical capital accumu- lation, an increasing openness of the economy, a reduction of government consumption, as well as non-oil natural resource rents that follow commodity price forecasts. The largest 23 contribution to the predicted growth rates come from education and investment, both areas that can be tackled by economic policy. 4. An additional improvement of the institutional quality, measured by indices of the rule of law and corruption, leads to a further rise in growth prospects by up to 1%-point in Scenario 3. Addressing present shortcomings in the transparency and enforceability of law and the access of civilians to justice and secure property rights, as well as delimiting corruption on all levels, appear to be fruitful strategies for enhancing Nigeria’s growth prospects. 7 Conclusions In this empirical assessment, the determinants of GDP per capita growth are studied using data of approximately 150 developing and developed countries during 1970 and 2014, focusing in particular on the role of natural resources, their interactions with institutional quality, and specific impacts in the country of Nigeria. Blessed by an enormous wealth of natural resources but at the same time afflicted by stagnating GDP growth rates, Nigeria is a prominent example of an economy that lacks economic development in spite of its resource-abundance. Based on the econometric results, Nigeria’s growth prospects are assessed under the assumption of different scenarios. Relying on the global sample, the empirical findings suggest that a sound institutional environ- ment, measured by an index of the rule of law, is associated with higher GDP per capita growth rates. In addition, the impact of natural resources on GDP per capita growth turns positive, once natural resources are interacted with the rule of law. The effect of natural resources thus depends on the quality of institutions, and resources can be a blessing in countries with transpar- ent and enforceable law, secure property rights, as well equality before the law of all citizens, and their freedom of movement and religion. The estimates further suggest that the growth impact of natural resources differs by the level of resource-dependence of the countries. Countries that are highly dependent on resources obtain lower growth-returns than countries that have comparably low shares of resource rents in GDP. Shifting the focus to Nigeria reveals that improvements of institutional quality—measured by the rule of law as well as by a corruption index—have a particular beneficial effect on growth rates of the country. In particular a reduction of corruption not only has a direct influence on economic growth, but also an indirect one through the improvement of the growth-enhancing potential of natural resources. If the level of corruption is held constant, higher natural resource rents are estimated to increase Nigeria’s growth rates. When omitting the corruption variable, the impact of natural resources appears to be detrimental to the growth of the economy. This suggests a strong link between resource endowments and the quality of institutions in the country, and a stabilization of the institutional environment at an improved level should be of high priority. 24 An additional important result is found with respect to the over-/undervaluation of Nigeria’s currency. Undervaluation is positively linked to GDP growth, and the long periods of overvaluation significantly reduced the country’s growth rates. Similar to institutional quality, there is an addi- tional indirect effect of that factor: when holding the degree of overvaluation constant, the impact of natural resources on growth increases. Such findings are in line with Dutch-disease effects, and managing the country’s real exchange rate can not only positively contribute to growth directly, but also indirectly by improving the growth-effect of resources. As especially the institutional variables are prone to endogeneity biases, similar growth regres- sions have been estimated using instrumental variable estimators at the cross-sectional level. The positive and negative growth impacts of the rule of law and corruption, respectively, are confirmed when using instrumental variables suggested in the literature. The results of the panel regressions are then used to assess Nigeria’s future growth potential based on different scenarios of the covariates. The stabilization of oil prices at a higher level than currently observed seems not sufficient for sustainable growth rates in the country. A steady accumulation of human as well as physical capital are the main ingredients for reaching an estimated GDP per capita growth rate of 5%, and a slight improvement of the institutional quality indicators is estimated to add another percentage point. 25 References Acemoglu, D., S. Johnson, and J. Robinson (2001): “The Colonial Origins of Comparative Development: An Empirical Investigation,” American Economic Review, 91(5), 1369–1401. Atkinson, G., and K. Hamilton (2003): “Savings, Growth and the Resource Curse Hypothesis,” World Development, 31(11), 1793–1807. Badinger, H., and E. Nindl (2014): “Globalisation and Corruption, Revisited,” The World Economy, 37(10), 1424–1440. Baland, J.-M., and P. Francois (2000): “Rent-seeking and resource booms,” Journal of De- velopment Economics, 61(2), 527–542. Bhattacharyya, S., and R. Hodler (2010): “Natural resources, democracy and corruption,” European Economic Review, 54(4), 608–621. Brunnschweiler, C. N. (2008): “Cursing the Blessings? Natural Resource Abundance, Institu- tions, and Economic Growth,” World Development, 36(3), 399–419. Bulte, E. H., R. Damania, and R. T. Deacon (2005): “Resource intensity, institutions, and development,” World Development, 33(7), 1029–1044. Campos, N. F., R. D. Dimova, and A. Saleh (2010): “Whither Corruption? A Quantitative Survey of the Literature on Corruption and Growth,” SSRN Scholarly Paper ID 1718935, Social Science Research Network, Rochester, NY. Coppedge, M., J. Gerring, S. I. Lindberg, S.-E. Skaaning, J. Teorell, E. Tzelgov, Y.- t. Wang, D. Altman, M. Bernhard, M. S. Fish, A. Glynn, A. Hicken, C. H. Knutsen, K. McMann, J. Staton, and B. Zimmermann (2015): “Varieties of Democracy Methodology v5,” Varieties of democracy project: Project documentation paper series, University of Gothen- burg, Kellogg Institute. Dreher, A., and S. Langlotz (2015): “Aid and Growth. New Evidence Using an Excludable Instrument,” CESifo Working Paper Series 5515, CESifo Group Munich. Easterly, W., and R. Levine (2003): “Tropics, germs, and crops: how endowments influence economic development,” Journal of Monetary Economics, 50(1), 3–39. Egger, P., and H. Winner (2005): “Evidence on corruption as an incentive for foreign direct investment,” European Journal of Political Economy, 21(4), 932–952. Feenstra, R. C., R. Inklaar, and M. P. Timmer (2015): “The Next Generation of the Penn World Table,” American Economic Review, 105(10), 3150–82. 26 Gylfason, T. (2001): “Natural resources, education, and economic development,” European Economic Review, 45(4–6), 847–859. Gylfason, T., T. T. Herbertsson, and G. Zoega (1999): “A Mixed Blessing,” Macroeconomic Dynamics, 3(02), 204–225. Hall, R. E., and C. I. Jones (1999): “Why do Some Countries Produce So Much More Output Per Worker than Others?,” The Quarterly Journal of Economics, 114(1), 83–116. Hodler, R. (2006): “The curse of natural resources in fractionalized countries,” European Eco- nomic Review, 50(6), 1367–1386. IIASA (2015): “SSP Data Set,” Supplementary material, Institute for Applied Systems Analysis (IIASA) and Vienna Institute of Demography (VID). Lane, P. R., and A. Tornell (1996): “Power, growth, and the voracity effect,” Journal of Economic Growth, 1(2), 213–241. Leite, C. A., and J. Weidmann (1999): “Does Mother Nature Corrupt? Natural Resources, Corruption, and Economic Growth,” SSRN Scholarly Paper ID 259928, Social Science Research Network, Rochester, NY. Lutz, W., and W. P. Butz (2014): World Population and Human Capital in the Twenty-First Century. Oxford University Press. Mehlum, H., K. Moene, and R. Torvik (2006): “Institutions and the Resource Curse*,” The Economic Journal, 116(508), 1–20. ´ lveda (2006): “Corruption, growth and political regimes: Cross coun- ´ndez, F., and F. Sepu Me try evidence,” European Journal of Political Economy, 22(1), 82–98. Nunn, N., and N. Qian (2014): “US Food Aid and Civil Conflict,” The American Economic Review, 104(6), 1630–1666. Raggl, A. K. (2014): “Economic growth in Ghana : determinants and prospect,” World Bank Policy Research Working Paper WPS6750, The World Bank. Rodrik, D. (2008): “The Real Exchange Rate and Economic Growth,” Brookings Papers on Economic Activity, 2008(2), 365–412. Sachs, J. D., and A. M. Warner (1995): “Natural Resource Abundance and Economic Growth,” Working Paper 5398, National Bureau of Economic Research. 27 (1997): “Sources of Slow Growth in African Economies,” Journal of African Economies, 6(3), 335–376. (2001): “The curse of natural resources,” European Economic Review, 45(4–6), 827–838. Sala-i-Martin, X., and A. Subramanian (2003): “Addressing the Natural Resource Curse: An Illustration from Nigeria,” Working Paper 9804, National Bureau of Economic Research. (2013): “Addressing the Natural Resource Curse: An Illustration from Nigeria,” Journal of African Economies, 22(4), 570–615. Torvik, R. (2001): “Learning by doing and the Dutch disease,” European Economic Review, 45(2), 285–306. (2002): “Natural resources, rent seeking and welfare,” Journal of Development Economics, 67(2), 455–470. van der Ploeg, F. (2011): “Natural Resources: Curse or Blessing?,” Journal of Economic Literature, 49(2), 366–420. Vial, V., and J. Hanoteau (2010): “Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence,” World Development, 38(5), 693–705. Werker, E., F. Z. Ahmed, and C. Cohen (2009): “How Is Foreign Aid Spent? Evidence from a Natural Experiment,” American Economic Journal: Macroeconomics, 1(2), 225–244. 28 A Appendix Table A.1 – Long-run GDP per capita growth, its standard deviation, natural resource and oil rents for NGA and various country-aggregates GDP per St.D. GDP per Natural Oil rents N capita growth capita growth resource rents NGA 1.06 7.47 43.44 41.10 1 Sub-Saharan Africa 0.87 5.23 13.70 3.91 41 Oil rents > 0 1.79 4.07 16.23 12.86 59 No oil rents 1.52 4.36 9.11 0.00 60 All 1.65 4.22 12.64 6.37 119 2.5 2 Growth impact (%-points) .5 1 0 1.5 2010 2020 2030 2040 year Column 3 Column 4 Column 5 Figure A.1 – Contribution to GDP per capita growth: improvement of the undervaluation index back to 0.38 (level 1965-69) until 2040-44 29 Table A.2 – Variable description: Panel setting Variable Description Source Dependent variable Growth in per capita GDP, 5-year averages WDI Explanatory variables Log GDP per capita Log of GDP per capita (constant 2005$), measured at start of each 5-yr WDI period Education Share of upper secondary and tertiary educated among the 15 to 64 year IIASA olds, 5-year average Governm. cons. General government final consumption expenditure (% of GDP), 5-year av- WDI erage Investment Gross fixed capital formation (% of GDP), 5-year average WDI Openness Imports plus exports of goods and services (% of GDP) filtered for its rela- WDI tion to log(area) and log(population), 5-year average Inflation Inflation, GDP deflator (annual %), 5-year average WDI Rule of law Equality before the law and civil liberties index (0, 100), deviations from V-Dem country-mean, 5-year average Corruption Index of political corruption, runs from less corrupt to more corrupt, (0, V-Dem 100), deviations from country-mean, 5-year average Natural resource rents Natural resources rents (% of GDP), 5-year averages WDI Oil rents Oil rents (% of GDP), 5-year averages WDI Mineral rents Mineral rents (% of GDP), 5-year averages WDI Forest rents Forest rents (% of GDP), 5-year averages WDI Non-oil rents Natural resource rents excluding oil (% of GDP), 5-year averages WDI Oil rents, 1st quartile Oil rents if oil rents belong to the lowest quartile in corresponding period, 0 otherwise (i.e. interaction of the oil rents variable with a dummy 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 corresponding period, 0 otherwise Note: The Variables oil rents 1st, 2nd, 3rd and 4th quartile add up to the variable oil rents Log(undervaluation) Measure of currency undervaluation based on Rodrik (2008) using the price PWT level of GDP and GDP per capita, WDI: World Development Indicators (2015), The World Bank; IIASA: IIASA-VID dateset on educational attainment (Lutz and Butz, 2014); V-Dem: Varieties of Democracy dataset (Coppedge, Gerring, Lindberg, Skaaning, Teorell, Tzelgov, Wang, Altman, Bernhard, Fish, Glynn, Hicken, Knutsen, McMann, Staton, and Zimmermann, 2015); PWT: Penn World Tables Version 8.1 (Feenstra, Inklaar, and Timmer, 2015) 30 Table A.3 – Variable description: Cross-sectional setting Variable Description Source Dependent variable Growth in per capita GDP, average 1980-2014 WDI Explanatory variables Log GDP per capita Log of GDP per capita (constant 2005$), initial value (1980) WDI Education Share of upper secondary and tertiary educated among the 15 to 64 year IIASA olds, initial value (1980) Investment Gross fixed capital formation (% of GDP), 5-year average WDI Openness Imports plus exports of goods and services (% of GDP) filtered for its WDI relation to log(area) and log(population), 5-year average Inflation Inflation, GDP deflator (annual %), 5-year average WDI Rule of law Equality before the law and civil liberties index, (0, 100), average 1980- V-Dem 2014 Corruption Index of political corruption, runs from less corrupt to more corrupt, V-Dem (0,100), average 1980-2014 Natural resource rents Natural resources rents (% of GDP), 5-year averages WDI Oil rents Oil rents (% of GDP), 5-year averages WDI Mineral rents Mineral rents (% of GDP), 5-year averages WDI Forest rents Forest rents (% of GDP), 5-year averages WDI Non-oil rents Natural resource rents excluding oil (% of GDP), 5-year averages WDI Oil rents, below median Oil rents if oil rents belong to the lowest quartile in corresponding period, 0 otherwise (i.e. interaction of the oil rents variable with a dummy variable indicating the first quartile) Oil rents, above median Oil rents if oil rents belong to the 2nd quartile in corresponding period, 0 otherwise Log(undervaluation) Measure of currency undervaluation based on Rodrik (2008) using the price PWT level of GDP and GDP per capita, Instruments EurFrac Fraction of the population speaking one of the major Western European HJ languages (English, Spanish, French, Portuguese, German) as a mother tongue EngFrac Fraction of the population speaking English as a mother tongue HJ WDI: World Development Indicators (2015), The World Bank; IIASA: IIASA-VID dateset on educational attainment (Lutz and Butz, 2014); V-Dem: Varieties of Democracy dataset (Coppedge, Gerring, Lindberg, Skaaning, Teorell, Tzelgov, Wang, Altman, Bernhard, Fish, Glynn, Hicken, Knutsen, McMann, Staton, and Zimmermann, 2015); HJ: Dataset used by Hall and Jones (1999); PWT: Penn World Tables Version 8.1 (Feenstra, Inklaar, and Timmer, 2015) 31 Table A.4 – Country and time coverage in the panel estimations (the years indicate the first year of the 5-year windows) Country MIN MAX t Conutry MIN MAX t Country MIN MAX t East Asia & Pacific Spain 1970 2010 9 South Asia Australia 1970 2010 9 Sweden 1970 2010 9 Afghanistan 2010 2010 1 Cambodia 1990 2010 5 Switzerland 1980 2010 7 Bangladesh 1980 2010 7 China 1970 2010 9 Tajikistan 1995 2010 4 Bhutan 1980 2010 7 Indonesia 1975 2010 8 Turkey 1970 2010 9 India 1970 2010 9 Japan 1970 2010 9 Turkmenistan 1995 2010 3 Maldives 2000 2005 2 Korea, Rep. 1970 2010 9 Ukraine 1995 2010 4 Nepal 1975 2010 8 Lao PDR 1985 2010 4 United Kingd. 1970 2010 9 Pakistan 1970 2010 9 Malaysia 1970 2010 9 Uzbekistan 2010 2010 1 Sri Lanka 2010 2010 1 Mongolia 1980 2010 7 New Zealand 1975 2010 8 Latin America & Caribbean Philippines 1970 2010 9 Argentina 1970 2010 8 Sub-Saharan Africa Thailand 1970 2010 9 Barbados 2010 2010 1 Benin 1980 2010 7 Timor-Leste 2000 2000 1 Bolivia 1970 2010 9 Botswana 2010 2010 1 Vanuatu 1980 2010 7 Brazil 1970 2010 9 Burkina Faso 1975 2010 8 Vietnam 1985 2010 6 Chile 1970 2010 9 Burundi 1970 2010 9 Colombia 1970 2010 9 Cabo Verde 2005 2010 2 Europe & Central Asia Costa Rica 1970 2010 9 Cameroon 1975 2010 8 Albania 1980 2010 7 Cuba 1970 2010 9 Cen. Afr. Rep. 1975 2010 8 Armenia 1995 2010 4 Dominican R. 1970 2010 9 Chad 2000 2010 3 Austria 1970 2010 9 Ecuador 1970 2010 9 Comoros 1980 2010 7 Azerbaijan 1995 2010 4 El Salvador 1970 2010 9 Congo, D. Rep. 1970 2010 9 Belarus 1990 2010 5 Guatemala 1970 2010 9 Congo, Rep. 1970 2010 9 Belgium 2000 2010 3 Guyana 1970 2010 9 Cote d’Ivoire 1970 2010 9 Bosnia & Herz. 2000 2010 3 Honduras 1970 2010 9 Eritrea 2010 2010 1 Bulgaria 1980 2010 7 Jamaica 1970 2010 9 Ethiopia 2010 2010 1 Croatia 1995 2010 4 Mexico 1970 2010 9 Gabon 1970 2010 9 Cyprus 1975 2010 8 Nicaragua 1970 2010 7 Gambia, The 1980 2010 7 Czech Republic 1990 2010 5 Panama 1980 2010 7 Ghana 1970 2010 9 Denmark 1970 2010 9 Paraguay 1990 2010 5 Guinea 1985 2010 6 Estonia 1995 2010 4 Peru 1970 2010 9 Guinea-Bissau 1975 2010 8 Finland 1970 2010 9 Suriname 1975 2005 7 Kenya 1970 2010 9 France 1970 2010 9 Trinidad & T. 1970 2010 9 Lesotho 1970 2010 9 Georgia 1995 2010 4 Uruguay 1970 2010 9 Liberia 2000 2010 3 Germany 1970 2010 9 Venezuela, RB 1970 2010 9 Madagascar 1970 2010 9 Greece 1970 2010 9 Malawi 1970 2010 9 Hungary 1990 2010 5 Middle East & North Africa Mali 1970 2010 8 Iceland 1995 2010 4 Algeria 1970 2010 9 Mauritania 2010 2010 1 Ireland 1970 2010 9 Egypt, Arab Rep. 1970 2010 9 Mauritius 1975 2010 8 Italy 1970 2010 9 Iran, Islamic Rep. 1970 2010 9 Mozambique 1980 2010 7 Kazakhstan 1990 2010 5 Iraq 2000 2010 3 Namibia 1980 2010 7 Kyrgyz Rep. 1995 2010 4 Israel 2010 2010 1 Niger 1980 2010 7 Latvia 1995 2010 4 Jordan 1975 2010 8 Nigeria 1980 2010 7 Lithuania 2000 2010 3 Lebanon 1990 2010 5 Rwanda 1970 2010 9 Macedonia 1990 2010 5 Morocco 1970 2010 9 Senegal 1970 2010 9 Moldova 1995 2010 4 Qatar 2000 2010 3 Sierra Leone 1980 2010 7 Montenegro 2005 2010 2 Saudi Arabia 1970 2010 9 South Africa 1970 2010 9 Netherlands 1970 2010 9 Syrian Arab Rep. 1970 2005 8 Sudan 2010 2010 1 Norway 1970 2010 9 Tunisia 1970 2010 9 Swaziland 1970 2010 9 Poland 1990 2010 5 Tanzania 1990 2010 5 Portugal 1970 2010 9 North America Togo 2010 2010 1 Romania 1990 2010 5 Canada 1970 2010 9 Uganda 1980 2010 7 Russian Fed.n 1990 2010 5 United States 1970 2010 9 Zambia 1970 2010 6 Serbia 2005 2010 2 Zimbabwe 1975 2010 8 Slovak Rep. 1990 2010 5 Slovenia 1995 2010 4 32 Table A.5 – Determinants of growth per capita: Panel fixed effects estimations, additional results (1) (2) (3) (4) (5) Log(GDP per capita) -2.980∗∗∗ -3.373∗∗∗ -3.155∗∗∗ -3.238∗∗∗ -3.321∗∗∗ (0.000) (0.000) (0.000) (0.000) (0.000) Education 0.199∗∗ 0.0428 0.0434∗ 0.0357 0.0394 (0.013) (0.131) (0.097) (0.194) (0.152) Education x GDP per capita -0.0158∗∗ (0.036) Education x NGA 0.321∗∗∗ (0.000) Log(undervaluation) -0.223 0.0392 0.0450 (0.613) (0.925) (0.914) Log(underval) x NGA 6.157∗∗∗ 5.552∗∗∗ 5.517∗∗∗ (0.000) (0.000) (0.000) Governm. cons. -0.0828∗∗ -0.0955∗∗ -0.102∗∗ -0.110∗∗∗ -0.104∗∗ (0.025) (0.011) (0.018) (0.007) (0.011) Investment 0.148∗∗∗ 0.154∗∗∗ 0.150∗∗∗ 0.159∗∗∗ 0.161∗∗∗ (0.000) (0.000) (0.000) (0.000) (0.000) Openness 0.0126∗∗ 0.0125∗∗ 0.0154∗∗ 0.0139∗∗ 0.0139∗∗ (0.047) (0.038) (0.025) (0.035) (0.035) Inflation -0.00159∗∗∗ -0.00160∗∗∗ -0.00179∗∗∗ -0.00177∗∗∗ -0.00171∗∗∗ (0.000) (0.000) (0.001) (0.001) (0.002) Rule of law 0.0130 0.0170∗∗ 0.0198∗∗∗ 0.0172∗∗ 0.0154∗∗ (0.106) (0.014) (0.003) (0.016) (0.028) Natural res. 0.0432 (0.125) Nat. res. x NGA 0.0828∗∗∗ (0.009) Oil rents 0.0737 0.0711 0.0693 (0.107) (0.113) (0.137) Oil rents x NGA -0.394∗∗∗ -0.0480 -0.285∗∗∗ -0.255∗∗∗ (0.000) (0.323) (0.000) (0.000) Non-oil rents 0.0239 0.0277 0.0304 0.0229 (0.491) (0.427) (0.396) (0.491) Non-oil rents x NGA 1.480∗∗∗ 0.637∗∗∗ 1.715∗∗∗ 1.686∗∗∗ (0.000) (0.000) (0.000) (0.000) Oil rents, 1st qu. 5.517∗∗∗ (0.000) Oil rents, 2nd qu. 0.616∗∗∗ (0.003) Oil rents, 3rd qu. 0.144∗∗ (0.017) Oil rents, 4th qu. 0.0785∗ (0.096) Corruption 0.0307∗∗ (0.034) Observations 1000 1000 960 956 956 R2 0.307 0.303 0.305 0.305 0.315 N g 150 150 142 142 142 ∗ ∗∗ Each specification includes country and period fixed effects, as well as a constant. p -values in parentheses. p < 0.1, p < 0.05, ∗∗∗ p < 0.01. 33 10 10 GDP per capita growth GDP per capita growth 5 5 0 0 -5 -5 1980 1990 2000 2010 2020 1980 1990 2000 2010 2020 Year Year Actual In-sample fit Actual In-sample fit 10 10 GDP per capita growth GDP per capita growth 5 5 0 0 -5 -5 1980 1990 2000 2010 2020 1980 1990 2000 2010 2020 Year Year Actual In-sample fit Actual In-sample fit 10 10 GDP per capita growth GDP per capita growth 5 5 0 0 -5 -5 1980 1990 2000 2010 2020 1980 1990 2000 2010 2020 Year Year Actual In-sample fit Actual In-sample fit 10 GDP per capita growth 0 -5 5 1980 1990 2000 2010 2020 Year Actual In-sample fit 34 Figure A.2 – Actual GDP per capital growth rates in Nigeria vs. in-sample prediction based on columns (1) to (7) in Table 2 10 10 GDP per capita growth GDP per capita growth 5 5 0 0 -5 -5 1980 1990 2000 2010 2020 1980 1990 2000 2010 2020 Year Year Actual In-sample fit Actual In-sample fit 10 10 GDP per capita growth GDP per capita growth 5 5 0 0 -5 -5 1980 1990 2000 2010 2020 1980 1990 2000 2010 2020 Year Year Actual In-sample fit Actual In-sample fit 10 GDP per capita growth 0 -5 5 1980 1990 2000 2010 2020 Year Actual In-sample fit Figure A.3 – Actual GDP per capital growth rates in Nigeria vs. in-sample prediction based on columns (1) to (5) in Table A.5 35 ISL GBR IRL ITA FRA DNK FIN SWE ESP PRT NLD CYP NOR CHE BEL DEU AUT KOR JPN MDV QAT USA GRC NZL CAN LBN AUS URY MUS SVN LVA LTU TUR TTO MEX EST SVK BRA HRV CRI POL CUB SAU CHL HUN PAN ARG DOM ZAF NAM TMP SUR SLV JAM CZE COL TUN SWZ BIH MAR GAB GTM MYS ECU VUT THA MKD JOR BGR CHN VEN ARM ROM PER PRY BTN KHM HND GEO SYR CIV LAO DZA ALB BOL IDN COM BLR SEN AZE PAK EGY IRN BFA CMR BEN RWA IRQ MOZ MLI BGD KAZ KEN SLE IND RUS VNM UGA TZA GMB GUY PHL MDA LSO GHA GNB UKR COG CAF TCD NPL ZWE MWI MNG BDI ZAR MDG GIN NER NGA KGZ TJK LBR -10 -5 0 5 10 Fixed Effects Figure A.4 – Country fixed effects based on specification (7) in Table 2 36 50 40 30 20 10 1980 2000 2020 2040 year Figure A.5 – Oil rents as shares in GDP: historic and expectations for 2015-2044, Scenarios 1–3 30 2 1.5 20 1 10 .5 0 0 -10 -.5 1980 2000 2020 2040 1980 2000 2020 2040 year year Figure A.6 – Institutional indicators: historic and expectations for 2015-2044 in Scenario 3: Rule of law (left) and corruption (right) 37 10 10 8 Contribution to growth Contribution to growth 6 5 2 4 0 0 2015 2020 2025 2030 2035 2040 2015 2020 2025 2030 2035 2040 Education Gov. cons. Education Gov. cons. Investment Openness Investment Openness Inflation Oil rents Inflation Oil rents Non-oil rents Rule of law Non-oil rents Rule of law Corruption Corruption 20 15 Contribution to growth Contribution to growth 15 10 10 5 5 0 0 2015 2020 2025 2030 2035 2040 2015 2020 2025 2030 2035 2040 Education Gov. cons. Education Gov. cons. Investment Openness Investment Openness Inflation Oil rents Inflation Oil rents Non-oil rents Rule of law Non-oil rents Rule of law Corruption Corruption Figure A.7 – Contributions of covariates to predicted GDP per capita growth rates in Scenario 0 (top left), 1 (top right), 2 (bottom left) and 3 (bottom right) 38