Master thesis The resource curse revisited: Are weak initial institutions the reason countries fall prey to the resource curse?

(1)

UNIVERSITY OF GRONINGEN AND UNIVERSITY OF GÖTTINGEN

The resource curse revisited: Are weak initial

institutions the reason countries fall prey to the

resource curse?

Master thesis

Author:

Name: Bsrat Measho

Study Program: Double Degree Master International Economics & Business Student numbers : 1934163 (Groningen) and 11333435 (Göttingen)

E-Mail: b.measho@student.rug.nl (alternatively: bsratmeasho@gmail.com)

Supervisor:

Dr. Robbert Maseland

Faculty of Economics and Business University of Groningen

Co-assessor: Dr. Timo Trimborn

Faculty of Economic Sciences University of Göttingen

(2)

Abstract

It is a well established fact in the economics literature that natural resource abundant countries tend to grow slower on average than their resource deficient counterparts. However, there are also exceptions to this rule such as Botswana or Malaysia. These two countries managed to still escape the so-called “resource curse” despite being highly dependent on their natural resource industries. This thesis hypothesizes that initial institutions, i.e. the institutions prevailing when a country first starts to produce oil and/or gas, are decisive, determining whether a country experiences a natural resource curse or a blessing. The results show that there is indeed evidence for this theory. However, it seems that there is more to this story and that initial institutions are one of the country-specific conditions that are crucial.

(3)

1 Introduction

The “world’s richest little isle” and “the envy of the South Pacific”, this is how Nauru was characterized by The New York Times in 1982. Three decades later media outlets like The Economist (2014) state that the island’s history is “stranger than fiction” ascribing its current economic state to “greed, phosphate and gross incompetence”.

When the island gained independence in 1968 the overall economic prospects were positive as Nauru had large phosphate reserves, generating high export earnings. As a result, the mono-economy had the highest per capita income in the world (US Department of State, 2015). In subsequent years, phosphate reserves declined substantially as well as global demand for it, depressing the world price. This unfortunate turn of events, coupled with dubious investments made by the government, reduced Nauru’s wealth dramatically (The Economist, 2001). In 2006, phosphate mining had to be ceased as primary reserves were exhausted. Secondary phosphate reserves have been explored, however, it is believed that these will only last for a few more decades (UN-OHRLLS, 2015). Today, Nauru struggles to repay its mountain of debt and is highly dependent on development aid (e.g. in 2009: 50% of GDP; Asian Development Bank, 2014).

Former president Stephen Marcus stated the following at the 63rd session of the UN General Assembly (2008): “As a consequence of mismanagement and corruption, past administrations took Nauru from what was then a bright future, to the edge of collapse. In so doing, national reserves and assets have been lost and we have been left with an unmanageable burden of domestic and external debt. […] We therefore seek the understanding and consideration of those countries and institutions to which we owe money to allow debt forgiveness or major write downs .” Ironically, three years later the president resigned due to corruption allegations (BBC, 2015).

(6)

2 However, the Southeast Asian nation seemingly managed to turn the “resource curse” into a “resource blessing” by investing revenues into the development of other industries. This has led to a diversification of its economy. For example, the government invested in physical and social infrastructure in Penang, formerly a poor fishing area, and turned it into an electronics center (IMF, 2011). Why is it that some countries, such as Botswana and Malaysia, managed to avoid the “resource curse” and other countries such as Nauru, the Democratic Republic of Congo or Nigeria became victims of it? One possible explanation is that if the prevailing institutions at the time when the natural resource is first produced are weak then different groups in society can engage in rent-seeking and predatory regimes can thrive (Angwafo and Chuhan-Pole, 2011). A further deterioration of the country’s institutional environment may be the result. This has detrimental effects on a country’s economic development as good institutions are of utmost importance for economic growth according to numerous economists (e.g. Acemoglu, Johnson and Robinson, 2001). Therefore, if primary goods production leads to rent-seeking and corruption, which results in a deterioration of institutions, the so-called “resource curse” may be an “institutions curse”, negatively affecting economic growth. However, if institutions at the time when natural resources are produced for the first time are well developed, then the potential for rent-seeking is minimized and a deterioration of the institutional environment in subsequent years may not occur.

(7)

3 The thesis is structured as follows: in section two, the most influential literature pertaining to the natural resource curse as well as the “oil-impedes-democracy” hypothesis is reviewed. Furthermore, the two hypotheses and the respective theoretical constructs are presented. In section three the methodology and the data employed are described. Section four presents and discusses the results. Finally, section five concludes by summarizing the main findings, by providing policy recommendations and by suggesting potential future research.

2 Literature review and theory

2.1 Natural resource dependence and economic development

The relationship between natural resource dependence and economic development has been analyzed for several decades now. An early attempt was made in 1950 by the economists Raùl Prebisch and Hans Singer. According to them the terms of trade of natural resource exporting developing countries (the “periphery”) is always going to deteriorate at the expense of the terms of trade of industrialized economies (the “center”). This theory, known as the “Prebisch-Singer Hypothesis”, is based on “elasticity pessimism”: Primary products face low income elasticities, implying that rising incomes in the “center” result in less than proportional increases in demand. Therefore, if countries in the “center” generate high economic growth, countries in the “periphery” only benefit to a limited extent (Taylor, 1998). This hypothesis stood in stark contrast to what other, more mainstream oriented economists thought about economic development. According to them, it is vital that developing countries in early stages of development export natural resources in order to generate foreign exchange for their imports and debt repayment (Auty, 2001).

(8)

4 weak institutional environment (4). The so-called “Dutch disease” describes that high natural resource exports can cause an appreciation of the currency, which makes the country’s exports less competitive on global markets. At the same time imports become relatively cheap for the country, which leads to higher demand for it. This higher demand for imports can potentially lead to a de-industrialization of the domestic economy. However, one should not neglect the fact that high demand for the domestic exports generates savings, which can be invested. Also, capable governments can implement certain mechanisms to prevent the “Dutch disease” from occurring. For example, Botswana’s government has managed to do that by having high fiscal savings, by maintaining a current account surplus and by channeling investments to infrastructure and human capital. The high fiscal savings can act as a cushion in times of price volatility and the investments in infrastructure and the education system can ensure that Botswana remains competitive. In addition, by having high fiscal savings, current consumption remains low, which can help prevent high inflationary pressures (Angwafo and Chuhan-Pole, 2011). In order to dampen the negative effects of price fluctuations that can result in a deterioration of the terms of trade, Indonesia has successfully employed “fiscal, monetary and exchange rate mechanisms” (Lindauer and Romer, 1994). As natural resources can be depleted at one point (e.g. phosphate reserves in Nauru) fiscal savings should be used to promote diversification of the economy. As mentioned in the introduction, Malaysia was a resource dependent country in the 1960s but has managed to diversify its economy in subsequent years (IMF, 2011). Hence, the first three factors that may cause a resource curse can be prevented or at least mitigated if the government adopts the right policies.

(9)

5 In contrast, Botswana, which already had good institutions when it became independent from Great Britain, is often presented as a success story. Its institutional development was seemingly not adversely affected by British colonization, because the Tswana states maintained a high degree of autonomy and therefore, pre-colonial institutions were still existent at independence. After independence the son of the Tswana king became the new leader and embedded pre-colonial Tswana institutions into the new state rather than imposing new ones that were developed exogenously (Englebert, 2000). Acemoglu and Robinson (2010) confirm that in Botswana the leaders did not resort to neo-patrimonial policies but they implemented good institutions such as secure property rights and they provided public goods. Botswana has had one of the highest GDP per capita growth rates for several decades, despite having an economy that is highly dependent on the diamond industry. Acemoglu, Johnson and Robinson (2001) attribute this success to its institutional environment.

Countries such as Canada, Norway or the United States are natural resource producers as well and at the same time, they are all high income countries. In Africa, Botswana and South Africa belong to the best performing countries but at the time their economies are highly dependent on natural resources (Angwafo and Chuhan-Pole, 2011). Therefore, natural resources do not have to be a curse but the institutional environment in which natural resources are produced and exported seem to be crucial.

2.2 The role of institutions in resource rich countries

The negative relationship between natural resource abundance and institutions has been often attributed to the fact that several fuel producing countries are so-called “Rentier States”. The term was coined by Mahdavy (1970) and describes countries that receive substantial external rents (e.g. oil revenues) on a regular basis. Many scholars, e.g. Huntington (1991), explained that those Rentier States are not dependent on tax revenues from their citizens as high rents are generated from producing and exporting natural resources. As a result, citizens demand little representation in politics and autocratic governments can emerge. Dunning (2008) states that governments of Rentier States “may have less need to share political power more broadly”. It is, therefore, argued that governments of Rentier States possess high fiscal autonomy, which reduces the government’s accountability and eventually, the institutional environment deteriorates.

(10)

6 Middle East are highly dependent on rents. For example, the average fiscal reliance for Kuwait was 91 percent for the years 1963 to 2006. Similarly, other countries in the Middle East such as Oman (average: 82%, 1967 to 2006) and Saudi Arabia (average: 79%, 1948 to 2006) are also highly reliant on their oil industries. These countries, therefore, constitute prime examples for what Mahdavy (1970) named Rentier States. Another prime example is Nauru as its citizens do not have to pay any taxes (The Economist, 2001), while the government has been corrupt and has made very poor investment decisions with the high rents generated from phosphate exports. Apparently, government accountability has been at such a low level that funds can be misused to benefit politicians and other elites.

Several scholars found empirical evidence for the negative relationship between natural resources and institutions. Ross (2001) tested the theory with pooled cross-section data for 113 countries, covering the years 1971 to 1997. His results confirm the negative relationship. The novel aspect about his paper was that the so-called “oil-impedes-democracy hypothesis” also holds for countries not located in the Middle East. A more recent paper by De Rosa and Iootty (2012) examines countries for the period 1996-2010 with panel data using pooled OLS, fixed effects and system GMM. The scholars measure natural resource dependence as fuel exports as a share of total exports. The results indicate that higher fuel export dependence lowers government effectiveness and competition in the economy. Finally, a paper by Tsui (2010) examines the effect of oil discoveries on democracy. The data is analyzed using OLS and IV-2SLS. The IV estimation is added, because oil discovery might be an endogenous variable as democratic countries may be more likely to possess the technology needed to discover oil. Therefore, oil endowment is taken as instrumental variable for oil discovery. The dependent variable is change in democracy and the crucial independent variable is a democracy variable, one decade before discovery and three decades after the discovery. The idea is that then long-term developments of the institutional variable can be captured. The results show that a 100 billion barrel discovery deteriorates a country’s democracy value by about 20%.

(11)

7 managed and misused. Furthermore, in an environment with weak institutions there are little mechanisms to limit rent-seeking from different groups in society. As the Rentier States literature explains, such a situation (particularly if tax rates are very low or non-existent) may fortify autocracies. On the other hand, when a country’s institutions are good (e.g. high executive constraint) when it starts generating these high rents, then it is more likely that the rents are invested in a productive way, so that the economy and the citizens benefit from it. The reason is that there are mechanisms in place that prevent or at least limit rent-seeking and the emergence of autocracies.

H1: Countries with weak initial institutions experience a severe deterioration of their institutions as soon as they start producing natural resources due to rent-seeking from politicians, business elites or different ethnic groups. Extreme cases (“Rentier States”) often exempt their citizens from paying taxes, reducing their accountability and allowing them to become autocratic. In contrast, those countries that initially had good institutions will not suffer from any adverse effects as there are sufficient mechanisms to limit the negative effects such as rent-seeking and the emergence of non-democratic governments.

2.3 The role of institutions in the industrialization process

“Virtually every country that experienced rapid growth of productivity and living standards over the last 200 years has done so by industrializing” (Murphy, Shleifer and Vishny, 1989). Based on the quote, it seems as if countries have to industrialize in order to develop economically and socially. This would imply that countries highly dependent on natural resources cannot generate sustainable economic growth and high welfare for their citizens. In the case of Nauru the worst-case scenario occurred as phosphate reserves have almost been depleted and the island does not have an industrial sector to rely on. Besides the exhaustibility argument, there is the Prebisch-Singer Hypothesis, arguing that primary products have low income elasticities. It has also often been put forward that natural resource industries are highly capital intensive (e.g. oil and gas industry) and therefore does not absorb sufficient workers (Arezki and Nabli, 2012). The manufacturing sector could possibly create more employment for the local population and increase incomes.

(12)

8 good institutions such as secure property rights cannot only attract FDI, promoting industrialization but effective policies are also more likely to be implemented in such an environment. The latter argument could explain why some resource rich countries such as Norway or the United States still industrialized despite being rich in resources. The other extreme would be a resource-rich country with weak intuitions (e.g. an autocratic regime). It is likely that autocratic regimes are mainly interested in restoring their power, implementing neo-patrimonial policies that are conducive to reaching that goal. In Rentier States high rents are available which can be allocated to different groups in society to fortify the position of the autocrat. Therefore, restoring the status quo in such a setting seems to be the main concern and attempts to industrialize are less likely.

H2: A resource dependent country with weak institutions has an underdeveloped industrial sector as necessary reforms for industrialization are not implemented. On the other hand, the government of a resource dependent country with good institutions is more accountable and implements the needed reforms.

3 Data and Methodology

3.1 Data for hypothesis 1

3.1.1 Starting years and inclusion of countries

(13)

9 Table 1 in Appendix A provides information on the availability of natural resources and states if a country is an oil or gas producer. The table also provides information on the starting year of oil and gas production. Different sources were consulted in order to determine a starting year. The main ones were: “Campbell’s Atlas of Oil and Gas Depletion” (Campbell, 2013), the US Energy Information Administration, the OPEC and the datasets from Haber and Menaldo (2011) and Ross (2013). The datasets were very useful in order to get an indication of when production might have started. If both datasets as well as a third source reported the same starting year, then it was assumed to be correct. If two sources provided the same information and a third source indicated another starting year, then the information provided by the two sources was assumed to be correct. If all sources provided conflicting information, then the country was excluded from the analysis (e.g. Bangladesh). Sometimes the sources provided similar information (i.e. deviation by a few years), then the earlier year was assumed to be correct. However, this technique was not used often as it can produce erroneous results. Another reason to exclude a country is when data on institutions are not available for the starting year of production (e.g. Tunisia). As a final step, the starting years were compared to the ones suggested by Thieme, Lujala and Rød (2007). The starting years are mostly equal or similar (deviation of 1-2 years) except for the following countries: Albania, Australia, Canada, Cuba, France, Italy and New Zealand. For those countries, the discrepancy was much higher. For example, for France the scholars state that oil production started in 1735, therefore, a difference of almost 200 years. The difference for the other countries is not as large but it is still substantial. The production levels were presumably not very high in the early years, but it could still produce erroneous results to choose the later year.

(14)

10 Menaldo (2011) data. If both datasets offered the same starting years, then Ross (2013) is preferred as it offers data until 2011. This resulted in the following division (also indicated in the third column of Table 1 which can be found in Appendix A):

“Ross (2013) countries” (50 countries):

Afghanistan, Albania*, Algeria, Angola, Australia*, Azerbaijan, Belarus, Benin, Brazil, Chad, Congo, Dem. Rep., Congo, Rep., Croatia, Denmark, Equatorial Guinea, France*, Gabon, Georgia, Ghana, Greece, Guatemala, India, Kazakhstan, Korea, Rep., Kyrgyzstan, Libya, Malaysia, Mauritania, Mongolia, Morocco, Netherlands, New Zealand*, Nigeria, Norway, Philippines, Qatar, Saudi Arabia, Serbia, Sudan, Suriname, Tajikistan, Tanzania, Thailand, Timor-Leste (East Timor), Trinidad and Tobago, Turkmenistan, Ukraine, United Kingdom, Uzbekistan, Vietnam and Yemen.

“Haber and Menaldo (2011) countries” (23 countries):

Argentina, Austria, Bolivia, Canada*, Colombia, Ivory Coast, Cuba*, Ecuador, Germany, Hungary, Indonesia, Iran, Israel, Italy*, Jordan, Mexico, Mozambique, Papua New Guinea, Sweden, Senegal, United Arab Emirates, United States and Venezuela.

(Country*: an asterisk indicates a strong discrepancy between own starting years and starting years reported by Thieme, Lujala and Rød (2007); Country: a country name in italic indicates that oil and/or gas has already been produced in colonial times; Country: a country name that is underlined indicates that oil and/or gas has been produced while the country was still part of the Soviet Union, i.e. prior to 1991; Country: a country name that is bold (i.e. Germany and Yemen) indicates that the country was divided when oil or/and gas production started).

To make it clear: the data of those two datasets is not combined into one large dataset. The two groups are analyzed separately as the variables differ.

(15)

11 Gabon, India, Morocco, Nigeria, Qatar, Trinidad and Tobago and the United Arab Emirates. Secondly, several countries that used to belong to the Soviet Union are part of the sample. Here, similar to the former colonies, the year of independence (1991) is taken as starting year even though production started prior to 1991. The following 10 countries are affected: Azerbaijan, Belarus, Croatia, Georgia, Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, Ukraine and Uzbekistan. However, as the datasets from Haber and Menaldo (2011) and Ross (2013) do not report data prior to 1991 and institutions data is also not available for the period in which the country was part of the Soviet Union, it is the best solution to assume that the production year was 1991. Finally, another type of problem is when countries were divided or unified in the time period that is being analyzed. Two countries are affected: Germany and Yemen. Sudan and South Sudan split in 2011, so it does not have a major impact on the analysis as the last year that is included in the analysis is 2011. South Sudan is not included, only Sudan. When it comes to Germany, Haber and Menaldo (2011) state in their data appendix: “For 1945-1990 we treat West Germany [..] and Germany as identical”. Therefore, East Germany is not included in the analysis. When it comes to the institutions data, fer the purpose of this analysis only the data for West Germany is considered and then starting from 1990 the institutions data for the unified Germany is chosen. Finally, Yemen was divided into North and South Yemen prior to 1990. However, the starting year for oil is assumed to be 1986. That means that in the period 1986-1989 the country was still divided. As a consequence, the years prior to unification (1986-1989) are not included in the analysis. The analyses are conducted with and without those “problematic” cases, in order to see if it makes a difference.

3.1.2 Oil and gas data

As data on natural resource production for the last two centuries is not easily accessible, the analysis is based on two different replication datasets.

Ross (2013)

(16)

12 nominal. Furthermore, the variable Natural resource dependence (NRD) was generated by dividing Total fuel value nominal by GDP (current US $). The latter is part of the World Development Indicators (World Bank, 2015). The variable Natural resource dependence can only be generated for 36 countries as the GDP data is not available for years prior to 1960. Since the start of production activities has to be captured, the variable was only generated for countries for which GDP data was available for the starting year of oil or/and gas production. When using this dataset one should keep in mind the limitations. In a codebook (2013) for the dataset, Ross states that the data from the US Geological Survey was converted from production volume to weight. The US Geological Survey used a standard formula to do this but Ross indicates that oil varies in weight depending on the quality. This implies that in some cases the values are understated and in some cases overstated. Furthermore, Ross joined different data sources, which had some conflicting information. Those corrections may have produced erroneous figures.

Haber and Menaldo (2011)

The other dataset is from Haber and Menaldo (2011) that was used for their paper “Do natural resources fuel authoritarianism? A reappraisal of the resource curse”. It covers the years 1800-2006 and includes 168 countries. They estimate their variables based on data from the Oil and Gas Journal, the US Energy Information Administration, the US Geological Survey, British Petroleum, the World Bank, the United Nations and Angus Maddison. They also consulted national institutions such as the central banks, treasury ministries and statistical offices. From this dataset, the variables Oil income per capita and Gas income per capita are added to generate Total fuel income per capita. As the dataset from Haber and Menaldo (2011) includes population data, the variable Total fuel income per capita is multiplied by the population data to create the variable Total fuel income.

(17)

13 3.2 Data for hypothesis 2

World Bank data

The World Bank provides data on natural resource rents as a percentage of GDP. It can serve as a proxy for a country’s dependence on natural resources. The variable Total natural resource rents (% GDP) was acquired for 153 countries, depending on whether institutions data is available. The variable is the sum of oil rents, natural gas rents, coal rents, mineral rents and forest rents.

As rents are calculated as the difference between production value and production costs, it is likely that the data is inaccurate for some countries or some time periods. The reason is that in order to estimate production costs extensive information is needed, which is not always available. This is particularly the case for low income and war-torn countries.

The World Bank also provides data on manufacturing as a percentage of GDP, which is taken as a proxy for industrialization in this thesis. More specifically, the variable name is Manufacturing, value added (% of GDP). Based on the level variable, a change variable was computed (i.e. value of year 2 minus value of year 1, etc.)., which is used in some regressions as well.

3.3 Data on institutions (used for both hypotheses)

The institution variables for the main analyses, namely Exconst (“Executive Constraint”) and Polity2, were taken from the Polity IV dataset. The variables Autoc (“Autocracy”) and Democ (“Democracy”) are employed for the robustness checks in this thesis and were also obtained from the Polity IV dataset. This dataset was chosen as it covers not only a long time period (unlike e.g. the Worldwide Governance Indicators from the World Bank Group) but also many countries. More specifically, it covers the years 1800-2013 and includes 167 countries.

(18)

14 position compared to most other countries in the world (i.e. 166 countries) for a specific year and also over time.

The variable exconst ranges from 1 to 7, the variables autoc and democ both range from 0 to 10. However, the data for those three variables also includes the values 66”, 77” and “-88”. The users’ manual (by Marshall, Gurr and Jaggers, 2014) explains that those values indicate “foreign interruption” (“-66”), cases of “interregnum or anarchy” (“-77”) and “cases of transition” (“-88”). Those values were transformed following the technique described in the manual: “-66” should be treated as missing values; “-77” should be converted to the neutral value 0 and “-88” should be transformed into smooth transition values. For the latter, the users’ manual provides an example. Assume country X has a value of -7 in 1957 and a value of +5 in 1961. The change amounts to 12. Then the user’s manual suggests to allocate a -4 for 1958, a -1 for 1959, a +2 for 1960. As “-88” is a transition phase, those figures indicate a smooth transition from the value -5 to +7. This same technique was applied for “-88” values in the data for exconst, autoc and democ. The variable polity2 was already transformed by the creators of the variable using the technique described.

In order to test the first hypothesis an institutional change variable is needed. This variable was simply computed by subtracting the value from last year from this year’s value (i.e. value for year 2 – value for year 1, etc.).

Regarding the data’s limitations, one could criticize that for each year the worldwide mean is zero. This would imply that there are no trends in institutional change, e.g. institutions improve worldwide or deteriorate. Kaufmann et al. (2010) state that in fact there has not been a worldwide trend. Therefore, choosing a fixed mean of zero for each year is appropriate. 3.4 Methodology

A description of every variable used can be found in Table 1 in Appendix B. Summary statistics for selected variables can be found in Table 2 and 3 in Appendix B. Furthermore, correlation tabled for key variables can be found in Tables 4, 5, 6, in Appendix B as well. 3.4.1 First hypothesis

The basic regression equation for the first hypothesis looks as follows:

Δ Institutionsit = ɑi + ßNatural resource variableit + γ Initial institutionsit + δ (Natural

(19)

15 The dependent variable (Δ Institutions) is change in institutions from the start of oil or gas production until either 2006 or 2011, depending on the natural resource variable (independent variable) used. The basic analysis first includes the normalized Polity IV variables Exconst and Polity2 as dependent variables. In a robustness check, the normalized Polity IV variables Democ (measures to what extent a country is democratic) and Autoc (measures to what extent a country is autocratic) are used as dependent variables. Besides the change variables, the level variables are also employed for some regressions.

Regarding the independent variables, a variable for oil or/and gas production is included and represented by Natural resource variable in the above equation. In the main analysis, this variable is Natural resource dependence as defined as gas and oil value divided by GDP (see Appendix B, Table 1). In the robustness checks, the variable total fuel income enters the regressions. The interaction variable Natural resource variable x Initial institutions measures whether initial institutions in combination with natural resource production has an effect on institutional change. Finally, regional dummy variables enter the regressions.

The dataset consists of panel data. The main regressions as well as the robustness checks are analyzed using pooled OLS and country fixed effects to account for heterogeneity across countries. The respective model assumptions were tested. The results of those tests are not included in this thesis but can be provided if needed. As heterogeneity was detected for the pooled OLS regressions, they were estimated using robust standard errors (i.e. “vce (robust)” command in Stata). Regarding the fixed effects regression, often heterogeneity and autocorrelation was detected. This was fixed by using clustered or robust standard errors (i.e. “cluster ()” to fix heterogeneity and autocorrelation; “robust” to fix heterogeneity).

(20)

16 3.4.2 Second hypothesis

For the second hypothesis the regression equation is defined as follows:

Manufacturing (% GDP)it = ɑi + βTotal natural resource rentsit + γ Institutionsit + δ (Total

natural resource rents x Institutions) + εit

This regression is estimated using either levels of the variables or changes. The dependent variable is therefore either the level version of Manufacturing, value added (% GDP) or the change variable. When it comes to the independent variables, either the level version of Total natural resources rents (% GDP) enters the regression or the change variable (if the dependent variable is also defined as change variable). In this case the natural resource variable is not only confined to oil and gas. This variable computed by the World Bank includes, besides oil and gas rents, also coal, mineral and forest rents. The institutions variable is again either Exconst (“Executive Constraint”) or Polity2 in the main analyses and for the robustness checks the variables Autoc (“Autocracy”) and Democ (“Democracy”) are tested. Finally, an interaction term consisting of the natural resources variable and one institutions variable enters the regression.

The dataset consists of panel data. It is analyzed using pooled OLS and country fixed effects to account for country heterogeneity as with the first hypothesis. As with the first hypothesis, heterogeneity and autocorrelation were identified. If only heteroskedasticity was the problem robust standard errors were used. If additionally autocorrelation was detected, then clustered standard errors were used. The respective model assumptions were tested and results of those tests can be provided if desired.

4 Results

4.1 Hypothesis 1: Main results

This subsection first presents results from pooled OLS and fixed effects regression analyses pertaining to the first hypothesis. Finally, impulse response functions are presented showing the effect of natural resource production (“impulse”) on changes in Polity2 and changes in Executive Constraint (“response”).

(21)

17 excluded as they had very high leverage, high squared residuals values or both as in the case of Algeria (see Appendix C, Figure 1 for the leverage residual plot). Surprisingly, removing those four countries had a large impact on the results. From the residual plot one may also conclude that Timor-Leste should be removed. However, if Timor-Leste is excluded the results become mostly insignificant. This indicates that the results are very sensitive to the countries included in the analysis.

In a second step, countries that were colonized, part of the Soviet Union or divided at the time oil or/and gas was first produced, were excluded. It is assumed that oil or/and gas production under those conditions did not have an impact on local institutions as they were most likely influenced by other factors (e.g. the colonizers introduced institutions that are conducive to reaching their goals) and oil and/or gas production presumably only had a negligible effect on institutional change. Please refer to Table 1, Appendix B for results including those “problematic” countries.

(22)

18 Table 4.2 presents results for pooled OLS regressions with changes in executive constraint as dependent variable. Again it is confirmed that natural resource dependence (“NRD”) exerts a negative influence on changes in institutions. Although this result is only confirmed in the last regression (7), it has a negative sign in the preceding regressions. The initial institutional variable (“Initial Exconst”) is insignificant in all regressions; however, the interaction term between NRD and Initial Exconst is positive. This again indicates that having good initial institutions can prevent a country from experiencing a deterioration of its institutional environment and possibly also from the dreaded resource curse.

(23)

19 Table 4.1

Dependent variable: Change in Polity 2 Pooled OLS

(1) (2) (3) (4) (5) (6) (7)

Change in Polity2 Change in Polity2 Change in Polity2 Change in Polity2 Change in Polity2 Change in Polity2 Change in Polity2

NRD -0.0173* -0.0142 -0.0175* -0.0173* -0.0159 -0.0163 -0.0190* (0.00853) (0.00893) (0.00868) (0.00856) (0.00907) (0.00877) (0.00884) Initial Polity2 -0.0202** -0.0241** -0.0216* -0.0202** -0.0192* -0.0305 -0.0147 (0.00771) (0.00911) (0.00855) (0.00771) (0.00828) (0.0250) (0.00778) NRD x Initial Polity2 0.0362* (0.00771) 0.0313 (0.00911) 0.0364* (0.00855) 0.0362* (0.00771) 0.0348* (0.00828) 0.0427* (0.0250) 0.0230 (0.00778) SSA -0.0291 (0.0276) MENA -0.0250 (0.0269) Eastern European and transition countries 0.00210 (0.0113)

Latin America and the Caribbean 0.0139 (0.0273) OECD 0.0269 (0.0544) Asia 0.0434 (0.0342) Constant 0.00740 0.0196 0.00862 0.00737 0.00503 0.00155 0.00168 (0.0116) (0.0118) (0.0124) (0.0117) (0.0134) (0.0202) (0.0127) N 429 429 429 429 429 429 429 R2 0.008 0.012 0.008 0.008 0.008 0.009 0.012

Standard errors in parentheses ; *p<0.05 **p<0.01 ***p<0.001

(24)

20 Table 4.2

Dependent variable: Change in Executive constraint Pooled OLS (1) (2) (3) (4) (5) (6) (7) Change in Exconst Change in Exconst Change in Exconst Change in Exconst Change in Exconst Change in Exconst Change in Exconst NRD -0.0367 -0.0328 -0.0370 -0.0367 -0.0335 -0.0351 -0.0372* (0.0190) (0.0191) (0.0191) (0.0190) (0.0193) (0.0185) (0.0186) Initial Exconst -0.0133 -0.0153 -0.0143 -0.0134 -0.0115 -0.0201 -0.0140 (0.00917) (0.0103) (0.0101) (0.00920) (0.0101) (0.0230) (0.00949) NRD x Initial Exconst 0.0345* 0.0306 0.0348* 0.0346* 0.0319* 0.0360* 0.0364* (0.0158) (0.0164) (0.0160) (0.0158) (0.0162) (0.0171) (0.0170) SSA -0.0124 (0.0299) MENA -0.0194 (0.0344)

Eastern European and transition countries

0.00773 (0.0107)

Latin America and the Caribbean 0.0223

(0.0355) OECD 0.0180 (0.0489) Asia -0.00684 (0.0416) Constant 0.00402 0.00844 0.00476 0.00391 0.000322 -0.00130 0.00491 (0.0121) (0.0139) (0.0128) (0.0122) (0.0136) (0.0214) (0.0130) N 429 429 429 429 429 429 429 R2 0.005 0.006 0.006 0.006 0.007 0.006 0.006

Standard errors in parentheses; *p<0.05 **p<0.01 ***p<0.001

(25)

21 Table 4.3

Dependent variables: Change in Polity2 (1) and Polity2 (2) Fixed effects

(1) (2)

Change in Polity2 Polity2

NRD 0.0684 (0.0643) -0.109 (-0.106) NRD x Initial Polity2 -0.101 (0.103) -0.387* (0.181) Constant -0.00929 (0.0158) 0.148*** (0.0277) N 429 446 R2 0.003 0.026

Standard errors in parentheses *p<0.05 **p<0.01 ***p<0.001

Table 4.4

Dependent variables: Change in Executive Constraint (1) and Executive Constraint (2) Fixed effects

(1) (2)

Change in Exconst Exconst

NRD 0.0449 (0.0703) -0.609*** (0.129) NRD x Initial Exconst -0.0945 (0.0850) 0.335* (0.146) Constant -0.00751 (0.0173) 0.0983** (0.0321) N 429 446 R2 0.003 0.051

For both tables:

(26)

22 4.2 Hypothesis 1: Robustness checks

Robustness checks were conducted using two different dependent variables: Democracy and Autocracy. Please consult Appendix C for the results. Table 4 in the Appendix shows a fixed effects analysis for democracy. The results indicate that natural resource dependence is negatively associated with Polity2 (level). The interaction effect of natural resource dependence and initial Polity2 is negative. The result is consistent with table 4.3, but it contradicts the results in tables 4.1, 4.2 and 4.4. The results in the latter three tables are based on change variables, though. When it comes to the autocracy results (Table 5, Appendix C), The table in the Appendix presents results based on pooled OLS. The only noteworthy result is that natural resource dependence is significant in some regressions and has a positive sign, i.e. natural resource dependency makes countries more autocratic, ceteris paribus.

(27)

23 Table 4.5

Dependent variable: Change in Polity2 (1) – (4) and Polity2 (5) Pooled OLS (1) - (4) and Fixed effects analyses (5)

(1) (2) (3) (4) (5)

Change in Polity2 Change in Polity2 Change in Polity2 Change in Polity2 Polity2

Total fuel income -3.54e-13 9.94e-13 2.32e-11 -3.92e-15 -8.07e-12***

(6.47e-13) (1.60e-12) (3.91e-11) (1.64e-12) (7.20e-13)

Initial Polity2 -0.0123 -0.0170 -0.00556 -0.0176

(0.00656) (0.0111) (0.0148) (0.0233)

Total fuel income x Initial Polity2 3.62e-13 6.64e-12* -1.76e-11 9.47e-13 -3.86e-11***

(2.04e-13) (3.16e-12) (3.77e-11) (1.49e-12) (9.48e-13)

Constant 0.00235 (0.00664) 0.00608 (0.0100) -0.00731 (0.0214) 0.00542 (0.0164) 0.145 (0.228) N 1106 565 177 476 644 R2 0.003 0.003 0.005 0.003 -

Note: Regressions (1) - (4) are based on pooled OLS regressions and (5) is the result of a fixed effects regression

(28)

24 Table 4.6

Dependent variable: Change in Democracy (1) and (2); Democracy (3) and (4) Pooled OLS (1) - (3) and Fixed effects analyses (4)

(1) (2) (3) (4)

Change in Democ Change in Democ Democ Democ

Total fuel income -5.84e-13* -6.77e-13 -1.13e-14 5.26e-12

(-2.06) (-0.04) (-0.01) (0.89)

Initial Democ -0.00921* -0.0118 0.543***

(-2.02) (-1.89) (26.03)

Total fuel income x Initial Democ 2.37e-13 2.31e-13 6.55e-13 -4.10e-12

(1.20) (0.02) (0.69) (-1.10)

Constant 0.00633 0.00963 0.453*** 0.144

(1.04) (1.02) (15.54) (0.64)

N 1041 604 1081 1062

R2 0.005 0.005 0.25 -

Note: Regressions (1) - (3) are based on pooled OLS regressions and (4) is the result of a fixed effects regression

(29)

25 4.3 Hypothesis 1: Impulse response functions for Polity2 and Executive Constraint

In order to gain a deeper understanding of what happens to a country’s institutional environment as soon as it starts producing oil and/or gas, cumulative impulse response functions (IRFs) were estimated. Therefore, instead of clumping all countries together and estimating one IRF, several IRFs were estimated focusing on only one country at a time. For this analysis, solely the most extreme cases were included in this analysis, i.e. countries with the weakest initial Polity2 and Executive Constraint values as well as countries with the highest values in the sample. Furthermore, it is based on the NRD variable, which is not available for all countries as GDP data was not always available for the starting year of each country. Therefore, some countries that had high or weak initial Polity2 and Executive Constraint values could not be included. Therefore, although the United States had a higher value for initial executive constraint, it was not included but Belarus was included. The countries included should, therefore, not be regarded as having had the highest or weakest values in general but only in relative terms compared to other countries in the sample and depending on data availability. Please consult Table 4.7 for a list of countries.

Table 4.7

IRF countries (alphabetic order, not according to values) Polity2

Low initial values

Polity2

High initial values

Executive Constraint Low initial values

Executive Constraint High initial values

Congo, Dem. Rep.* Australia Benin* Australia

Equatorial Guinea* Denmark Chad* Belarus*

Sudan Greece* Philippines Denmark

Turkmenistan* Norway Thailand* Norway

Uzbekistan* Sweden* Uzbekistan* Trinidad and Tobago

*: Statistically significant IRFs

(30)

26 IRFs for Polity2

Congo, Dem. Rep.: Equatorial Guinea:

Turkmenistan: Uzbekistan:

(31)

27 IRFs for Executive Constraint

Benin: Chad: Thailand: Uzbekistan: Belarus: -.2 -.1 0 .1 0 5 10 15 20 varbasic, D1_NRD, levelchanges_exconst 95% CI cumulative irf step

(32)

28 In contrast, the graph for Equatorial Guinea fluctuates, but also remains below zero throughout the whole period. Conversely, the IRF for Benin shows a different pattern: as soon as the country started producing oil in 1983, the institutional environment seemingly improved substantially, and then fluctuated until it became relatively stable after about 10 years (above zero).

Another set of countries that can be grouped together are those that were formerly part of the Soviet Union, i.e. Belarus, Turkmenistan and Uzbekistan. The graph for Belarus first shows a stark decline below zero and then fluctuates. Eventually, its executive constraint stabilizes below zero after about 8 years. In contrast, Turkmenistan experienced a slight deterioration initially and then its institutional environment improved and it stabilized after about 5 years. When it comes to Uzbekistan, its Polity2 also improved initially and then fluctuated with a slight downward trend. Eventually, the line stabilizes after 12 years between 0 and 0.5. Finally, Greece and Sweden can be grouped together as they are both OECD countries and were considered to have relatively high initial Polity2 values. Here, the graph for both countries indicates that after the start of oil and/or gas production, the Polity2 values improved. Particularly the graph for Greece indicates a considerably improvement in Polity2 as a result of oil and gas production.

Overall, the IRFs indicate that some countries suffered from a deterioration of their institutional environment (i.e. Belarus, Chad, Congo, Dem. Rep., Equatorial Guinea, Thailand, Uzbekistan) and others experienced an improvement (Benin, Greece, Sweden, Turkmenistan) as a result of starting fuel production. However, one cannot neglect the fact that out of 20 countries for which IRFs were computed for 9 countries the IRFs were insignificant.

(33)

29 4.4 Hypothesis 2: Main results

The main results for hypothesis 2 are presented in tables 4.8, 4.9, 4.10 and 4.11. The first table (4.8), reports results for pooled OLS regressions. The first regression includes all countries in the sample and the subsequent regressions either include only a specific region (i.e. Sub-Saharan Africa; Middle East North Africa; Latin America and the Caribbean, South and East Asia) or a group of countries (i.e. OPEC, OECD, former Soviet countries). The reason for splitting the sample into different subsets is that from the previous analysis it became clear that the consequence of fuel production seems can be very different across countries and grouping similar countries (e.g. similar in terms of income (OECD) or location) may provide more useful results than analyzing them all at once.

The first regression in Table 4.8 shows that total natural resource dependence is associated with less dependence on the manufacturing sector. As manufacturing (% of GDP) can be seen as a proxy for industrialization, the result may imply that a higher dependence on natural resources hinders a country from industrializing. As already explained in the literature review, industrialization is considered an important part of economic development by many economists (e.g. Murphy, Shleifer and Vishny, 1989). Furthermore, the variable Polity2 is mostly positive and significant, indicating that if a country has good institutions it is likely to be more dependent on the manufacturing sector, i.e. the country is likely to be more industrialized. In regression (4) it is negative and significant, indicating that having democratic values leads to less dependence on the manufacturing sector. As regression (4) only includes the Middle East and North Africa, this result seems to be very specific to this region as for all other regions or groups the variable is positive. Finally, the interaction effect of total natural resource rents and Polity2 is significant in most regressions. Unexpectedly, the sign is often negative indicating that having good institutions and being dependent on natural resources results in lower levels of industrialization. The theory suggested in the literature review predicts that it should be the other way around: having good institutions and being dependent on natural resources should exhibit a positive effect on a country’s ability to industrialize as effective policies are more likely to be implemented.

(34)

30 Table 4.8

Dependent variable: Manufacturing (value added, % of GDP) Institutions variable: Polity2

Pooled OLS (1) (2) (3) (4) (5) (6) (7) (8) (9) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) TNRR1 -0.231*** -0.215*** -0.158*** -0.173*** -0.291*** -0.0210 -0.140*** -0.0803* -0.172*** (0.00802) (0.00870) (0.0132) (0.0299) (0.0224) (0.0492) (0.0312) (0.0333) (0.0373) Polity2 1.588*** 2.024*** 2.016*** -2.660*** 4.284*** 0.723 1.872*** -0.436 0.628 (0.157) (0.133) (0.258) (0.693) (0.404) (0.400) (0.564) (0.731) (0.437) TNRR1 x Polity2 -0.0604*** -0.0525*** -0.0603*** 0.0371 -0.116*** -0.463*** -0.0371 0.0986*** 0.0502 (0.00669) (0.00685) (0.0132) (0.0223) (0.0162) (0.0511) (0.0443) (0.0237) (0.0329) Constant 16.54*** 16.20*** 12.55*** 13.67*** 19.21*** 19.41*** 17.70*** 19.38*** 18.04*** (0.142) (0.132) (0.214) (0.780) (0.487) (0.446) (0.487) (0.611) (0.413) N 4534 4387 1474 391 272 809 639 220 537 R2 0.201 0.224 0.167 0.495 0.599 0.160 0.073 0.240 0.104

Notes: (1): all countries; (2): without outliers (i.e. Saudi Arabia; Swaziland; Trinidad and Tobago; Turkmenistan; Uzbekistan); (3) only Sub-Saharan Africa; (4) only Middle East and North Africa; (5) only OPEC; (6) only OECD; (7) only Latin America and the Caribbean; (8) only former Soviet countries ; (9) only East and South Asia

(35)

31 Table 4.9

Dependent variable: Change in manufacturing (value added, % GDP) Institutions variable: Change in Polity2

Fixed effects (1) (2) (3) (4) (5) (6) (7) (8) (9) Change in man. (% GDP) Change in man. (% GDP) Change in man. (% GDP) Change in man. (% GDP) Change in man. (% GDP) Change in man. (% GDP) Change in man. (% GDP) Change in man. (% GDP) Change in man. (% GDP) TNRR1 -0.00221 -0.00103 -0.00934 -0.0127 -0.0135 -0.0105 -0.00707 0.0256*** 0.0181 (0.00419) (0.00424) (0.00499) (0.00693) (0.00610) (0.0139) (0.00744) (0.00377) (0.0116) Change in Polity2 -0.261 -0.239 -0.172 -0.744 -1.121 0.0460 -0.164 0.0927 0.121 (0.133) (0.128) (0.110) (0.481) (0.529) (0.302) (0.378) (1.332) (0.295) TNRR1 x Change in Polity2 0.0121 0.0138* 0.0149** 0.0294 0.0370** -0.0629** -0.0290 0.123 -0.0557 (0.00665) (0.00699) (0.00467) (0.0461) (0.0113) (0.0216) (0.0508) (0.149) (0.0521) Constant -0.0821 -0.103* 0.0754 0.344* 0.421 -0.162*** -0.0810 -1.050*** -0.0767 (0.0437) (0.0437) (0.0657) (0.149) (0.194) (0.0359) (0.0603) (0.0869) (0.0998) N 4427 4349 1445 389 272 787 654 241 529 R2 0.001 0.001 0.004 0.010 0.028 0.004 0.007 0.012 0.014

Notes: (1): all countries; (2): without outliers (i.e. Iran and Swaziland); (3) only Sub-Saharan Africa; (4) only Middle East and North Africa; (5) only OPEC; (6) only OECD; (7) only Latin America and the Caribbean; (8) only former Soviet countries ; (9) only East and South Asia

(36)

32 explanation for that may be that those countries already industrialized as part of the Soviet Union (e.g. Stalin’s Five Year Plans) and therefore, they already had at least basic industry. Those countries may use additional resource rents to further develop the already existing industry. The interaction effect of total natural resource rents and change in Polity2 is significant and positive in the regressions excluding outliers (2), only including Sub-Saharan Africa (3) and only including OPEC countries (3). The interaction effect is significant and negative for OECD countries. One reason for this result may be that OECD countries already industrialized and their economies today are usually highly service oriented. Therefore, if an OECD country receives high natural resource rents, while having good institutions, it is not likely to industrialize as a result. In contrast, developing countries with an underdeveloped industrial sector are likely to push the industrialization process forward in the presence of natural resource rents and a good institutional environment.

More regression analyses were conducted with Executive Constraint as institutional variable. The tables can be found in Appendix (Table 6 and 7). The signs are largely as expected with the exception of a few. One rather surprising finding is that the interaction term (“TNRR x Exconst”) is negative for most regressions except for the one including only former Soviet countries (Pooled OLS) or Latin America and the Caribbean (Fixed effects). For Sub-Saharan Africa, the interaction may be negative due to the nature of the sample. Among those countries there are not sufficient countries included that actually have high executive constraint. The sample consists of countries with similar levels of executive constraint. Therefore, the regression analysis did not produce reliable coefficients.

4.5 Hypothesis 2: Robustness checks

(37)

33 Table 4.10

Dependent variable: Manufacturing (value added, % of GDP) Institutions variable: Democracy

Pooled OLS (1) (2) (3) (4) (5) (6) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) TNRR1 -0.186*** -0.0361 -0.0486 -0.0861 -1.339*** 1.555 (0.0442) (0.390) (0.0444) (0.0813) (0.264) (1.985) Democ -6.662*** -20.11*** 0.356 -0.0724 13.22*** 0.695 (1.019) (3.870) (1.045) (2.018) (1.945) (13.46) TNRR1 x Democ 0.0192 0.702 -0.000261 -0.0553 -1.241*** -1.483 (0.0404) (0.368) (0.0330) (0.0798) (0.222) (1.758) Constant 24.42*** 15.81** 17.76*** 17.24*** 32.71*** 11.04 (1.045) (5.056) (1.462) (2.322) (2.288) (14.98) N 164 40 39 68 32 24 R2 0.576 0.872 0.062 0.158 0.636 0.729

Standard errors in parenthesis *p<0.05 **p<0.01 ***p<0.001

Notes: (1) without outliers (i.e. Mongolia; Swaziland; Trinidad and Tobago; Turkmenistan; Ukraine; Venezuela are excluded); (2) only Sub-Saharan Africa; (3) only OPEC; (4) only Latin America and the Caribbean; (5) only former Soviet countries ; (6) only East and South Asia

(38)

34 Table 4.11

Dependent variable: Manufacturing (value added, % of GDP) Institutions variable: Autocracy

Fixed effects (1) (2) (3) (4) (5) (6) (7) (8) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) Man. (% GDP) TNRR1 -0.123*** -0.0934*** -0.229*** -0.106*** 0.0330 -0.131*** -0.0876 -0.170*** (0.00894) (0.0116) (0.0239) (0.0141) (0.0503) (0.0376) (0.104) (0.0326) Autoc -0.212 0.239 -1.479** -0.149 -5.070*** 1.862*** -0.0428 0.0404 (0.123) (0.186) (0.486) (0.382) (0.341) (0.457) (1.212) (0.326) TNRR1 x Autoc 0.0256*** 0.00414 0.102*** 0.0309** 0.594*** -0.0865* -0.0591 0.0128 (0.00637) (0.00986) (0.0144) (0.00971) (0.0540) (0.0387) (0.0524) (0.0232) Constant 15.50*** 11.32*** 15.12*** 13.25*** 16.35*** 18.66*** 18.79*** 17.83*** (0.100) (0.173) (0.681) (0.444) (0.279) (0.379) (0.987) (0.300) N 4469 1475 439 344 809 697 201 537 R2 0.042 0.052 0.217 0.180 0.274 0.035 0.056 0.060

Standard errors in parenthesis *p<0.05 **p<0.01 ***p<0.001

Notes: (1): without outliers (i.e. Lithuania; Qatar; Saudi Arabia; Swaziland; Turkmenistan; Ukraine; Uzbekistan are excluded); (2) only Sub-Saharan Africa; (3) only Middle East and North Africa; (4) only OPEC; (5) only OECD; (6) only Latin America and the Caribbean; (7) only former Soviet countries ; (8) only East and South Asia

1

(39)

35 values, the most democratic country may still be not democratic and therefore, also not act as such.

Table 4.11 presents the robustness check conducted with fixed effects and autocracy as institutions variable. It partly confirms what the other regressions showed. However, again there are also unexpected signs such as the positive sign for the interaction effect (for regressions (1), (3), (4) and (5)). However, this may again reflect the composition of the sample. The exception is regression (6) which has a negative sign for the interaction effect, indicating that being autocratic and dependent on natural resource rents is negatively related to industrialization.

4.6 Discussion and limitations

This section presented the results for hypothesis 1 and 2 based on pooled OLS and fixed effects regressions as well as impulse response functions. The main institutions variables employed were either Polity2 or Executive Constraint obtained from the Polity IV database. In addition, democracy and autocracy variables were employed to conduct robustness checks. Overall, there is evidence for the two hypotheses. For the first hypothesis, the pooled OLS regressions are largely in line with the proposed hypotheses. In addition, the impulse response functions provide evidence for the effect of natural resource production on institutional development. However, it is not clear why for some countries the institutional environment deteriorates after production starts (e.g. Congo, Dem. Rep.), for others it leads to an improvement of the institutional environment (e.g. Greece and Sweden) and again for others, it does not have any effect (9 out of the 20 countries that were selected). It might be the case that only the most extreme initial institutions have an impact. For example, the Democratic Republic of Congo has one of the worst institutional environments worldwide and its IRF shows a deterioration of its institutions. Sweden, on the other hand, has one of the best institutional environments and the IRF shows an improvement after fuel production started. However, it may also be the case that there are different country-specific factors that are difficult to capture with the analyses conducted.

(40)

36 institutions in the world deteriorated overall. Then, it seems as if the country’s institutions improved, although it did not change. This limitation also applies to the World Governance Indicators by the World Bank as they were computed the same way. The World Bank, however, states that there has been no trend in the world. Therefore, if the overall trend remained relatively stable, the effect on the results should be minor (Kaufmann, Kraay and Mastruzzi, 2010). When it comes to the IRFs, they were computed with Vector Autoregressions (VARs). Those VARs do not account for heterogeneity or nonlinear relationships. Furthermore, the number of lags, which has a strong influence on the outcomes, may be chosen incorrectly. Finally, another limitation of this analysis is that lags could have also been included in the pooled OLS and fixed effects regressions. The reason is that the start of fuel production may not immediately lead to changes in institutions, which may also be the reason why the change variables are often insignificant.

5 Conclusion and recommendations

(41)

37 Formulating a universally applicable theory based on these results is not feasible. The reason as explained is that countries are heterogeneous and they seem to react differently to oil and/or gas production. As soon as a country produces oil and/or gas different initial conditions may interact rather than only institutions. Dunning (2008) put this “conditional theory” forward providing the example of Venezuela. He argues that the oil producing country actually became more democratic as a result of receiving high oil rents, while its neighboring countries became more autocratic. Dunning (2008) suggests that the wealth distribution at the time resource rents are generated is a crucial factor. He explains that in a more unequal society, elites do not fear redistribution of those rents and a country can become more democratic. Conversely, if a country is more homogenous in terms of wealth distribution, the pressure to redistribute wealth is lower and elites attempt to fortify their status using those rents. This argument can be easily expanded: How about ethnic divisions in a country? It has been often reported that different ethnic groups fight for rents. Therefore, the composition of ethnic groups (i.e. number of groups, but also size) may also play a role, particularly in African countries. Another possible scenario would be when there are more groups of elites rather than just one big group of politicians. If there are several groups of business elites and elites in politics, it may again lead to a different outcome as rents are not only allocated to a few individuals.

Regarding the second hypothesis, the results again provide some support. However, some results also contradict the hypothesis. For example, in some regressions the interaction term consisting of total natural resource rents and institutions is negative, indicating that countries that have good institutions and high rents are less dependent on the manufacturing sector and are less industrialized. This result, for example, was obtained for Sub-Saharan Africa. However, this may also be due to the nature of the data: if all African countries have weak institutions, it may not be statistically feasible to get any other result as the sample is too homogeneous in that respect.

(42)

38 the price for primary goods on world markets fluctuates and the resources can be exhausted at one point (e.g. in Nauru).

(43)

39

6 References

Acemoglu, Daron, Simon Johnson and James A. Robinson. 2001. An African success story: Botswana. MIT Department of Economics Working Paper No. 01-37.

Acemoglu, Daron, Simon Johnson and James A. Robinson, 2001. The colonial origins of comparative development: An empirical investigation. The American Economic Review, 91 (5): 1369-1401.

Acemoglu, Daron and James A. Robinson, 2010. Why is Africa poor? Economic History of Developing Regions, 25 (1), pp. 21-50.

AfdB, OECD, UNDP and UNECA. 2013. African Economic Outlook: Structural transformation and natural resources. Paris: OECD Publishing.

Angwafo, Manka and Punam Chuhan-Pole. 2011. Yes Africa can: Success stories from a dynamic continent. Washington, DC: World Bank.

Arezki, Rabah and Mustapha K. Nabli. 2012. Natural resources, volatility, and inclusive growth: Perspectives from the Middle East and North Africa. IMF Working Paper. WP/12/111.

Asian Development Bank. 2014. ADB country operations business plan: Nauru 2015-2017. Available at: < http://www.adb.org/sites/default/files/linked-documents/cobp-nau-2015-2017-sd-03.pdf> [accessed May 26, 2015].

Auty, Richard M. 2001. Resource abundance and economic development. Oxford: Oxford University Press.

BBC. 2015. Nauru profile- timeline. Available at: <http://www.bbc.com/news/world-asia-pacific-15433901> [accessed May 26, 2015].

Campbell, Colin J. 2013. Campbell’s Atlas of oil and gas depletion. Second edition. New York: Springer.

De Rossa, Donato and Mariana Iootty. 2012. Are natural resources cursed? An investigation of the dynamic effects of resource dependence on institutional quality. World Bank Policy Research Working Paper 6151.

Dunning, Thad. 2008. Crude democracy: Natural resource wealth and political regimes. New York: Cambridge University Press.

(44)

40 Haber, Stephen and Victor Menaldo. 2011. Haber and Menaldo 2011 APSR Dataset. Available at: < http://stephen-haber.com/data/ > [accessed 03 April 2015].

Huntington, Samuel. 1991. The third wave: Democratization in the late twentieth century. Norman: University of Oklahoma Press.

IMF. 2011. Resource-rich countries can learn from history, says Collier. Available at: http://www.imf.org/external/pubs/ft/survey/so/2011/INT092711B.htm [accessed February 22, 2015].

Kaufmann, Daniel, Aart Kraay and Massimo Mastruzzi. 2010. The Worldwide Governance Indicators: Methodology and analytical issues. Draft research working paper.

Lal, Deepak and H. Myint. 1996. The political economy of poverty, equity and growth. Oxford: Clarendon Press.

Lindauer, David L. and Michael Roemer. 1994. Asia and Africa: Legacies and Opportunities in Development. San Francisco: ICS Press.

Mahdavy, Hossein. 1970. The pattern and problems of economic development in rentier states: The case of Iran. Published in : Studies in the Economic History of the Middle East. Oxford: Oxford University Press.

Marshall, Monty G., Ted Robert Gurr and Keith Jaggers. 2014. Polity IV project: Political Regime Characteristics and Transitions, 1800-2013: Dataset Users’ Manual. Polity IV, Center for Systemic Peace.

Murphy, Kevin M., Andrei Shleifer and Robert W. Vishny. 1989. Industrialization and the big push. Journal of Political Economy, 97 (5): 1003-1026.

Ross, Michael L. 2001. Does oil hinder democracy? World Politics, 53 (3): 325-361.

Ross, Michael L. 2013. Oil and Gas Data, 1932-2011. Available at: <http://hdl.handle.net/1902.1/20369> [accessed 03 April 2015].

Ross, Michael L. 2014. What have we learned about the resource curse? Working paper. UCLA, Department of Political Science.

Sachs, Jeffrey D. and Andrew M. Warner. 1995. Natural resource abundance and economic growth. NBER Working Paper No. 5398.