Resources Policy 66 (2020) 101656 Available online 19 March 2020 0301-4207/© 2020 Elsevier Ltd. All rights reserved. Domestic revenue displacement in resource-rich countries: What’s oil money got to do with it? Daniel Ofoe Chachu Department of Economics, University of Ghana, Department of Economics, University of Ghana, Box LG 57, Legon, Accra, Ghana A R T I C L E I N F O Keywords: Hydrocarbon revenues Non-hydrocarbon revenues Institutions A B S T R A C T Cross-country studies on the effect of hydrocarbon revenues and non-hydrocarbon tax effort are only now emerging. Using an expanded global dataset in a two-stage least squares framework, we confirm a displacement effect. A percentage point increase in hydrocarbon revenues displaces non-hydrocarbon revenues by 0.2 to 0.3 percentage points. With low levels of domestic revenue and a debt crises looming for many developing countries, resource-rich countries need to leverage on their resource wealth to invigorate the non-resource sectors of their economies. This should widen the tax base and optimize the tax take for oil-rich countries over the long haul. 1. Introduction 1.1. Overview of the “resource curse” The concept of a “resource curse” has been discussed variously and could be broadly defined within the context of evaluating the adverse effect of natural resource wealth on development outcomes (Sachs and Warner, 1995; Collier, 2007, 2010; Collier and Goderis, 2007; Collier and Hoeffler, 2009; Badeeb et al., 2017; Yanikkaya and Turan, 2018; Mawejje, 2019; Bildirici and Gokmenoglu, 2020; Matallah, 2020). While the neo-classical or conventional view depict natural resources as input for production and therefore a potential contributor to growth and welfare improvements (Davis and Tilton, 2005), alternative views exist. For instance, the Prebisch-Singer hypothesis argues that the terms of trade for primary commodities decline over time due to their low-income elasticity of demand relative to the demand for manufac- tured products. More so, a natural resource-rich economy may experi- ence a “resource pull” effect, which manifests through a ‘pooling away’ of physical and human resources from the non-resource sector of the economy thus depriving it the ability to expand (Klein, 2010). This invariably undermines the tax base. A shrinking of the tax base has implications for revenue mobilization in the non-resource sector when the natural resource sector diminishes or depletes over time. 1.2. Resource rents and tax revenue mobilization A growing body of literature explore the relationship between natural resource rents and fiscal capacity (see for example Knack, 2009; Jensen, 2011; Ossowski and Gonzales, 2012; Thomas and Trevi~no, 2013; Crivelli and Gupta, 2014; Mohtadi et al., 2016; Bhattacharyya et al., 2017) .1 Knack (2009) finds that windfalls from natural resource extraction are associated with inefficient tax policy and administration. He indicates that previous studies had neither directly looked at the impact of resource windfalls on the quality of tax systems nor on taxes in manufacturing and services sectors, which mainly fall within the non-resource sector. Bornhorst, et al, (2009) break the trend using novel data which they compile on 30-oil producing countries for the period 1992 to 2005. Using panel econometric techniques, they find that among resource-rich nations, a 10 percentage point increase in hydro- carbon revenues diminishes other domestic revenue sources by as much as 2 percentage points. For example, prior to the global financial crisis (i.e. between 1980 and 2005), natural resource revenues saw an average increase of 7.7 percentage points of GDP for resource rich countries in Sub-Saharan Africa (Keen and Mansour, 2009). On the other hand, non-resource related taxation rose by less than 1 percentage point of GDP for the same period (Keen and Mansour, 2009). The trend in tax revenue mobilization has been attributed to the increasing importance of natural resource taxation. However, compared to oil producing countries, non-oil producing countries have performed better in raising non-oil related tax revenue as a share of their Gross Domestic Product (Afri- can Development Bank, 2010). The question that arises is whether the poor performance in non-oil revenue mobilization in oil-rich countries is directly as a result of a crowding out effect of hydrocarbon revenues. E-mail addresses: dochachu@st.ug.edu.gh, doc.chachu@gmail.com. 1 See Mawejje (2019), Gnangnon and Brun (2019), Oyarzo and Paredes (2019) and Abdulahi et al. (2019) for a survey of more recent related discussions. Contents lists available at ScienceDirect Resources Policy journal homepage: http://www.elsevier.com/locate/resourpol https://doi.org/10.1016/j.resourpol.2020.101656 Received 22 November 2019; Received in revised form 13 February 2020; Accepted 9 March 2020 mailto:dochachu@st.ug.edu.gh mailto:doc.chachu@gmail.com www.sciencedirect.com/science/journal/03014207 https://http://www.elsevier.com/locate/resourpol https://doi.org/10.1016/j.resourpol.2020.101656 https://doi.org/10.1016/j.resourpol.2020.101656 https://doi.org/10.1016/j.resourpol.2020.101656 http://crossmark.crossref.org/dialog/?doi=10.1016/j.resourpol.2020.101656&domain=pdf Resources Policy 66 (2020) 101656 2 In this paper, we evaluate the causal effect of hydrocarbon revenues on non-hydrocarbon tax effort using a global panel dataset. We build on Bornhorst et al. (2009), which is most related to our work. Our starting point is to examine the reproducibility of their main results. We do this by obtaining their data on hydrocarbon and non-hydrocarbon revenues for a global sample of 30 hydrocarbon-producing countries over the 1992–2005. As our contribution to this literature, we expand this dataset for 77 countries over the period 1992 to 2008 by relying on the IMF’s Article IV Country Review Reports for these countries. We adopt a more robust empirical strategy to address concerns about endogeneity, not adequately addressed in previous work. In evaluating the causal impact of hydrocarbon revenues on non-hydrocarbon tax effort, we construct an instrument which exploits exogenous variation in a weighted real global price of crude oil over the period. Our identification stems from the fact that the only way the weighted real global price of crude oil impacts on non-resource revenues is through its effect on resource revenues, hold- ing other factors constant. Furthermore, we explore the extent to which the quality of in- stitutions mediates the relationship between hydrocarbon revenues and non-hydrocarbon revenue mobilization. We find that a 10 percentage point increase in hydrocarbon revenues as a share of GDP displaces the share of non-hydrocarbon revenues by 2 to 3 percentage points of GDP. The magnitude of the displacement effect is higher in some of our specifications, compared to Bornhorst et al. (2009). In effect, in the presence of oil and gas resources, revenue mobilization outside the non-resource sector diminishes. Maximizing revenue flows from newly discovered oil is important. While good quality institutions are desir- able, we find their moderating influence on the adverse impact of resource revenues on non-resource revenues to be weak. The tendency of reneging on effort to improve fiscal capacity outside the natural resource sector in the face of a boom in the latter is a concern. Sustaining a strong non-oil related revenue base is important for reducing fiscal risks but also building and maintaining the capacity of revenue man- agement institutions. A well-diversified domestic revenue base would ensure better smoothening of tax rates over time instead of panic in- creases in tax rates in response to the diminishing or depletion of non-renewable natural resources. The rest of the paper is as follows. Section two describes our empirical strategy. The results are discussed in section three. The fourth section delivers a conclusion to the study. 2. Empirical framework and description of data As a first step, we reproduce the main specifications of Bornhorst et al. (2009).2 Next, we expand their dataset and propose an alternative identification strategy to re-examine the evidence provided by Born- horst et al. (2009). We identify and address endogeneity concerns in their paper. Furthermore, we introduce a set of omitted governance indicators such as government stability, bureaucratic quality and investor (risk) profile which could influence revenue mobilization efforts. We follow Bornhorst et al. (2009), in defining hydrocarbon revenues as a percentage of GDP as our measure of natural resource dependence. This motivation is informed by our interest in re-examining the hy- pothesis that dependence on oil-related revenues undermines the po- tential for generating and sustaining non-oil revenues. 2.1. Step one: replicating Bornhorst et al. (2009) The base model specification is given by the following: Yit ¼ β1 X’ 1it þ β2X2it þ ϕt þ πi þ εit (1) Where Yit is the main outcome (dependent) variable of interest, X1it is a vector of control variables, X2it is the main explanatory variable, ϕt is a time dummy and πi is a country-specific fixed effect. The dependent variable is non-hydrocarbon revenues as a percentage of Gross Domestic Product (GDP) while the main explanatory variable is hydrocarbon revenues as a percentage of GDP. The list of control variables are foreign grants as a percentage of GDP, non-hydrocarbon trade openness as a percentage of GDP, agriculture (value-added) as a percentage of GDP and corruption perception index. The basis for their choice of covariates is informed by the following. Khattry and Rao (2002) find that trade liberalization depresses tax revenue effort in low income countries and upper middle income countries as a result of falling incomes and trade taxes. Gupta (2007) notes an ambiguous relationship between trade liberalization and tax revenues although he finds a decline in revenues when trade liber- alization takes the form of reduction in tariffs. Keen and Mansour (2009) also observe that between 1980 and 2005, trade tax revenues as a share of GDP and total tax revenue fell in Sub Saharan Africa thus presenting a major challenge for non-resource rich countries especially. Rodrik (1998) associates trade openness with larger role for government (ful- filled partly through increases in taxes) in providing social protection against the downside risks of running an open economy. Thus, Bornhorst et al. (2009) control for the level of openness with a variable they call non-hydrocarbon openness as a percentage of GDP. Bornhorst et al. (2009) use GDP per capita as a proxy for the level of development in a country. The state of development of a country is likely to be correlated with the level of revenue that can be raised from both hydrocarbon investments and non-hydrocarbon investments. The structure of the economy is also relevant in determining how much revenue can be raised. Gupta (2007) finds that besides GDP per capita and trade openness, the share of agriculture in GDP is a strong deter- minant of revenue. Similarly, he finds that foreign aid has a significant impact on revenue performance. Institutional factors such as the level of corruption also affect how much revenue can be generated. Holding the level of revenue potential constant, a corrupt country is less likely to reap much benefit from its revenue potential than a less corrupt country. In their first specification, Bornhorst et al. (2009) run an Ordinary Least Squares regression of non-hydrocarbon revenues as a percentage of GDP on hydrocarbon revenues as a percentage of GDP, controlling for grants as a percentage of GDP with country fixed effect and time effects in their full sample of 30 countries. In their second specification they include all other control variables. A variant of the second specification is shown in the third column of their table (see appendix), which has outliers (Russia, Kuwait and Norway) dropped out of the sample of countries used in that regression. All their panel OLS regressions have country and time fixed effects. The country fixed effect controls for time-invariant country specific factors that might be correlated with their explanatory and dependent variables while the time effects con- trols for factors changing in the same way over time that affect all countries and are correlated with the explanatory and dependent vari- ables. The authors express a concern that non-hydrocarbon revenues are persistent over time and that a shock to the revenues in one period may persist and therefore affect non-hydrocarbon revenues in the subsequent year. If this concern is not accounted for in the regression, then the effect of hydrocarbon revenues on non-hydrocarbon revenues will be biased downwards. To address this, they introduce a one-year lag of non-hydrocarbon revenues as a control variable in the fourth column of their table. Doing this actually brings up the problem of auto correlation – the error term of the lagged dependent variable would be correlated with the error term of the regression. In response to this problem, the authors make use of a generalized method of moment (GMM) estimator popularly known as Arellano and Bond (1991) estimator. The impor- tance of the Arellano and Bond estimator stems from its ability in 2 We acknowledge the role of the authors in sharing their data on hydro- carbon and non-hydrocarbon revenues for the 30 countries they use in their sample. D.O. Chachu Resources Policy 66 (2020) 101656 3 helping to deal with the following. (i) Panel data with smaller time periods (T) and larger entities (N) (ii) Explanatory variables that are correlated with the error term (iii) Fixed effects that are contained in the error term (iv) Autocorrelation that arises from including a lagged dependent variable Bornhorst et al. (2009) however provide limited information on how they specify the Arellano and Bond estimator. As a result, we attempted several specifications towards getting as close as possible to the esti- mates in column four of Table 3 of their paper (see appendix). For instance, they indicate that they instrument for “potentially endogenous variables, such as corruption”. In the specification we present in this paper, we use the first difference option, instrumenting both for cor- ruption and the lagged dependent variable. The instrument for corrup- tion is all available lag levels of the variable from t-1 and earlier while the instrument for the lag of non-hydrocarbon revenues is all available lags for the variable from t-2 and earlier. The summary statistics we present in Table 1 is based on the base model specification in Table 3 of our results. The average level of non-hydrocarbon revenues as a percentage of GDP is 13.5 with a standard deviation of 8.7 percentage points. Mean- while, the average level of hydrocarbon-revenues as a percentage of GDP is 16.8 percent of GDP with a wider standard deviation of about 12 percentage points. 2.2. Step two: Re-examining the evidence from a larger sample In the second step, we re-examine the evidence by extending the period of our sample to 2008, prior to the full pass-through effect of the global financial crises. The sample covers 77 countries. The extended sample consists of the 30 oil producing countries used in Bornhorst et al. plus 47 other countries that neighbor the former and produced insig- nificant (15 countries) or no oil at all (32 countries) in the year 2008. The essence is to allow for a difference-in-difference analysis, account- ing for the fact that not all countries were raking in oil and gas revenues over the period under consideration.3 Data for hydrocarbon and non- hydrocarbon revenue as a percentage of GDP for the extended period was collated from the International Monetary Fund (IMF) Article IV country reports. This was complemented by domestic revenue data for non-oil producing countries from the World Bank’s World Development Indicator database and IMF’s Global Development Finance database. As Gupta (2007) notes, countries that struggle to raise revenue from traditional sources (direct and indirect taxes) have the tendency to look to other sources such as aid and debt. Similarly, it is plausible to suggest that hydrocarbon revenues are endogenous in the model used by Bornhorst et al. (2009). Countries that have a limited tax base such as those with large informal economies or those that have an elastic tax base may face difficulties in raising non-hydrocarbon tax revenues. Furthermore, in a highly competitive world where countries try to attract foreign capital, it becomes more difficult to mobilize revenue by increasing taxes on firm-specific economic rents as compared to location-specific rents (Zodrow, 2006). One basic difference between firm-specific rent and location-specific rent is that the latter accrues to firms whose economic activity are geographically defined as in the case of mining whereas economic activities relating to the former are more geographically dispersed. Persistent difficulties and limited opportu- nities for mobilizing non-hydrocarbon revenues might compel countries to devote more effort into raising hydrocarbon revenues so long as that source is available. The scenario above raises the issue of reverse cau- sality and therefore the need to instrument for hydrocarbon-revenues as a percentage of GDP. The advantage of using commodity prices to analyze the effect of natural resources is in the fact that they are “typically unaffected by the behavior of individual countries” (Collier and Goderis, 2009).4 It is however important to recognize for instance that there have been situ- ations where the supply of crude oil by a leading producer fell signifi- cantly (for example due to the unrests in Libya), which necessitated efforts by members of the Organization for Petroleum Exporting Coun- tries (OPEC) to fill the gap in order to stabilize prices. This notwith- standing, it is reasonable to assume that the real price of crude oil is beyond the control of a single country. The expanded data allows for a treatment group (countries endowed with hydrocarbons) and control group (countries with no hydrocarbons). The proposed instrument for hydrocarbon revenues is defined as follows: Instrumentit ¼Pt � � Oili2008 GDPi2008 � � 100% (2) Where Pt is the average world price of crude oil in year t (measured in 2000 constant prices to adjust for inflation); Oili2008 is total crude oil production level in barrels in year 2008 for each country and ​ GDPi2008 is the 2008 GDP for each country measured in 2000 constant prices. The oil production and GDP levels are held constant over time in con- structing the instrument. In this way all the independent variation in the instrument once we control for fixed effects, would be coming from factors outside the control of a country (world price of crude oil adjusted for inflation over the period under review). Thus, our exclusion re- striction is that changes in the weighted real global crude oil prices only affects non-hydrocarbon revenues through its impact on hydrocarbon revenues. Oil production is not allowed to change over time since its change over time may be driven by omitted variables that are also correlated with non-hydrocarbon revenues. Interacting world oil prices with oil production splits the data between treatment and control groups. In effect, this is a difference-in-differences approach. Oil prices have a significant effect on hydrocarbon revenues for countries that have much oil and little or no effect on countries that have little or no oil. A potential confounder to our exclusion restriction is that hydro- carbon revenues may affect non-hydrocarbon revenues through its impact on corruption, conflict or any of the possible channels of the natural resource curse. We respond to this concern by controlling for these channels using variables commonly used in the literature or their proxies. We do this both within a fixed effect econometric (FE) frame- work as well as employing a Generalized Method of Moments estimator (GMM). For instance, we control for corruption in our preferred speci- fications. We also tackle macro or political institutional concerns by accounting for proxies of conflict, war and general macroeconomic risks Table 1 Summary statistics for estimating Bornhorst et al. (2009). Variable Obs. Mean Std. Dev. Min Max Non-hydrocarbon revenues (net of grants) as a % of GDP 298 13.5 8.7 -4.5 46.0 Hydrocarbon revenue as a % of GDP 298 16.8 11.9 0.9 67.8 Grants (net of debt relief) as a % of GDP 298 0.9 1.5 0 11.2 Corruption Perception Index 298 4.4 1.0 1.0 6.0 Log GDP per capita (2000 constant $) 298 8.0 1.4 5.6 10.6 Non-hydrocarbon openness as % of GDP 298 63.5 35.1 15.0 181.1 Agriculture as a % of GDP 298 9.6 8.5 0.4 48.6 3 For instance, although Ghana is currently an oil producing and exporting nation, with significant revenues from the sector, it was not the case during the period under review. 4 They actually cite the work of Deaton and Miller (1995).. D.O. Chachu Resources Policy 66 (2020) 101656 4 using a set of institutional variables. A key caution, however, is that our estimates would be misleading if there are other plausible channels that have not been accounted for. With the instrument, we run a Two-Stage Least Squares (2SLS) regression model with fixed effects. We also apply the Arellano and Bond estimator to the expanded sample, with the inclusion of the new in- strument. In addition, we specify interaction terms in order to look at the differential impact of hydrocarbon revenues on countries with different governance and administration arrangements. The interaction terms include corruption, and government stability. In alternative specifications, we drop the corruption variable since it is also part of the causal effect we are trying to estimate. In other words, corruption could be seen as one of the channels through which hydro- carbon revenues offsets non-hydrocarbon revenues. Estimates from this approach would however be misleading if reasons for corruption in a country stems from other factors other than the governance and insti- tutional challenges oil rents bring. The summary statistics for our extended sample in Table 2 is based on column 2 of Table 5. There are three new variables introduced in Table 2: government stability index, bureaucratic quality index and investment profile index. These all come from the International Country Risk Group (ICRG) database. Government stability index assesses the strength of a gov- ernment by giving a measure of the ability of a government to stay in office without the risk of being overthrown. The measure is scaled from 0 to 12 with zero capturing the lowest risk and 12 capturing the highest risk. There are three sub-components: government unity, popular sup- port and legislative strength, with each scaled from 0 to 4. Bureaucratic quality measures the effectiveness of the institutions of state to carry out their mandate irrespective of changes in government. The index ranks countries from 0 to 4. Countries that score higher marks have strong institutions with effective means of maintaining the quality of their human resource base. The index on investment profile combines three sub-components, which measure risks relating to the ability of a country to enforce contracts, the extent of payment delays and the ease with which profits of foreign investors could be repatriated. The score ranges from 0 to 4. The higher the score, the higher the risks involved in investing in a country. 3. Empirical results 3.1. Results from step one The results from attempts to replicate the main specification in Table 3 of Bornhorst et al. (2009) are shown in Table 3. Column 1 of Table 3 assesses a bivariate relationship, controlling for only grants as a share of GDP. In columns 2 to 4 we introduce the remaining set of control variables, including both country fixed effects and time effects. Coefficient estimates in columns 2 and 4 are without outliers. In column 1 of Table 3, a percentage point increase in hydrocarbon revenues is associated with a 0.16 percentage point reduction in non-hydrocarbon revenues, holding foreign grants constant. This compares with a decline of 0.19 percentage points decline in non-hydrocarbon revenues as stated in Bornhorst et al. (2009). When additional controls are introduced in column 2, a percentage point increase in hydrocarbon revenues is associated with a 0.17 percentage point decline in non-hydrocarbon revenues compared to an offset of about 0.16 per- centage points in Bornhorst et al. (2009). When outliers (Russia, Kuwait and Norway) are excluded, the non-hydrocarbon revenue offset becomes 0.15 percentage points compared to 0.16 percentage points in Bornhorst et al. (2009). In the fourth column of their table of results, the authors provide limited information on how they apply the Arellano and Bond estimator. I use the first difference option and treat the lag (t-1) of non-hydrocarbon revenues as an endogenous variable. We make use of previous lags from t-2 and earlier as instruments for the lag of non-hydrocarbon revenues. Since their paper indicate that they treat the corruption variable as endogenous, we also make use of previous lags from t-1 and earlier as instruments for corruption. The rest of the explanatory variables then become my list of controls. We also control for time effects while first differencing controls for country effects. From the results in Table 3, the coefficient on the lag of non- hydrocarbon revenue is 0.35 compared to 0.27 in Bornhorst et al. (2009). The long-run effect of a marginal increase in hydrocarbon rev- enue is given by the formula: Long run effect ¼ β1 1� β2 ; where β1 is the coefficient on hydrocarbon revenue as a percentage of GDP and β2 is the coefficient on the lag of non-hydrocarbon revenue as a percentage of GDP. In the long run, a one percentage point increase in hydrocarbon revenues as a percentage of GDP leads to a decline in non-hydrocarbon revenues by 0.24 percentage points of GDP compared to 0.32 percentage points in Bornhorst et al. (2009). Both are statistically significant at the 1 percent level. On the whole, the results are comparable to Bornhorst et al. (2009). The signs, levels of significance and the coefficient on the explanatory variables are largely similar except for a few instances. The differences in those instances could be attributed to some differences in the vari- ables used and rounding errors. For example, in Table 1, the mean values of non-hydrocarbon revenues as a percentage of GDP and non-hydrocarbon openness as a percentage of GDP are approximately 13.5 percent and 63.5 percent respectively whereas Bornhorst et al. (2009) state a value of 15 percent and 53 percent respectively. The next section looks at results from the use of the expanded sample of 77 countries (see Table 2). 3.2. Results from step two – expanded sample Columns 1 to 6 of Table 5 present results on the expanded sample using a fixed effects estimator (i.e. columns 1 and 2), a two-stage least squares estimator (i.e. columns 3 and 4) and an Arellano and Bond estimator without and with our external instrument (i.e. column 5 and 6 respectively). In column 1, a one percentage point increase in hydro- carbon revenues is associated with a reduction in non-hydrocarbon revenues by 0.23 percentage points, holding foreign grants constant. Including the list of control variables reduces the offset in non- hydrocarbon revenues to 0.20 percentage points. In columns 3 and 4, the paper addresses the concern about the potential endogeneity of hydrocarbon revenues with an instrument. The specification also ac- counts for error-in-measurement bias. We define the instrument as the average real world price of crude oil weighted by the amount of oil a country had in 2008, expressed as a percentage of that country’s 2008 Gross Domestic Product in year 2000 prices. Prior to discussing the results, the next couple of paragraphs briefly look at the trend in crude oil prices, the first stage regression and the Table 2 Summary statistics for expanded sample with 77 countries. Variable Obs. Mean Std. Dev. Min Max Non-hydrocarbon revenues (net of grants) as a % of GDP 725 19.2 12.4 -7.4 49.7 Hydrocarbon revenue as a % of GDP 725 8.4 12.4 0 67.8 Grants (net of debt relief) as a % of GDP 725 2.2 3.8 0 27.8 Corruption Perception Index 725 3.0 1.3 0.1 6.0 Log GDP per capita (2000 constant $) 725 8.0 1.6 5.0 10.9 Non-hydrocarbon openness as % of GDP 725 80.7 53.8 15.0 413.9 Government stability index 725 8.8 1.8 3 12 Bureaucratic Quality index 725 2.2 1.1 0 4 Investment Profile index 725 8.0 2.4 0 12 Agriculture as a % of GDP 725 12.4 11.8 0.1 62.0 Note: sample matches regression specification 2 in Table 5. Numbers converted to one decimal place. D.O. Chachu Resources Policy 66 (2020) 101656 5 strength of the instrument. The trend in the average world (real) price of crude oil and the nominal prices from 1992 to 2008 as shown in Fig. 1. The real price of crude oil fluctuated in the decade between 1992 and 2002. However, after 2002 to the end period, crude oil prices were on an upward trend (see Fig. 1). Table 4 shows the first stage regression. The first stage regression indicates a strong correlation between hydrocarbon revenue and the instrument. The test of relevance also yields an F-test value of 23.64 (against a minimum benchmark of 10) however this value reduces significantly when I allow for the clustering of standard errors by country (the F-test value of the revenue instrument reduces to 8.94). In either case the instrument remains strongly correlated with hydrocarbon revenues. Table 4: First stage regression with instrument for hydrocarbon revenues. Following from the first stage regression, column 3 and 4 of Table 5 provide estimates of the instrumented regression results. Its shows an offset of 0.24 percentage points in non-hydrocarbon revenues for a percentage point increase in hydrocarbon revenues. Column 5 and 6 provides Arellano and Bond estimates without and with the use of the new instrument, respectively. The result in column 6 indicate that the offset in non-hydrocarbon revenues is largely unaffected by the use of the new instrument. The offset in non-hydrocarbon revenues changes from 0.17 percentage points in column 5 to about 0.18 percentage points in column 6. In all specifications, the effect of a percentage point increase in foreign grants, offsets non-hydrocarbon revenues by more the twice the offset caused by hydrocarbon revenues. The results seem to suggest that dependence on foreign grants is unhelpful to the effort of countries to mobilize domestic resources. There could however be the argument that the result could stem from a case of reverse causality. That is, countries that have low non-oil revenues attract foreign grants. The result none- theless corroborates Knack (2009), who finds that foreign grants tend to reduce the incentive for countries to implement tax reforms. He also finds that this effect is even stronger than the effect of natural resource rents. The results also indicate that countries that have a large agricultural sector face a huge hurdle as far as mobilizing domestic revenues are concerned. This result is statistically significant across all specifications and can be attributed to the highly informal and subsistent nature of the agricultural sector in many developing countries. Data is also lacking on production and consumption levels. Taxing the agricultural sector carries with it significant administrative costs in developing countries because there are so many small holders. It therefore becomes economically inefficient to incur high administrative costs to chase unpredictable revenues which are in any case transfers. Table 5 also indicates that how open a country is with the rest of the world has a positive effect on non-hydrocarbon revenues however the effect is not statistically significant in its difference from zero. Similarly, the effect of an increase in corruption perception on non-hydrocarbon revenues is positive but statistically insignificant. In columns 5 and 6 however, the positive effect of corruption becomes statistically signifi- cant. Following Bornhorst et al., 2009, we correct for its endogeneity by Fig. 1. Annual crude oil prices: 1992–2008. Table 3 Replicating Bornhorst et al. (2009). We leave out alternative specifications that use hydrocarbon and non-hydrocarbon GDP as denominators for hydrocarbon and non-hydrocarbon revenues respectively. First is the challenge with data availability, accuracy and consistency. The second is the fact that they do not significantly impact the main results. Dependent variable: non-hydrocarbon revenue as a percentage of GDP Independent variables (1) (2) (3) (4) FE FE FE GMM Hydrocarbon revenues as a % of GDP -0.164*** (0.0236) -0.169*** (0.0286) -0.154*** (0.0270) -0.156*** (0.0436) Non-hydrocarbon revenues as a % of GDP (t -1) 0.354*** (0.110) Grants (net of debt relief) as a % of GDP -0.688*** (0.0488) -1.188*** (0.116) -1.157*** (0.114) -1.296*** (0.170) Corruption Perception Index 0.307 0.236 0.117 (0.218) (0.220) (0.254) Non-hydrocarbon trade openness -0.00786 (0.00647) -0.00883 (0.00641) 0.0123 (0.00764) Ln GDP per capita in 2000 constant $ -0.793 (0.776) -0.0876 (0.827) 2.265 (1.820) Agriculture as a % of GDP -0.173*** -0.136** -0.0289 (0.0659) (0.0648) (0.102) Observations 381 298 266 249 Country effect Yes Yes Yes Yes Time effect Yes Yes Yes Yes Outliers excluded No No Yes No R-squared 0.952 0.965 0.889 Number of countries 29 26 23 26 Based on Long run effect Std. Error P>|z| [95% Interval] Confidence Column 4 -0.24 0.05 0.000 -0.33 - 0.15 Robust standard errors in parentheses; ***, **, * represent 1%, 5% and 10% levels of significance respectively. Table 4 First stage regression. Dependent variable: Hydrocarbon revenue as a % of GDP VARIABLES (1) Instrument for Hydrocarbon Revenue 0.139*** (0.0289) Grant (net of debt relief) as a % of GDP 0.0494 (0.0587) Corruption Perception Index -0.393* (0.203) Non-hydrocarbon trade openness 0.0332** (0.0131) Log GDP per capita in 2000 constant US$ 0.307 (1.959) Agriculture as a % of GDP -0.0233 (0.0510) Observations 709 Number of countries 61 Robust standard errors in parentheses ***, **, * represent 1%, 5% and 10% levels of significance respectively. With country fixed effect and time effects. D.O. Chachu Resources Policy 66 (2020) 101656 6 using its lagged values as instruments. The result is however contrary to what one would expect. A plausible storyline could however be that corrupt politicians care about the size of the revenue cake. The bigger the size, the bigger their share. Therein lies an implicit incentive to grow fiscal capacity. The coefficient on the lag (t-1) of non-hydrocarbon revenues is about 0.18. Table 6 shows that, in the long-run, the effect of a percentage point increase in hydrocarbon revenues offsets non-hydrocarbon revenues by 0.21 percentage points compared to 0.24 percentage points in the 30- country sample of oil producers. 3.3. Robustness checks with alternative specifications In Table 7, we introduce a set of institutional variables to account for their potential role in influencing non-resource revenue effort. An argument can be made that government instability in a country can discourage investment by both citizens of that country as well as for- eigners. Such a situation could contribute to reducing the tax base. The level of government instability might also be correlated with the amount of hydrocarbon revenues a country would want to raise. In one instance, a country might be willing and able to raise more hydrocarbon revenues in order to appease factions or finance conflicts. In another instance, political instability might reduce a country’s ability to raise hydrocar- bon revenues. Nigeria offers a classic example. Conflicts in the Niger delta region, where most of the oil is produced, undermines production and therefore revenue mobilization efforts. The situation described might bias upward the estimated effect of hydrocarbon revenues on non- hydrocarbon revenues if the level of political stability is not controlled for. In column 1 of Table 7, controlling for government stability slightly reduces the offset in non-hydrocarbon revenues although the govern- ment stability variable turns out not statistically significant at conven- tional levels. A percentage point increase in hydrocarbon revenues is associated with a decline in non-hydrocarbon revenues by about 0.2 percentage points. Column 2 shows that the magnitude of the coefficient on hydrocar- bon revenue is largely unaffected with the inclusion of bureaucratic quality. Moreover, the latter turns out statistically insignificant at con- ventional levels. The sign on bureaucratic quality is not typically what one would expect as it suggests that an improvement in bureaucratic quality is associated with a reduction in non-hydrocarbon revenues. However, this effect is muted due to large standard errors. Controlling for all three measures of institutional quality and stability in column 3 also leaves the coefficient on hydrocarbon revenues unchanged. The coefficient on government stability, bureaucratic quality and investment profile index are all statistically insignificant in their difference from zero. In column 4, the study investigates whether there are differences in the levels of non-hydrocarbon revenue offset for an increase in hydro- carbon revenues in countries that have more stable governments compared to those that have no stability or are less stable governments. Thus, our interest is in the coefficient of the interaction term between hydrocarbon revenues and government stability. The evidence suggests a worse offset in non-hydrocarbon revenues for less or non-stable gov- ernments however the interaction term between hydrocarbon revenue and government stability is not statistically significant at conventional levels. Column 5 also suggests that there is no evidence of a non-linear relationship between hydrocarbon revenues and non-hydrocarbon rev- enues. The coefficient of the squared term of hydrocarbon revenues comes out as statistically insignificant in its difference from zero. The grants and agriculture variables remain statistically and economically significant across all specifications. A 10 percentage point increase in share of agriculture to GDP reduces non-hydrocarbon reve- nues by about 2 percentage points across all specifications in Table 8. Similarly, a 10 percentage point increase in foreign grants reduces non- hydrocarbon revenues by about 8 percentage points. The coefficient on corruption perception index is statistically insignificant in its difference from zero across all specifications. In another set of specifications based on Table 8, we drop the corruption variable and find that the results are largely unaffected. From Table 8, the interaction term between hydrocarbon revenues and corruption perception index indicates that the offset in non- hydrocarbon revenues for an increase in hydrocarbon revenues is exacerbated with increasing levels of corruption. The interaction term is however not statistically significant in its difference from zero. In col- umn two, we introduce the instrument for hydrocarbon revenues as Table 5 Re-examining the evidence: results from expanded sample. Dependent variable: non-hydrocarbon revenue as a percentage of GDP Independent Variables (1) FE (2) FE (3) 2SLS (4) 2SLS (5) GMM (6) GMM Hydrocarbon revenues as a % of GDP -0.233*** (0.0266) -0.201*** (0.0272) -0.238*** (0.0725) -0.238* (0.122) -0.173*** (0.0541) -0.177*** (0.0539) Non-hydrocarbon revenues as a % of GDP (t -1) 0.181** (0.0874) 0.179** (0.0869) Grants (net of debt relief) as a % of GDP -0.706*** (0.0570) -0.839*** (0.111) -0.842*** (0.109) -0.842*** (0.193) -0.594*** (0.126) -0.598*** (0.126) Corruption Perception Index 0.172 (0.183) 0.230 (0.199) 0.230 (0.336) 0.598** (0.247) 0.587** (0.248) Non-hydrocarbon trade openness 0.0109 (0.00746) 0.0103 (0.00731) 0.0103 (0.0123) 0.0112 (0.00773) 0.0115 (0.00773) Log GDP per capita in 2000 constant US$ -0.925 (0.935) -0.708 (1.008) -0.708 (1.733) 1.064 (2.643) 1.020 (2.626) Agriculture as a % of GDP -0.219*** (0.0647) -0.219*** (0.0638) -0.219* (0.112) -0.0976* (0.0591) -0.0990* (0.0594) Country effect Yes Yes Yes Yes Yes Yes Time effect Yes Yes Yes Yes Yes Yes Instrumented No No Yes Yes No Yes Observations 986 725 709 709 584 584 R-squared 0.922 0.948 0.245 0.245 Number of countries 76 61 61 61 60 60 Robust standard errors in parentheses; ***, **, * represent 1%, 5% and 10% levels of significance respectively. Column 4 has robust clustered (by country) standard errors in parenthesis. Columns 5 and 6 report Arellano and Bond estimates with robust standard errors in parenthesis. Table 6 Arellano-Bond estimator: Estimation of Long-run effect of hydrocarbon revenues. Based on Long run effect Std. Error. P>|z| [95% Interval] Confidence Column 5 -0.21 0.06 0.000 -0.32 -0.10 Column 6 -0.23 0.06 0.000 -0.33 -0.11 D.O. Chachu Resources Policy 66 (2020) 101656 7 described in equation (2) while controlling for government stability index and find that a percentage point increase in hydrocarbon revenues offsets non-hydrocarbon revenues by approximately 0.26 percentage points. Columns 3 and 4 then report results on Arellano and Bond estimates without and with instrumenting for hydrocarbon revenues respectively while controlling for government stability. Including government sta- bility in the list of control variables does not change the basic results of the Arellano and Bond estimates although instrumenting for hydrocar- bon revenues worsens the decline of non-hydrocarbon revenues only marginally. The long run effect of hydrocarbon revenues is identical to the results recorded in Table 8. The offset in non-hydrocarbon revenues for a marginal increase in hydrocarbon revenues remain statistically robust at conventional levels across all specifications. 4. Conclusion The paper re-examines the evidence that hydrocarbon revenues offset a country’s effort to mobilize non-hydrocarbon revenues. We do this by first replicating the main econometric specifications in Table 3 of Bornhorst et al. (2009). Our contribution is two-fold: we expand the data and then explore the effect of other measures of institutional quality on the relationship between hydrocarbon revenues and non-hydrocarbon revenue effort. Using an instrumental variable approach, the paper find that the non-hydrocarbon revenue offset is about 2 to 3 percent of GDP for a 10-percentage point increase in hydrocarbon revenues. Gov- ernment stability, bureaucratic quality and the protection of foreign capital do not seem to significantly matter in improving non-resource revenue effort. Further work remains in this area. This includes global efforts in providing a methodologically consistent and extensive dataset on resource revenues as well as other fiscal data. With regards to the latter, database initiatives being pursued by the United Nations Uni- versity’s World Institute of Development Economics Research, the In- ternational Monetary Fund, among others are critical to increasing our understanding of how natural resource wealth impact on domestic revenue efforts. They would have implications for further research in this area, including providing insights into how resource-rich devel- oping and emerging countries can leverage resource wealth to expand other sectors of the economy (and therefore the tax base) while main- taining a stable statutory tax rate. Declaration of competing interest None. CRediT authorship contribution statement Daniel Ofoe Chachu: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Writing - original draft, Writing - review & editing. Acknowledgement I would like to acknowledge Prof. Jon Bakija of Williams College, MA, USA for guidance and very valuable feedback. Table 7 Re-examining the evidence with alternative specifications I. Dependent variable: non-hydrocarbon revenue as a percentage of GDP Independent variable (1) FE (2) FE (3) FE (4) FE (5) FE Hydrocarbon revenues as a % of GDP -0.206*** (0.0507) -0.208*** (0.0497) -0.206*** (0.0499) -0.292*** (0.0938) -0.264*** (0.110) Grants (net of debt relief) as a % of GDP -0.839*** (0.206) -0.837*** (0.202) -0.840*** (0.204) -0.827*** (0.209) -0.840*** (0.211) Corruption Perception Index 0.117 (0.331) 0.224 (0.347) 0.300 (0.354) 0.134 (0.335) 0.178 (0.324) Non-hydrocarbon trade openness 0.0109 (0.0134) 0.0105 (0.0131) 0.00832 (0.0137) 0.0112 (0.0132) 0.0107 (0.0133) Log GDP per capita in 2000 constant US $ -1.068 (1.624) -0.893 (1.682) -0.531 (1.747) -0.996 (1.574) -0.661 (1.613) Agriculture as a % of GDP -0.220* (0.119) -0.223* (0.118) -0.226* (0.116) -0.220* (0.119) -0.218* (0.120) Government stability index 0.123 (0.136) 0.0953 (0.131) 0.170 (0.132) 0.0653 (0.163) Bureaucratic quality index -0.454 (0.373) -0.252 (0.393) Investment profile index -0.273 (0.168) Hydrocarbon revenues*Government stability 0.00832 (0.00977) (Hydrocarbon revenues)2 0.00124 (0.00150) Country effect Yes Yes Yes Yes Yes Time effect Yes Yes Yes Yes Yes Instrumented No No No No No Observations 725 725 725 725 725 R-squared 0.948 0.948 0.949 0.948 0.948 Number of countries 61 61 61 61 61 Robust clustered (by country) standard errors in parentheses; ***, **, * represent 1%, 5% and 10% levels of significance respectively. Table 8 Re-examining the evidence with alternative specifications II. Dependent variable: non-hydrocarbon revenue as a percentage of GDP Independent Variable (1) FE (2) FE (3) GMM (4) GMM Hydrocarbon revenues as a % of GDP -0.175*** (0.0685) -0.259*** (0.127) -0.173*** (0.0520) -0.177*** (0.0518) Non-hydrocarbon revenues as a % of GDP (t -1) 0.181** (0.0869) 0.179** (0.0865) Grants (net of debt relief) as a % of GDP -0.842*** (0.210) -0.842*** (0.185) -0.594*** (0.127) -0.598*** (0.127) Corruption Perception Index 0.243 0.156 0.596** 0.583** (0.370) (0.348) (0.243) (0.244) Non-hydrocarbon trade openness 0.0116 0.0103 0.0111 0.0114 (0.0128) (0.0122) (0.00754) (0.00755) Log GDP per capita in 2000 constant US $ -0.884 (1.611) -0.863 (1.712) 1.125 (2.635) 1.079 (2.616) Agriculture as a % of GDP -0.222* -0.220** -0.0974 -0.0988* (0.119) (0.111) (0.0593) (0.0596) Government stability index 0.174 -0.00938 -0.00541 (0.126) (0.151) (0.150) Hydrocarbon revenues*Corruption Perception Index -0.0146 (0.0198) Country effect Yes Yes Yes Yes Time effect Yes Yes Yes Yes Instrumented No Yes No Yes Observations 725 709 584 584 R-squared 0.948 0.245 Number of countries 61 61 60 60 Robust clustered (by country) standard errors in parentheses; ***, **, * represent 1%, 5% and 10% levels of significance respectively. Column 3 and 4 report Arellano and Bond estimates with robust standard errors. D.O. Chachu Resources Policy 66 (2020) 101656 8 APPENDIX A. List of Countries Algeria Iran Albania Mauritius Angola Kazakhstan Cape Verde Moldova Azerbaijan Kuwait Cyprus Morocco Bahrain Libya Dominican Republic Namibia Bahamas, The Mexico Fiji Nicaragua Bhutan Nigeria Finland Panama Burkina Faso Norway Ghana Paraguay Burundi Oman Greece Portugal Brunei Qatar Honduras Senegal Cameroon Russia Jamaica Seychelles Chad Saudi Arabia Kenya Sierra Leone Congo Sudan Kyrgyz Republic Singapore Ecuador Syria Latvia Slovenia Equatorial Guinea Trinidad and Tobago Lebanon Spain Gabon UAE Luxembourg Sri Lanka Indonesia Venezuela Madagascar Switzerland Iceland Vietnam Maldives Tajikistan Ireland Yemen Mali Togo Israel Zambia Malta Uganda B. Table 3 of Bornhorst et al., (2009) (European Journal of Political Economy) D.O. Chachu Resources Policy 66 (2020) 101656 9 References Abdulahi, M.E., Shu, Y., Khan, M.A., 2019. Resource rents, economic growth, and the role of institutional quality: a panel threshold analysis. Resour. Pol. 61, 293–303. African Development Bank Group, 2010. Domestic Resource Mobilization across Africa: Trends, Challenges and Policy Options. Committee of Ten Meeting Washington DC. October 2010. Arellano, Manuel, Bond S, 1991. Some test of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev. Econ. Stud. 58, 277–297. Badeeb, R.A., Lean, H.H., Clark, J., 2017. The evolution of the natural resource curse thesis: a critical literature survey. Resour. Pol. 51, 123–134. https://doi.org/ 10.1016/j.resourpol.2016.10.015. December 2016. Bhattacharyya, S., Conradie, L., Arezki, R., 2017. Resource discovery and the politics of fiscal decentralization. J. Comp. Econ. 45 (2), 366–382. https://doi.org/10.1016/j. jce.2017.02.003. Bildirici, M.E., Gokmenoglu, S.M., 2020. Precious metal abundance and economic growth: evidence from top precious metal producer countries. Resour. Pol. 65, 101572. https://doi.org/10.1016/j.resourpol.2019.101572. December 2019. Bornhorst, F., Gupta, S., Thornton, J., 2009. Natural resource endowments and the domestic revenue effort. Eur. J. Polit. Econ. 25 (4), 439–446. https://doi.org/ 10.1016/j.ejpoleco.2009.01.003. Collier, P., 2010. The Plundered Planet: Why We Must – and How We Can – Manage Natural Resources for Global Prosperity. Oxford University Press, Inc. 198 Madison Avenue, New York, New York, p. 10016. Collier, P., Hoeffler, A., 2009. Testing the neocon agenda: democracy in resource-rich societies. Eur. Econ. Rev. 53. Pg. 293 – 308. Collier, P., Goderis, B., 2007. Commodity Prices, Growth and the Natural Resource Curse: Reconciling a Conundrum. CSAE WPS/2007-15. Department of Economics, University of Oxford, Oxford, United Kingdom. Collier, P., 2007. The Bottom Billion: Why the Poorest Countries Are Failing and what Can Be Done about it. Oxford University Press, Inc. 198 Madison Avenue, New York, New York, p. 10016. Crivelli, E., Gupta, S., 2014. Resource blessing, revenue curse? Domestic revenue effort in resource-rich countries. Eur. J. Polit. Econ. 35 https://doi.org/10.1016/j. ejpoleco.2014.04.001. Davis, A.G., Tilton, E.J., 2005. The natural resource curse, 2005 Nat. Resour. Forum 29, 233–242. United Nations. Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden Street, MA 02148, USA. Deaton, A., Miller, R.I., 1995. International Commodity Prices, Macroeconomic Performance, and Politics in Sub-saharan Africa. International Finance Section, Department of Economics, Princeton University, Princeton, NJ. Gnangnon, S.K., Brun, J.F., 2019. Tax reform and fiscal space in developing countries. Eurasian Economic Review 1–29. Gupta, S. Abhijit, 2007. “Determinants of Tax Revenue Effort in Developing Countries.” International Monetary Fund Working Paper WP/07/184. Jensen, A.D., 2011. State-building in resource-rich economies. Atl. Econ. J. 171–193. https://doi.org/10.1007/s11293-011-9269-z. June. Keen, M., Mansour, M., 2009. Revenue in mobilization in sub saharan Africa: challenges from globalization. IMF Working 1–48. WP 09/157. Khattry, B., Rao, J.M., 2002. Fiscal faux pas? An analysis of the revenue implications of trade liberalization. World Dev. 30 (8), 1431–1444. Klein, N., 2010. The linkage between the oil and non-oil sectors – a Panel VAR approach. IMF Working Paper WP 10, 118. Knack, S., 2009. Sovereign rents and quality of tax policy and administration. J. Comp. Econ. 37, 359–371, 2009. Matallah, S., 2020. Economic diversification in MENA oil exporters: understanding the role of governance. Resour. Pol. 66, 101602. https://doi.org/10.1016/j. resourpol.2020.101602. January. Mawejje, J., 2019. Natural resources governance and tax revenue mobilization in sub saharan Africa: the role of EITI. Resour. Pol. 62, 176–183. https://doi.org/10.1016/ j.resourpol.2019.04.001. April. Mohtadi, H., Ross, M.L., Ruediger, S., 2016. Oil , taxation and transparency (vol. 55108). Revisited. Resour. Pol. 30 (2), 1–40. https://doi.org/10.1016/j. resourpol.2005.05.001. Ossowski, R., Gonz�ales, A., 2012. Manna from Heaven: the Impact of Nonrenewable Resource Revenues on Other Revenues of Resource Exporters in Latin America and the Caribbean (No. IDB-WP-337). IDB Working Paper Series. Oyarzo, M., Paredes, D., 2019. Revisiting the link between resource windfalls and subnational crowding out for local mining economies in Chile. Resour. Pol. 64, 101523. Rodrik, D., 1998. Why do more open economies have bigger governments? J. Polit. Econ. 106 (5), 997–1031. Sachs, Jeffrey, Warner, Andrew, 1995. Natural Resource Abundance and Economic Growth. In: National Bureau of Economic Research Working Paper Series 1–47. W5398. Thomas, M.A.H., Trevino, M.J.P., 2013. Resource Dependence and Fiscal Effort in Sub- saharan Africa (No. 13-188). International Monetary Fund. Yanikkaya, H., Turan, T., 2018. Curse or blessing? An empirical Re-examination of natural resource-growth nexus. J. Int. Dev. 1–19. https://doi.org/10.1002/jid.3374. Zodrow, G.R., 2006. Capital mobility and source-based taxation of capital income in small open economies. Int. Tax Publ. Finance 13 (2–3), 269–294. D.O. Chachu http://refhub.elsevier.com/S0301-4207(19)30907-9/sref1 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref1 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref2 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref2 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref2 http://refhub.elsevier.com/S0301-4207(19)30907-9/optciTzsg6HUn http://refhub.elsevier.com/S0301-4207(19)30907-9/optciTzsg6HUn http://refhub.elsevier.com/S0301-4207(19)30907-9/optciTzsg6HUn https://doi.org/10.1016/j.resourpol.2016.10.015 https://doi.org/10.1016/j.resourpol.2016.10.015 https://doi.org/10.1016/j.jce.2017.02.003 https://doi.org/10.1016/j.jce.2017.02.003 https://doi.org/10.1016/j.resourpol.2019.101572 https://doi.org/10.1016/j.ejpoleco.2009.01.003 https://doi.org/10.1016/j.ejpoleco.2009.01.003 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref11 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref11 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref11 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref12 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref12 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref13 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref13 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref13 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref14 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref14 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref14 https://doi.org/10.1016/j.ejpoleco.2014.04.001 https://doi.org/10.1016/j.ejpoleco.2014.04.001 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref17 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref17 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref17 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref18 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref18 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref18 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref19 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref19 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref20 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref20 https://doi.org/10.1007/s11293-011-9269-z http://refhub.elsevier.com/S0301-4207(19)30907-9/sref22 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref22 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref23 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref23 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref24 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref24 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref26 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref26 https://doi.org/10.1016/j.resourpol.2020.101602 https://doi.org/10.1016/j.resourpol.2020.101602 https://doi.org/10.1016/j.resourpol.2019.04.001 https://doi.org/10.1016/j.resourpol.2019.04.001 https://doi.org/10.1016/j.resourpol.2005.05.001 https://doi.org/10.1016/j.resourpol.2005.05.001 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref30 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref30 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref30 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref31 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref31 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref31 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref32 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref32 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref37 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref37 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref37 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref34 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref34 https://doi.org/10.1002/jid.3374 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref36 http://refhub.elsevier.com/S0301-4207(19)30907-9/sref36 Domestic revenue displacement in resource-rich countries: What’s oil money got to do with it? 1 Introduction 1.1 Overview of the “resource curse” 1.2 Resource rents and tax revenue mobilization 2 Empirical framework and description of data 2.1 Step one: replicating Bornhorst et al. (2009) 2.2 Step two: Re-examining the evidence from a larger sample 3 Empirical results 3.1 Results from step one 3.2 Results from step two – expanded sample 3.3 Robustness checks with alternative specifications 4 Conclusion Declaration of competing interest CRediT authorship contribution statement Acknowledgement APPENDIX Acknowledgement A. List of Countries B. Table 3 of Bornhorst et al., (2009) (European Journal of Political Economy) References