The Impact of Oil Rent, Currency Overvaluation, and Institution Quality, on Economic Growth of Oil-Rich Countries: A Heterogeneous Panel Data Study

This paper investigates several key challenges faced by oil-rich countries regarding their economic growth and development. First, it discusses how to determine currency overvaluation for these countries (if any). To determine the overvaluation, the real exchange rate (RER) is calculated and the Balassa–Samuelson effect is estimated via a regression model. Next, the study presents an empirical model for assessing the impact of oil rent on economic growth in the context of currency overvaluation and the institutional quality in every country. As a dynamic model, both endogeneity and heterogeneity are expected across cross-sections because countries are different in culture, customs, and political institutions. Consequently, heterogeneous panel data analysis is undertaken using the error correction model cointegration technique and the mean group estimation method in an autoregressive distributed lag model. Finally, the study concludes the findings and provides policy recommendations by offering a new perspective on an ongoing dilemma, discussing the challenges and limitations facing developing oil-rich countries and how their path to success may differ from other countries.


INTRODUCTION
This study aims to investigate the role of currency overvaluation, oil rent, and institutional quality on the economic growth of oil-rich countries. 'Oil-rich country' is defined as one for which oil exports comprise 10% or more of its GDP (Mehlum et al., 2006). Because the driving force of growth changes during different stages in economic development, economists analyze the composition of various factors in different countries and regions. Such models look closely at the impact of structural change as well as political and financial institutions on economic growth and the effect of a variety of subfactors including distribution conflicts, market failures, credit constraints, migration, urbanization, changes in mortality and fertility rates through health improvement, inequality, and trade.
One key to evaluating modern economic growth is to examine the causes underlying cross-country differences in income per capita (Acemoglu, 2009). As the focus in macroeconomic modeling shifts from short-to long-run relationships amongst variables, it becomes necessary to observe the interaction of variables over time. Hence, a dynamic approach, or the use of lag variables, is an essential part of current macroeconomic models. When considering an oil price shock, a change in the money supply, or a change in a country's trade policy, understandably, it may take more than a year or two to see the full effect of these shocks or disturbances on the economy. Similarly, one's consumption decisions are most likely related to last month's salary more than to this month's. Moreover, since there are more detailed data available for many countries, by combining the crosssectional data of different countries to time series data, modern This Journal is licensed under a Creative Commons Attribution 4.0 International License macroeconomic models are not only more dynamic but can also make a more efficient analysis.
Thus, this study aims to investigate how, if at all, natural resources' rent, specifically for crude oil, causes slower economic growth rates in oil-rich countries with lower quality institutions. The study analysis also includes an exploration of how the combination of natural resources' rent and overvaluation may cause growth rates slow even more. Moreover, this study explores how natural resources' rent works and the role of exchange rates.
Although academic literature and empirical studies in this area are plentiful, this research offers three new contributions to the existing literature. First, as it has been suggested that petroleum creates the majority of problems for most countries (Ross, 2012), this research distinguishes between the types of natural resources by focusing solely on crude oil. Furthermore, this study differentiates between natural resource abundance (reserves) and dependency (exports). Brunnschweilera and Bulte (2008) were pioneers in distinguishing between these two aspects; however, unlike these scholars, it appears that natural resource dependency (i.e., the ratio of crude oil exports to GDP) is a more meaningful variable to demonstrate how the economic growth of oil-rich countries may be related to their natural resources. Secondly, the study introduces an interaction variable 'OilRent-Overvaluation' to better explains the rent-seeking impact on growth. Third, this study employs a heterogeneous panel data model to address both endogeneity and heterogeneity problems, as Cavalcanti et al. (2011) argued that cross-country heterogeneity could produce biased and misleading results.

Oil-Rich Countries
The phrase 'oil-rich country' may be indicative of either a country's oil reserves by volume or the percentage their GDP comprisedc of oil exports. Oil reserves measure the oil abundance of a country, whereas the oil export percentage to GDP ratio refers to the oil dependence of a country (Table 1). The Organization of the Petroleum Exporting Countries (OPEC) database lists 33 countries as the world's oil exporters based on their barrel per day exports.
Countries such as Syria and Libya do not have enough data, and data for Russia is only available after 1990 following the collapse of the Soviet Union; hence, they have been removed from consideration.
For the analysis of oil dependency in countries, it is more common to consider those countries for which oil exports comprise 10% or more of their GDP. Including this factor removes another four countries, i.e., Vietnam, China, Equatorial Guinea, and Sudan. The remaining 26 countries were included in most regression models in this study using data for the period between 1980 and 2017. Saudi Arabia, which exports 7.5 million oil barrels per day, is the biggest oil exporter worldwide and Trinidad and Tobago, which exports 65,000 barrels per day, is the smallest exporter ( Figure 1).

Rent-Seeking
In his recent book, The Price of Inequality: how today's divided society endangers our future, Stiglitz (2013) explains rent-seeking, noting that "the term 'rent' was originally used to describe the returns to land" (49) because the landlord received such payments based on his ownership of the land and not because of anything he did. The term was then extended to include monopoly rents, which is the income one receives from the control of a monopoly that is usually an exclusively government-granted right, for instance, to import a limited amount or quota of a good (Stiglitz, 2013). Natural resource-rich countries are known for rent-seeking activities as it is "far easier to get rich in these countries by gaining access to resources at favorable terms than by producing wealth" (39-40). In this way, wealth can be taken away from others, creating an unequal balance of power and slower economic growth. On the other hand, rent-seeking can be a mechanism to "capture the economic rent arising from price distortions and physical controls caused by excessive government intervention, such as licenses, quotas, interest rate ceilings, and exchange control" (Todaro and Smith, 2012. p. 522). Acemoglu et al. (2008) claim that institutions are the fundamental cause of differences in economic growth  (Acemoglu et al., 2008). Mehlum et al. (2006) claim that the quality of institutions plays a crucial role in growth diverging experiences across resourcerich economies the resource curse since the growth performance differences among resource-rich countries is mainly due to how the institutional arrangement distributes resource rents. Aslaksen (2008) suggests that Oil has adverse impoct on corruption in both democratic and nondemocratic countries. However, Bhattacharyy and Hodler (2010) argue that natural resource feeding corruption depends on the quality of democratic institutions. Hence, it is difficult, if not impossible, to talk about the role of natural resources in economic growth without considering the quality of institutions in the model.

Institutions
How to measure the quality of institutions collectively is a challenging task. Fortunately, in recent years several organizations compile reliably more comprehensive data in this regard in a historical database that can be used easily in regression models ( Table 2). Two of them that more being cited by scholars are Freedom House (FH) and Polity IV. FH provides data for many countries from 1972 in two main categories: political rights (PR) and civil liberties (CL). PR and CL measured on a 1 to 7 scale, with one,1, representing the highest degree of freedom and seven,7, the lowest. Countries whose combined average ratings for PR and CL fall between 1.0 and 2.5 are designated "free," between 3.0 and 5.0 "partly free," and between 5.5 and 7.0 "not free." Polity IV collects, creates, and provides data for more countries and more years relating to countries' political regimes, characteristics, and transitions. Recently, they created a specific time series variable named POLITY2 to describe a country's quality of institutions based on a combination of two factors: DEMOC, which determines institutionalized democracy based on (1) the presence of institutions and procedures that citizens can express actual preferences about alternative policies and leaders, (2) the existence of traditional constraints on the exercise of power by the executive, and (3) the guarantee of CL to all citizens in their daily lives and acts of political participation; and AUTOC, which measures institutionalized autocracy based on the lack of legalized political competition and political freedoms. The POLITY2 variable computed on a scale of +10 (solidly democratic) to -10 (strongly autocratic) by subtracting the AUTOC score from the DEMOC score 1 . Several countries, such as Brunei and Colombia, have not been rated by Polity IV. Therefore, in some regressions, either FH data used for all countries or the POLITY2 equivalent of FH data calculated and replaced for these two countries. Some countries in this study have a solid score for all of the study period . The scores for some other countries have improved during the years or declined. The table shows the average score of all years for each country.
The Economist Intelligence Unit created the Figure 2 for the 2019 Democracy index. According to this index, all developing oil-rich countries in the Middle East, North Africa, and South America are in authoritarian regimes category either dark or light orange color.

Exchange Market
For a pure floating regime, without central bank intervention, the exchange market is like any other commodity market, in which demand and supply of a commodity determine its price. If a nation's total demand for a foreign currency exceeds its total foreign exchange earnings, the exchange rate increase; therefore, 1 More details about these variables can be found in POLITY IV Project Dataset Users' Manual 2017. the local currency will depreciate, and the foreign currency will appreciate. However, in the real world, no currency is free of central bank intervention.
Also essential to the exchange market is inflation, which, in developing countries, is generally demand-side and largely has roots in the excess money supply. Here, the institution of an 'independent central bank' plays a key role in preventing excess money supply and resultant inflation. However, in many developing countries, central banks are not independent. Therefore, in these countries, governments usually resolve their budget deficit problems easily by ordering their central bank to print money directly or increase their money supply indirectly through other means.

Oil Rent and Currency Overvaluation
Now imagine a country like Iran, which has between 30% and 58% annual inflation over the past several years. If the Iranian central bank keeps the fixed, dollar-pegged exchange rate year after year, should we expect Iranian currency to be heavily overvalued? Two points are important to consider here: first, natural resources such as oil and gas are under government control in most countries and the oil rent is a vast part of the government's revenue. Therefore, it is very important to consider who has access to the trickle-down of this wealth, which may be a source of government corruption. Secondly, one must consider the overvalued exchange rate. If a country keeps printing money and still defends fixed and pegged exchange rates, this leads to overvaluation and encourages more capital to exit the country. To prevent such capital flight, countries impose restrictions, often giving rise to a black market for foreign currency exchange, although in some cases, such markets exist openly and are tolerated by the government. More currency being overvalued means that the increased oil rent earnings make this resource more lucrative-for those who have access to it. For this reason, the interaction of the variables of oil rent and an overvalued exchange rate can more fully explain the natural resource cures phenomena.

Data for This Study
At the time of writing, the most current Penn World Table (PWT) was version 9.1, which includes variables data including real GDP, purchasing power parity (PPP), exchange rate, output, consumption, capital formation, and productivity information for 182 countries between 1950 and 2017. For comparison and robustness, data have also been obtained from the Madison Database and World Bank World Development Indicators (WDI). The quality institution's data were obtained from two resources: Freedom House and Polity IV. Data regarding oil exports, reserves, production, and historical prices can be found at the OPEC Oil and Gas Data website 2 . However, at the time of this writing, only data through 2013 was available on that website. OPEC's Data Services Department provided the author with the full database, including data through 2017.
This study also includes an interaction variable in the model. It is oil rent multiplied overvaluation. For robustness, the author collected both oil rent data and total rent data. The latter is the ratio between the total rent on all of a country's natural resources and its GDP. However, in the case of the oil-rich countries included in this study, these data are mainly the total of crude oil and natural gas rent to the country's GDP. Thus, the second interaction variable is either the component of oil rent or total rent concerning the overvaluation data. In the same way, the first interaction variable is either the component of POLITY2 or FH data and oil rent or total rent data.

LITERATURE REVIEW
According to Helpman (2004) and Barro and Sala-i-Martin (2004), the study of economic growth can be separated into two 'waves 2 See https://www.opec.org/library/Annual%20Statistical%20Bulletin/ interactive/current/FileZ/Main-Dateien/oilgasdata.html. of research.' Although classical economists from Adam Smith to Thomas Malthus, Frank Ramsey, and Joseph Schumpeter discussed the essential elements of economic theories such as the accumulation of physical and human capital, diminishing returns, and technological progress, the first wave began during the 1950s with the neoclassical Solow-Swan model that had the key assumption of exogenous technological progress and the concept of conditional convergence. The focus of the neoclassical growth model was more on explaining the short-run growth rate and business cycles. The concept of conditional convergence was achieved by removing the assumption that different countries had the same initial economic positions. Allowing such heterogeneity across countries is the main idea behind different economies growing faster from their steady-state values (Barro and Sala-i-Martin 2004).
The second wave began with the work of Romer (1986) and Lucas (1988) and the realization of the crucial importance of long-run growth over the mechanics of business cycles. Taking steps in this path required linking the per capita growth rate with technological progress that can be defined inside the model. Thus, the most important contribution of the second wave economic growth theories was stating the long-run growth rate within the model, termed endogenous growth models. By introducing research and development (R&D) and imperfect competition into the growth model by Romer (1986Romer ( , 1990, Aghion andHowitt (2009), andHelpman (2004), technological progress was the result of planned R&D activities, which are rewarded by some type of monopoly. This technological progress through R&D is the engine of longrun economic growth. Hence, the long-run growth depends on the government via taxation, protection of property law, providing infrastructural investment, regulation of the financial market, international trade and so on. Therefore, government and institutions play a key role in long-run economic growth 3 .

METHODOLOGY
This study was conducted in two phases. First, was the determination of currency over or undervaluation in oil-rich countries. In the second phase, overvaluation data attained in phase one were used to find the relationship between oil rent-overvaluation and economic growth for oil-rich countries considering their quality of institutions.

Phase One: Finding Currency Overvaluation
To determine currency overvaluation, two leading papers were considered: Dollar (1992) and Rodrik (2008). In his influential paper, Dollar (1992) showed a way to calculate, what he called, distortion from PPP or the Balassa-Samuelson effect. Although PPP is a widely used theory of exchange rate determination in international economics, Balassa (1964) and Samuelson (1964) independently provided a theoretical model, known today as the Balassa-Samuelson effect, showing that countries with higher incomes have higher technology advances and, therefore, more productive labor 3 To see the variety of modern economic growth models in more details look at Romer (2001) on tradable goods than low-income countries. Though, as the 'law of one price' does not work on non-tradable goods, by increasing the wages in non-tradable goods in high-income countries, they have more expensive non-tradable goods compared to low-income countries. Consequently, different productivities lead to a deviation from PPP (Asea and Corden, 1994).
From Rodrik (2008), I computed every country's under/ overvaluation for the period of this study in three steps.
Step 1 used the PWT and divided the exchange rate (XR) by the PPP conversion rate to calculate the RER: where i is for countries and t is for time. The PWT versions after Version 7 do not directly provide PPP; however, they provide a variables correspondence link that explains how PPP can be calculated as "pl_gdpe" multiple "xr". Both XR and PPP are national currency units per US dollar (USD). According to the Balassa-Samuelson effect, non-tradable goods are cheaper in less developed countries. In step 2, the RER was adjusted for this effect by regressing lnRER on GDP per capita, using rgdpo (output-side real GDP at chained PPP) divided by population (pop), both obtained from the PWT: where f t is the time fixed effect and u i is the error term. Through robust pooled regression, a very significant β of -0.1125664 was found. Then, by determining the fitted value for each country and year, the Balassa-Samuelson adjusted rate was calculated. In Step 3, the RER was subtracted from the above-adjusted rate: According to Rodrik (2008), when UNDERVAL exceeds unity, the currency is undervalued (i.e., imports are expensive, but exports are relatively cheap). When it is below unity, the currency is overvalued (i.e., imports are cheap, and exports are relatively expensive). Thus, to see the relationship between overvaluation and economic growth, Rodrik (2008) used the following equation or estimation model: growth = α+β lnGDPpc i,t-1 +δ lnUNDERVAL it +f i +f t +u it (4) where GDP per capita annual growth is the dependent variable, f i and f t are the relative country and time fixed effect, respectively. The empirical result obtained using STATA. Despite overvaluation is this study's concern, to maintain consistency between this and Rodrik's (2008) model, I kept the variable name "LnUNDERVALit". However, the interpretation of results should be undertaken with care. According to Rodrik (2008), whenever the UNDERVAL was below unity, the currency was overvalued and when it exceeded unity, the currency was undervalued. Figure 3 shows how overvaluation and economic growth move together in these selected countries.
Also, the results of the time-series regressions for selected countries (Table 3) and panel data fixed effect for all countries (Table 4) all showed a significant positive relationship between currency over/undervaluation and economic growth. This is the expected result as show overvalued currency has a reverse relationship with economic growth. Notice that again when the variable UNDERVAL is below unity, the currency is overvalued and Figure 1 shows how growth rate decrease when the currency is overvalued.
A more recent study by Habib et al. (2016) confirmed Rodrik's (2008) results, and in particular, emphasized that the RER was a more important factor for economic growth in developing countries than in developed countries. They concluded "a strong and statistically significant positive (negative) effect of real depreciation (appreciation) on real per capita growth…" (Habib et al., 2016).

Phase Two: Oil Rent and Economic Growth
This study assumes that factors such as institutions, customs, history, and stage of development may be different amongst countries or between cross-sections, as well as that each country's characteristics may evolve over a long period. Some   countries may experience shocks such as war or revolution or, in the case of global shock, countries may absorb it differently depending on their levels of resilience as determined by their unique infrastructures and institutions. In other words, in dynamic time-variant macroeconomic models, the heterogeneity may not differ only in each cross-section among countries, which can be easily captured through intercept ɑ i , but the slope for each country, the coefficient β i , may also differ over time. For this reason, Pesaran and Smith (1995) showed that none of the FE or RE methods are consistent in this situation. To resolve this bias estimation problem in dynamic panel data with heterogeneous slopes, Pesaran and Smith (1995), Pesaran et al. (1997Pesaran et al. ( , 1999, and Im et al. (2003)  A dynamic (ARDL) model with the assumption that all coefficients can vary across cross-sectional units can be written as 4 : where i = 1, 2,…, N for countries and t = 1, 2,…, T for time, the long-run parameter for country i is: i i 1 i    = − and the MG estimators for the whole panel will be: Nevertheless, as soon as a lag variable is added to the model, autocorrelation between the model's current year and the lag variable is expected. Therefore, the mean and/or variance of a non-stationary time series depends on time (Asteriou and Hall, 2016). Identifying non-stationarity in a model is the same as the presence of unit roots in the model. We know that in the simplest AR (1) model such as y t = Φ y t-1 +ε t, only if Φ = 1 will there be a unit-root and the series will be non-stationary. We also know that, at least in this simple case, we can obtain stationarity by first-order differencing, i.e.: y t -y t-1 = Φ (y t-1 -y t-1 )+ε t ≥ ∆y t =ε t .

Cointegration
The most macroeconomic variables such as GDP, inflation index (CPI) and money supply are time depended or trended therefore in most cases they are non-stationary as their means are constantly increasing. The problem with trended or non-stationary data is that the OLS estimator can be easily inconsistent and biased 5 . The idea of finding a solution that has both short-and long-run results while at the same time maintaining stationarity among all variables has led scholars to produce a model by combining two accumulated error terms, which are called stochastic trends 6 . In a special case that their combination eliminating non-stationarity, the variables are cointegrated. Nevertheless, when variables are cointegrated, even if they may have increasing trends over time, there should be a common trend that links them together. Therefore, most macroeconomic models either are cointegrated or they suffer spurious regression problem. Hence, cointegration is so important in dynamic macroeconomic models.

Error correction model (ECM)
The most popular model that can create cointegrated stationarity is called the ECM, in the simple form of: where Y t and X t have a long-run relationship because t t 1 2 Y --ˆX = μ β β and t~ I(0) μ , and b 1 produces the short-run effect or the immediate impact of a change in X t on Y t . The coefficient 2 β produces the long-run relationship between X t and Y t through the equation: and π is the adjustment effect that shows how much disequilibrium has been corrected. In general, this process is similar to the reparametrization of the linear ARDL.
Thus, the multivariable dynamic ARDL model for the present study, is as follows 7 : where, t = 1, 2,…, T is for the time periods, I = 1, 2,…, N is for the countries, Y it is economic growth, and X i,t-j is the vector (kx1) of explanatory variables for the countries. For the explanatory variables, j begins from zero instead of one to allow the inclusion of the current year (time t) for explanatory variables in the model. Furthermore, μ i is the fixed effect and ε it is the error term iid ~ (0, σ 2 ). Then, the above model can be reparametrized as an ECM: 5 For more details, look at Baltagi (2013), Asteriou and Hall (2016) and Nymoen (2019). 6 Variable is unit-root non-stationary (Nymoen 2019, 316). 7 The model adapted based on Asteriou and Hall (2016).
where, θ i is the error correction parameter at the equilibrium and φ i is the long-run parameter if we use the MG estimator. When using the PMG estimator, because the long-run coefficient will be the same across countries, we have φ ' instead of φ i . An important feature of the ECM is its multicollinearity reduction by combining differenced data with lagged levels of variables (Nymoen, 2019). Then, based on Apergis and Payne (2014), Cavalcanti et al. (2011), andMehlum et al. (2006) the following econometric empirical model was developed: ln y j,t = a j +d j,t +β j1 ln OR j,t +β j2 ln Ov j,t +β j3 ln IQ j,t +μ j,t where, lny j,t is the logarithm of GDP per capita for countries j = 1,…, J and time period t = 1,…, T, a j represents country-specific fixed effects and d j,t denotes heterogeneous country-specific trends. OR j,t is the oil rent, Ov j,t is the currency overvaluation, and IQ j,t is quality of the institutions. Then, the ECM form of the above model is as follows: where ω j,t-1 = ln y j,t-1 +β j1 ln OR j,t +β j2 ln Ov j,t +β j3 ln IQ j,t Table 5 presents the results of the Granger causality test to determine whether there was any simultaneous effect between overvaluation and the GDP per capita. The null hypothesis states there is no causal relationship between the two variables. Because the above result shows only a causality relationship between overvaluation and GDP per capita, the ARDL model instead of a panel vector autoregressive (PVAR) model has been used in this study. The ARDL model works well when variables are cointegrated and the reduced form as an ECM established.

EMPIRICAL RESULTS
The following IPS unit-root test shows that the variables were stationary at first difference level, as such they are integrated one; I (1) ( Table 6). In this analysis, although the standard augmented Dickey-Fuller (ADF) test on the regression residuals with a null hypothesis that the variables are not cointegrated, and an alternative hypothesis that they are cointegrated can do the job, the Pedroni test for cointegration and the Im-Pesaran-Shin (IPS) unit-root test considers heterogeneity in different sections of the panel data and across the error terms.
To ensure the best number of lags in the model, Schwartz criteria are used, which indicates that ARDL (1,1) works best in this study ( Figure 4).
Finally, using the MG regression estimators, in three scenarios for all 26 countries from 1980 to 2017, countries with strong institution quality, and countries with weak institution quality.
In the first scenario, GDP per capita was the dependent variable and overvaluation was the independent variable. Several other controlling variables that are typical growth models literature variables, such as human capital and capital formation, were considered in the empirical regression model. As Table 7 shows, the long-run coefficient of overvaluation, θ, is significant and negative, as expected. This indicates a reverse relationship between overvaluation and economic growth. The short-run coefficient, β, is not significant in this empirical model. The error correction coefficient, γ, is the adjustment coefficient. Finally, the FH institution quality was entered as a fixed variable because it did not have enough variation for the period of this study.
We should be careful about how we interpret the FH coefficient. Freedom House ranks countries from 1 to 7, where 1 specifies the best and healthiest quality of institutions in a country and 7 displays the worst condition. Thus, as the FH variable decreases, the quality of institutions in a country improves. Consequently, the negative sign of the FH coefficient should be considered as a positive relationship between the quality of institutions and economic growth, as expected. Because FH is a fixed variable in the model, there is only a short-run coefficient available for it. Hence, for all countries, FH in the short run is positive and significant.
In the second and third scenarios, the independent variable is an interaction variable, oil rent multiplied by overvaluation. This interaction variable alongside several other controlling variables was considered in the empirical model. In the second scenario, these variables were estimated for countries with a strong quality of institutions and in the third scenario for countries with weak quality of institutions. Table 8 provides the results of these scenarios. As expected, the combination of oil rent and overvaluation created a variable that allows us to better measure the effect of oil rent and overvaluation on economic growth. Furthermore, the distinction among countries with weak and robust quality of institutions enables us to examine how institutions can influence this process. The significant negative long-run relationship between OilRent-Overvaluation and economic growth in countries with weak institutions and positive coefficient for countries with strong institutions while the coefficient for the FH institution variable is positive and significant for both groups of countries, support the theory that rent-seeking behavior and currency overvaluation can hurt the steady economic growth process in oil-rich countries with weak institutions in the long-run. Table 9 shows the short-run coefficients for selected countries. Although, as discussed short-run coefficients are not very important in macroeconomics study, we can see some different impacts of interaction and institution quality variables on these selected countries. The first country, Australia, is a country with strong institution quality and in this regression model, the coefficient of the FH variable is not significant. However, for the next country, Iran, with weak institution quality the FH coefficient is positive and significant. The results for other countries in Table 9 either are not significant or have mixed coefficient signs. As mentioned, the important is the long-run relationship between these variables. In general, we expect a more influential role of political and economic institutions, such as an independent judiciary system, protection of property law, and independent central banks, in developing countries than developed countries.

CONCLUSION
Given the results of this study, the present research puts forward three contentions. First, while some scholars argue that investigations into resource abundance (as measured by resource reserves) or resource dependence (as measured by the percentage of resource export to the GDP) should be considered, this study argues that resource dependence is a better explanatory variable for such research. If the purpose of the study is to determine the influence of resource rent, in this case, oil, on economic growth, then it is not clear how resource reserves, which are underground, and yet unused, can be an effective explanatory variable. As part of a nation's wealth, resource reserves can play a role of endowment on the nation's initial stage of growth for the conditional convergence purposes in the neoclassical growth model, but not necessarily on the nation's dynamic economic growth rate after the initial step. Moreover, OPEC members use their most updated oil reserves to bargain their production quota and market share. The biggest concern with choosing resource reserves over resource exports is an endogeneity problem in the model. While resource reserves are typically assumed to be exogenous in the model, the percentage of resource export to the GDP is endogenous in such a model (Brunnschweilera and Bulte, 2008). However, MG/ PMG and PVAR techniques can address this issue in their models (Antonakakis et al., 2017).
Second, the results of this study suggest that the welfare effect of natural resources could be very different from the growth effect. The results of this study indicate that oil-rich countries may have a good welfare policy by overvaluing their currency to make imported necessities cheaper for their citizens, but such a welfare policy may not be an excellent economic growth policy. Moreover, as the scholars (Boschini et al., 2007(Boschini et al., , 2012Durlauf et al., 2005;Isham et al., 2005;Ross, 2005) suggest, and this study used as its foundation, crude oil is by far the most significant natural resource that can undesirably affect economic growth, for countries with weak institutions.
Third, this study contends that in oil-rich developing countries with weak institutions, rent-seeking behavior and currency overvaluation slow economic growth. Indeed, this study confirms that oil rent has a substantial adverse effect on economic growth in the case of countries with weak institutions. For these countries, the interaction of such rent with overvaluation is even more negatively significant. As Arezki and Van der Ploeg (2007) suggested, higher natural resource revenue is an open invitation to rent-seeking activities.
Despite the fact that this study's main objective was to investigate the role of currency overvaluation and oil rent on the economic growth of oil-rich countries, it is also essential to consider sustainable growth in developing countries beyond the standard growth components such as the efficient allocation of scarce resources, capital accumulation, incomes, and output. Any scrutiny of economic growth in developing countries must consider the stage of economic development in those countries and development features, such as Kuznets's structural change and urbanization process (Acemoglu, 2009). Further, economic development is more than improvements in incomes and output and typically involves radical changes in institutions and social structures as well as customs and beliefs (Todaro and Smith, 2012).