Exogenous resource shocks and economic freedom.

O'Reilly, Colin ; Murphy, Ryan H.

INTRODUCTION

Early explanations of the relationship between natural resource abundance and slow economic growth--the resource curse--considered conventional macroeconomic channels (e.g., the Dutch Disease, see Sachs and Warner, 2001). As empirical evidence grew for endogenous growth models and the consequent importance of institutions (Acemoglu et ai, 2001, 2014), governance and institutions as channels for the resource curse have garnered more support (Mehlum et al, 2006). We contribute to this literature by testing if oil rent windfalls impact economic institutions.

Natural resource windfalls may influence relative prices and thus the incentives facing the economic and political agents shaping institutions, and in turn, economic performance (North, 1990). Regardless of whether institutions are the primary channel through which natural resource abundance lowers growth, the possibility has received substantial theoretical (Besley and Persson, 2010; Robinson et al, 2006) and empirical attention (Mehlum et al, 2006; Sarmidi et al, 2014; Mavrotas et al, 2011). Most of the literature studying the institutional channel of the natural resource curse has focused on the relationship between resource abundance and political institutions. For example, resource rents are associated with an increased ability for authoritarian leaders to stay in office (Omgba, 2009; Cuaresma et al, 2011), possibly through increased repression (De Mesquita and Smith, 2010). These dynamics shape political institutions; Arezki and Bruckner (2011) find that oil rents diminish political rights and Caselli and Tesei (2015) indicate that resource windfalls make authoritarian regimes more authoritarian. (1)

The focus on political institutions in explaining the pattern between resource wealth and economic performance is puzzling given that the association between economic performance and political institutions like democracy is tenuous (Acemoglu et al, 2008; Flachaire et al, 2014; Murphy and O'Reilly, 2017). De Haan and Sturm (2000) emphasize the difference between political and economic institutions; see "Empirical Approach" section for more on this distinction. In contrast to political institutions, the relationships between economic performance and those such as the rule of law, contract enforcement (Acemoglu and Johnson, 2005), property rights (Bhattacharyya, 2009), and other liberal economic institutions (De Haan et al, 2006; Faria and Montesinos, 2009; Hall and Lawson, 2014) are far more robust.

Two recent studies test if resource rent windfalls affect aspects of the quality of governance, including concepts more closely related to property rights and economic institutions. Using system GMM to address concerns of endogeneity, Busse and Groning (2013) find that natural resource exports increase corruption, but have no effect on measures of law and order or bureaucratic quality. Blanco et al (2015) find that oil rents are associated with reduced bureaucratic quality, but they find no effect on the rule of law. The difficulty in establishing a link between the resource curse and such essential institutions gives credence to a strand of the literature that challenges the association between resource abundance and economic growth (Alexeev and Conrad, 2011; Cotet and Tsui, 2013; Badeeb et al, 2017). Certain case studies of transition economies have questioned the resource curse, suggesting that for some countries, oil wealth has simply been a blessing (Pomfret, 2011; Kalyuzhnova and Patterson, 2016).

There is a small literature testing if rent windfalls are associated with a deterioration of economic institutions. Campbell and Snyder (2012) find a negative correlation between oil exports and economic freedom as measured by the Fraser Institute; however, without any attempt to address the endogeneity of resource rents, this cross-sectional relationship is hardly causal. Using panel data, Kennedy and Tiede (2013) find that oil rents cause the rule of law to deteriorate (as measured by the International Country Risk Guide and the Heritage Foundation), but that oil rents improve the related measures by the Fraser Institute of legal and property rights protections as well as the measure of regulation. However, they make some questionable econometric modeling decisions: they perhaps improperly apply an extreme bounds analysis, (2) use a random effects model rather than a fixed effects model, (3) and make no attempt otherwise to control for the endogeneity of oil rents.

In this paper we test whether exogenous oil rent windfalls affect economic institutions, as measured by the Fraser Institute's Economic Freedom of the World index and its subcomponents. Fraser's index has been shown to correlate with a host of economic and development outcomes. ("Empirical Approach" section provides more details on the index). Following the work of Lei and Michaels (2014), we use "giant" oil field discoveries as our measure of exogenous resource rent windfalls to address concerns of endogeneity. (4)

As expected, we find evidence that oil rent windfalls are associated with an increase in the size of government component of the Fraser index in the short run, primarily via an increase in government consumption and an increase in transfers and subsidies. However, these effects are only statistically significant over a 5-year interval, dissipating in size and significance over a longer interval. We find no evidence that giant oil field discoveries influence the overall index of economic institutions. At least for this set of resource discoveries, there is no resource curse for economic freedom.

EMPIRICAL APPROACH

Methodology

We estimate Eq. (1) to determine if oil discoveries influence economic institutions, [E.sub.it], as measured by either the full index of economic institutions or one of the subcomponents of the index. Subscript, i,

indicates countries, and t indicates 5-year periods.

[E.sub.it] = [beta][E.sub.it-1] + [gamma][D.sub.i,t-1] + [delta][X.sub.i,t-1] + [f.sub.i] + [tau] + [[epsilon].sub.i,t] (1)

The parameter of interest, [gamma], corresponds to the variable [D.sub.i,t-1] which is an indicator equal to 1 if a giant oil field was discovered during the period t-1 and 0 otherwise. Control variables that evolve over time are included in, [X.sub.i,t], and fixed effects are included, [f.sub.i] to capture country-specific effects. To capture any long run trends in institutions, we include time period dummy variables, [tau]. All independent variables are included with a lag since we are interested in how rent windfalls change institutions subject to conditions at the time of the windfall. Also motivating this modeling decision is the finding that oil production and oil exports begin to increase following an oil discovery after approximately 4 and 6 years, respectively (Lei and Michaels, 2014).

As a baseline we estimate Eq. (1), with both fixed effects and a lagged dependent variable, though acknowledging the presence of Nickell bias in these results (Nickell, 1981). To address the bias caused by the lagged dependent variable we employ a quasi-maximum likelihood (QML) estimator, which is asymptotically equivalent to, but has advantages over, the widely used system GMM estimation. In particular, the QML estimator may outperform system GMM when analyzing persistent series in finite samples (Moral-Benito, 2013). After examining these initial results, we look more closely at the components of our dependent variable and rerun the results using a longer interval.

The maximum likelihood estimator that we employ here was first suggested by Hsiao et al. (2002) to overcome the problem of "incidental parameters" in maximum likelihood estimation of fixed effects models. The procedure obtains consistent parameter estimates by first differencing the series and then maximizing a transformed likelihood function. Hsiao et al. (2002) find that this estimator outperforms system GMM in terms of bias and root-mean-square error. Specifically we implement this quasi-maximum likelihood (QML) estimator via the ST ATA command developed by (Kripfganz, 2016).

Data selection

The aforementioned study (De Haan and Sturm, 2000) provides a useful discussion of the difference between political and economic institutions, a distinction later emphasized by Flachaire et al. (2014). Political institutions include freedom to participate in the political decision process, for instance fair and competitive elections, whereas economic institutions include the rules that govern trade, institutional transaction costs, and the protection of property rights. It is clearly possible that a country can have institutions consistent with political freedom but inconsistent with economic freedom, and vice versa (Gwartney and Lawson, 2003). Making distinctions between these two dimensions of institutions does not preclude the likelihood that political and economic institutions are intertwined (Sobel and Coyne, 2011). Our data on economic freedom originate in the Economic Freedom of the World (EFW) index (Gwartney et al, 2016) published by the Fraser Institute. Its components include measures ("Areas") of the Size of Government (Area 1), the Legal System and Property Rights (Area 2), Sound Money (Area 3), Freedom to Trade Internationally (Area 4), and Regulation (Area 5).

The report has been used extensively as an explanatory variable (Hall and Lawson, 2014); greater economic freedom has been shown to be robustly related to economic growth (De Haan et al., 2006). As in the broader institutions and growth literature, the issue of endogeneity arises when testing the relationship between economic freedom and economic growth (Hanson, 2003), though the intractability of the issue is disputed (Heckelman, 2005). In a survey of the literature De Haan et al. (2006) indicate that studies "generally find support for a positive relationship between changes in EF and growth." More recent studies have demonstrated that the relationship between economic freedom and economic growth is robust to a series of methodologies, including Granger causality (Justesen, 2008), two stage least squares (Faria and Montesinos, 2009), and generalized method of moments (Kacprzyk, 2016).

Certainly other measures of economic institutions exist including: the International Country Risk Guide measures, the Corruption Perceptions Index, the Economic Freedom Index by the Heritage Foundation, among others. The influence of oil rents on the corruption, (5) the law and order, and the bureaucratic quality dimensions of the International Country Risk Guide have been investigated in the aforementioned studies (Blanco et al., 2015; Arezki and Bruckner, 2011; Busse and Groning, 2013). Ram (2014) emphases that there can exist substantial differences between the Heritage Foundation index and the Fraser Institute index, despite a high degree of correlation between these measures as demonstrated by De Haan and Sturm (2000). As illustrated by Murphy (2016a), the correlation between the indices of economic freedom is actually stronger than the correlation between two common indices of political institutions. Our choice of the Fraser index reflects that its methodology is more transparent and less subjective, as found by Cummings (2000: 31-33).

The Economic Freedom of the World index includes, with one-fifth weight, measures of the size of government. Many countries in Western Europe and Northern Europe have achieved very high scores in EFW and live lives relatively free of undue regulations but with an effective legal system, while maintaining very large welfare states. In some ways, this is almost inevitable due to the close connection between growing incomes and growing government spending per capita (i.e., "Wagner's Law"). Finland, Denmark, and the Netherlands rank 20th, 21st, and 25th, respectively, in EFW, comfortably within the top quartile, despite very high levels of government spending.

However, it is useful to distinguish countries that have a high degree of liberalization on every margin except the size of their governments from those which have liberalization in conjunction with small government sizes. Excluding city-states, the highest ranking countries in the world in EFW are New Zealand, Switzerland, Canada, Georgia, and Ireland. While these countries expend far more resources than would be necessary to maintain a night-watchman state, they have smaller governments than do Finland, Denmark, or the Netherlands.

Ultimately, which set of institutions constitutes "economic freedom" is an issue of semantics, and "economic freedom" need not necessarily perfectly correspond to better outcomes in every possible context. One can differentiate "market liberal" economic institutions that EFW measures, with perhaps "social liberal" economic institutions that would exclude government spending (Huskinson and Lawson, 2014). The goal of this exercise, however, is to test the hypothesis that there is a resource curse for economic institutions, as defined by economic freedom, in turn defined by market liberal institutions. It should be noted that the inclusion of a limited state sector as an important aspect of defining economic freedom has its roots in Classical Liberalism (Hayek, 1960; Friedman, 1962; Rothbard, 1978).

A growing literature has sought to explain what causes certain countries to have more economic freedom than others. Such findings include negative effects from war (O'Reilly and Powell, 2015), neutral or positive effects from immigration (Clark et al, 2015), mixed but likely negative effects from membership in intergovernmental organizations (Murphy, 2016b), negative but small effects from inequality (Murphy, 2015), mixed effects from access to the internet (Sheehan and Young, 2015), negative effects from foreign aid (Young and Sheehan, 2014), and positive effects from the ability of citizens to "vote with their feet" and emigrate (Hall, 2016).

Most recent cross-country studies of the resource curse have used system GMM or other instrumental variable methods because oil rents are potentially endogenous to political and economic institutions. To address the problem of endogeneity, we follow Lei and Michaels (2014) who use a data set of giant oil field discoveries from Horn (2004) as sources of rent windfalls that are exogenous to violent conflict. (6) Lei and Michaels argue that "giant" oil field discoveries, defined as those containing at least 500 million barrels of recoverable oil, are plausibly exogenous natural resource windfalls. Giant oil field discoveries are rare and random, so that "countries have little control over the timing of such finds" (Lei and Michaels, 2014). An advantage of the quasi-natural experimental identification strategy is that our results do not rely on assumptions about initial conditions or suffer from the weak instrument problem.

To assuage concerns that oil discoveries could be more likely in countries with more or less economic freedom, we test for any contemporaneous correlation between the degree of economic freedom and the discovery of giant oil fields (see "Appendix"). Oil discoveries were most common in the 1960s and 1970s, and economic freedom has increased on average since the 1970s; to avoid identifying a spurious correlation time specific effects must be included. If no period fixed effects are included, there is a significant association between economic freedom and oil discoveries. However, once time specific effects are controlled for, we find no significant relationship between the economic freedom of a county and the likelihood of discovering a "giant" oil field.

To be included in the dataset an oil field must contain at least 500 million barrels of "estimated ultimate recoverable reserves" which is an estimate of the total amount of oil or natural gas that will ever be recovered from the field. Though this is a subjective estimate prone to measurement error, these discoveries clearly map into rent. A discovery is linked to increasing oil production by 35-50 percentage points, increasing oil exports by 20-50 percent and increasing the likelihood of violent conflict within the first decade of discovery (Lei and Michaels, 2014).

Since oil revenue per person differs based on the population of the country, an oil discovery may have a different effect on a small country than on a large country. However, if oil discoveries are viewed as rent windfalls that tend to be contested by a narrow subset of society then the absolute size of the rent is the relevant measure. A windfall of 500 million barrels of oil is a large rent windfall regardless of the size of the country. By using such large rent windfalls our measure captures the "overabundance" of production that induces rent seeking and the resource curse (Oskenbayev et al, 2013). We follow the parsimony of Lei and Michaels (2014) by electing to use an indicator variable for oil discoveries rather than adjusting for country size. (7) Our controls are standard country level panel method controls. Our data on real logged economic output are from the Penn World Tables 9.0 (Feenstra et al, 2015). We also use the polity2 variable from the Polity IV data set, which measures the quality of a country's democracy on a strictly bounded [-10, 10] scale of integers (Marshall et al, 2014). Finally, our measure of human capital is from the Penn World Tables. This variable includes data on years of schooling from Barro and Lee (2013) and from Cohen and Leker (2014). For all variables, descriptive statistics can be found in Table 1 and a correlation matrix of these variables is found in Table 2. However, these control variables are used to reduce bias and improve precision of the coefficient on the oil discovery; the oil discovery itself is already assumed to be exogenous.

RESULTS AND DISCUSSION

Table 3 presents our baseline results. Regression (1) provides complete results for the effect of oil discoveries on economic freedom and its components. We find no statistically significant effect on overall economic freedom, with a negative, economically small point estimate. An oilfield discovery corresponds to a hundredth of a standard deviation decrease in EFW. (8) Regressions (2)-(6) provide results for its effects on each of the five areas of economic freedom in the EFW index. In Regression (2), we find clear, statistically significant, negative results, implying that an oilfield discovery leads to a larger government (i.e., Area 1) 5 years later. While none of the other areas approaches a meaningful coefficient, their net effect apparently balances out the negative effects on the size of government. Though there may be evidence that the discovery of an oil field increases the size of government, in the aggregate, negative effects are insufficient either to reject the null hypothesis or for the magnitude of the point estimate of effect to be large enough to otherwise be meaningful.

Table 4 employs the quasi-maximum likelihood technique, and thus its estimates are unaffected by Nickell bias. However, these results yield qualitatively identical and quantitatively very similar results. The overall effect on economic freedom (Regression 7) remains negative, statistically insignificant, and small in magnitude. The point estimate of an oilfield discovery now corresponds to a 0.075 standard deviation decrease in EFW. The effect on the size of government (Regression 8) is modestly reduced in its absolute value (from -0.351 to -0.327) but is otherwise the same. All other null results are similar to those found in Table 4 and are similarly uninteresting.

Table 5 looks more closely at three of the components of Area 1, size of government, first using fixed effects and the lagged dependent variable, then using QML. The components of Area 1, which are rated on a 0-10 scale and are not raw data, are government consumption as a share of total consumption (1A), transfers and subsidies as a share of GDP (IB), government enterprises and investment as a share of investment (1C), and the top marginal tax rate (ID). We ignore the scores for the top marginal tax rate. Lei and Michaels (2014) indicate that the positive relationship between exogenous discoveries and government consumption per capita (not share) in their data is not robust to the inclusion of control variables; our examination of this relationship differs in two respects. First, the measures constructed in the EFW index are more concerned with the size of government relative to the economy as an institution, rather than a raw quantity of spending, and the differences in definitions reflect that. Secondly, transfers and subsidies may importantly reflect the role in how the discovery of these oilfields changes government behavior, and Lei and Michaels do not include this (perhaps because it is not directly reflected in national accounts due to conventions); EFW is one of the few data sources which provides truly comprehensive data on each dimensions of government spending.

Regressions 13-18 break down each of the three relevant EFW sub-measures of size of government. Regressions (13) and (16) use government consumption as the dependent variable, Regressions (14) and (17) use transfers and subsidies, and Regressions (15) and (18) use government investment. The effects on government consumption and transfers and subsidies are negative (more spending) and statistically significant. The effect on government investment is negative and economically large, but statistically insignificant due to large standard errors. This is true for both methodologies, which each find very similar results.

Next, we substitute 10-year intervals for 5-year intervals. Effects over 5-year intervals may best be thought of as policies as opposed to institutions, as generally institutions are thought to be slow moving and only rarely change drastically. We look at the five variables the previous analysis found to be of the most interest: the entire EFW index, the total measure of size of government, and the three relevant components of size of government. These regressions (QML only) are found in Table 6. In none of these specifications are results significant for our variable of interest. The size of the coefficient of the effect of a discovery on government spending is now very small (-0.082 instead of -0.327), suggesting the effect actually dissipates and the lack of statistical significance is not simply the result of larger standard errors.

Taken together, we see the evidence overall showing an increase in the relative size of the public sector following an exogenous oil discovery, but this is a short run increase in government size. This can be interpreted as a temporary decrease in economic freedom from the "market liberal" conception of economic freedom or as an increase in economic freedom (or neutral) from a "social liberal" conception. Alternatively, within the framework of Solow convergence, the increase in wealth could build additional capital that will yield a short run increase in economic growth. Regardless, the discovery does not reflect a negative effect on the underlying institutional constraints on the public sector. Ultimately the effect on the one area of economic freedom is not large enough to be distinguishable in the aggregate index. We take all this to suggest that increases in public outlays are reasonably expected following one of these discoveries, but a general effect on economic freedom cannot be discerned.

Some potential concerns regarding our interpretation include the serial correlation of oil discoveries, as well as the possibility that countries with existing oil rents may differ in their ability to manage new discoveries of oil. The later concern may be overstated; countries with existing oil discoveries appear to be unable to manage oil discoveries well enough to prevent violent conflict over new oil discoveries (Lei and Michaels, 2014). Further, all of the results presented thus far regarding the sensitivity of economic institutions to oil discoveries are robust to the inclusion of an indicator variable for past oil discoveries. (9) Should these results be interpreted as a general test of the resource curse? Doing so offers a severe trade-off. To the extent that the argument of Lei and Michaels (2014)--that these oilfields should be taken seriously as an exogenous discovery--holds, our analysis offers a stronger result than most previous studies from the standpoint of pure econometrics. However, one could argue that these results are not generalizable, and other resources, whether blood diamonds or sheep's milk, are more susceptible to curse their country with diminished institutional quality than is a giant oilfield. Still, in terms of the raw size of the rents, large oilfields undeniably reflect a significant windfall to their discoverer, large enough to increase the likelihood of violent conflict (Lei and Michaels, 2014). Thus, these results strongly draw into question whether a resource curse through the channel of economic institutions--or one large enough to be statistically distinguishable from zero --exists, even if certain caveats must be attached to this statement.

These results have important policy implications given that political and economic institutions have been shown to reduce the detrimental effects of the resource curse (Mehlum et ai, 2006). Studies have found that the negative effects of the resource curse can be mitigated by institutions that "promote accountability and state competence" (Robinson et al, 2006), by democratic institutions (Bhattacharyya and Hodler, 2010), and by majoritarian electoral systems (Andersen and Aslaksen, 2008). Most relevant to this study, economic institutions as measured by the Fraser Institutes' index reduce the negative growth effects of the resource curse (Farhadi et al., 2015) and reduce the risk of oil induced separatist conflicts (O'Reilly and Murphy, 2016).

A variety of political and economic institutions can mitigate aspects of the resource curse, but political institutions in particular appear to be vulnerable to the resource curse (Ross, 2015). To the extent that our null result indicates that economic institutions are not influenced by oil rent windfalls, economic institutions may be a more robust bulwark vis-a-vis the natural resource curse than are political institutions.

CONCLUSION

We find that, for one particular class of resource shocks, economic institutions as measured by the Economic Freedom of the World index are unaffected. Pursuant to this analysis, we utilize data on "giant" oil field discoveries that have been argued to be exogenous due to their size. To the extent that giant oil field discoveries are exogenous, the present study is the first well-identified model of the relationship between oil resources and economic freedom. There is evidence that these discoveries cause government spending to increase in the short run, but overall effects on economic institutions are not present statistically. Moreover, all effects dissipate over a 10-year interval, suggesting that any observed changes reflect a short run effect on policy, with no harm visited upon the underlying institutions.

Political institutions can mitigate aspects of the resource curse, but the presence of resource rents has been shown to be detrimental to those same institutions. Our result, that economic institutions are not subject to the same risk of deterioration, indicates that free economic institutions may be a more robust safeguard against the natural resource curse. This finding, along with studies that show that economic institutions mitigate the risk of conflicts over resource rents, offers channels through which economic institutions relax the negative relationship between economic growth and resource rents. Perhaps it is not surprising that effective market institutions allocate abundant resource rent productively, while in some circumstances, political institutions are influenced by the very resources that they are designed to allocate.

REFERENCES

Acemoglu, D, Gallego, FA and Robinson, JA. 2014: Institutions, human capital, and development. Annual Review of Economics 6: 875-912. doi:10.1146/annurev-economics-080213-041119.

Acemoglu, D and Johnson, S. 2005: Unbundling institutions. Journal of Political Economy 113: 949-995.

Acemoglu, D, Johnson, S and Robinson, JA. 2001: The colonial origins of comparative development: an empirical investigation. American Economic Review 91: 1369-1401. doi:10. 1257/aer.91.5.1369.

Acemoglu, D, Johnson, S, Robinson, JA and Yared, P. 2008: Income and democracy. American Economic Review 98: 808-842. doi:10.1257/aer.98.3.808.

Alexeev, M and Conrad, R. 2011: The natural resource curse and economic transition. Economic Systems 35: 445-461. doi:10.1016/j.ecosys.2010.10.003.

Andersen, JJ and Aslaksen, S. 2008: Constitutions and the resource curse. Journal of Development Economics 87: 227-246. doi:10.1016/j.jdeveco.2007.12.005.

Arezki, R and Bruckner, M. 2011: Oil rents, corruption, and state stability: evidence from panel data regressions. European Economic Review 55: 955-963. doi:10.1016/j.euroecorev.2011.03. 004.

Badeeb, RA, Lean, HH and Clark, J. 2017: The evolution of the natural resource curse thesis: a critical literature survey. Resources Policy 51: 123-134. doi:10.1016/j.resourpol.2016.10.015.

Barro, RJ and Lee, JW. 2013: A new data set of educational attainment in the world, 1950-2010. Journal of Development Economics 104: 184-198. doi:10.1016/j.jdeveco.2012.10.001.

Besley, T and Persson, T. 2010: State capacity, conflict, and development. Econometrica 78: 1-34. doi:10.3982/ECTA8073.

Bhattacharyya, S. 2009: Unbundled institutions, human capital and growth. Journal of Comparative Economics 37: 106-120. doi:10.1016/j.jce.2008.08.001.

Bhattacharyya, S and Hodler, R. 2010: Natural resources, democracy and corruption. European Economic Review 54: 608-621. doi:10.1016/j.euroecorev.2009.10.004.

Blanco, LR, Nugent, JB and O'Connor, KJ. 2015: Oil curse and institutional changes: which institutions are most vulnerable to the curse and under what circumstances? Contemporary Economic Policy 33: 229-249. doi:10.1111/coep.12077.

Busse, M and Griming, S. 2013: The resource curse revisited: governance and natural resources. Public Choice 154: 1-20. doi: 10.1007/sl 1127-011 -9804-0.

Campbell, ND and Snyder, TJ. 2012: Economic growth, economic freedom, and the resource curse. Journal of Private Enterprise 28: 23-46.

Caselli, F and Tesei, A. 2015: Resource windfalls, political regimes, and political stability. Review of Economics and Statistics 98: 573-590. doi:10.1162/REST_a_00538.

Clark, JR, Lawson, R, Nowrasteh, A, Powell, B and Murphy, R. 2015: Does immigration impact institutions? Public Choice 163: 321-335. doi: 10.1007/s 11127-015-02S4-y.

Cohen, D and Leker, L. 2014: Health and education: another look with the proper data (CEPR Discussion Paper No. 9940). C.E.P.R. Discussion Papers.

Cotet, AM and Tsui, KK. 2013: Oil and conflict: what does the cross country evidence really show? American Economic Journal: Macroeconomics 5: 49-80. http://www.aeaweb.org/aej-macro/.

Cuaresma, JC, Oberhofer, H and Raschky, PA. 2011: Oil and the duration of dictatorships. Public Choice 148: 505-530. doi: 10.1007/sl 1127-010-9671-0.

Cummings, JT. 2000: Economic freedom indices: their use as tools for monitoring and evaluation. SCS Working Paper 00/01.

De Haan, J and Sturm, J-E. 2000: On the relationship between economic freedom and economic growth. European Journal of Political Economy 16:215-241. doi:10.1016/S0176-2680(99)00065-8.

De Haan, J, Lundstrom, S and Sturm, J-E. 2006: Market-oriented institutions and policies and economic growth: a critical survey. Journal of Economic Surveys 20: 157-191. doi: 10.1111/j. 0950-0804.2006.00278.x.

De Mesquita, BB and Smith, A. 2010: Leader survival, revolutions, and the nature of government finance. American Journal of Political Science 54:936-950. doi:10.111 l/j. 1540-5907.2010.00463.X.

Farhadi, M, Islam, MR and Moslehi, S. 2015: Economic freedom and productivity growth in resource-rich economies. World Development 72: 109-126. doi:10.1016/j.worlddev.2015.02.014.

Faria, HJ and Montesinos, HM. 2009: Does economic freedom cause prosperity? An IV Approach. Public Choice 141: 103-127. http://link.springer.com/journal/volumesAndlssues/11127.

Feenstra, RC, Inklaar, R and Timmer, MP. 2015: The next generation of the penn world table. American Economic Review 105: 3150-3182. doi:10.1257/aer.20130954.

Flachaire, E, Garcia-Penalosa, C and Konte, M. 2014: Political versus economic institutions in the growth process. Journal of Comparative Economics 42(1): 212-29. doi:10.1016/j.jce.2013.05.001.

Friedman, M. 1962: Capitalism and freedom. Chicago, IL: University of Chicago Press.

Gwartney, J and Lawson, R. 2003: The concept and measurement of economic freedom. European Journal of Political Economy, Economic Freedom 19:405-430. doi:10.1016/S0176-2680(03)00007-7.

Gwartney, J, Lawson, R and Hall, J. 2016: Economic freedom of the world: 2016 annual report. Vancouver, BC: Fraser Institute.

Hall, JC. 2016: Institutional convergence: exit or voice? Journal of Economics and Finance 40: 829-840. doi: 10.1007/sl2197-015-9345-3.

Hall, JC and Lawson, RA. 2014: Economic freedom of the world: an accounting of the literature. Contemporary Economic Policy 32: 1-19. doi:10.1111/coep.12010.

Hanson, JR. 2003: Proxies in the new political economy: caveat emptor. Economic Inquiry 41: 639-646. doi:10.1093/ei/cbg033.

Hayek, FA. 1960: The constitution of liberty. Chicago, IL: University of Chicago Press.

Heckelman, JC. 2005: Proxies for economic freedom: a critique of the hanson critique. Southern Economic Journal 72: 492-501.

Hsiao, C, Pesaran, MH and Tahmiscioglu, KA. 2002: Maximum likelihood estimation of fixed effects dynamic panel data models covering short time periods. Journal of Econometrics 109: 107-150. doi:10.1016/S0304-4076(01)00143-9.

Horn, MK. 2004: Giani fields 1868-2004 (CD-ROM). Revision to 2003. VersionAAPG/Datapages, Houston (Version 1.2, 2004).

Huskinson, T and Lawson, R. 2014: Clusters of economic freedom. Applied Economics Letters 21: 1070-1074. doi: 10.1080/13504851.2014.909565.

Justesen, MK. 2008: The effect of economic freedom on growth revisited: new evidence on causality from a panel of countries 1970-1999. European Journal of Political Economy 24: 642-660. doi: 10.1016/j.ejpoleco.2008.06.003.

Kacprzyk, A. 2016: Economic freedom: growth nexus in European Union countries. Applied Economics Letters 23: 494-497. doi:10.1080/13504851.2015.1083076.

Kalyuzhnova, Y and Patterson, K. 2016: Kazakhstan: long-term economic growth and the role of the oil sector. Comparative Economic Studies 58: 93-118. doi:10.1057/ces.2015.31.

Kennedy, R and Tiede, L. 2013: Economic Development assumptions and the elusive curse of oil. International Studies Quarterly 57: 760-771. doi:10.1111/isqu,12065.

Kripfganz, S. 2016: Quasi-maximum likelihood estimation of linear dynamic short-T panel-data models. The Stata Journal 16(4): 1013-1038.

Leamer, EE. 1983: Let's take the con out of econometrics. The American Economic Review 73: 31-43.

Lei, Y-H and Michaels, G. 2014: Do giant oilfield discoveries fuel internal armed conflicts? Journal of Development Economics, Land and Property Rights 110: 139-157. doi:10.1016/j.jdeveco. 2014.06.003.

Marshall, M, Gurr, T and Jaggers, K. 2014: Polity IV project: political regime characteristics and transitions 1800-2013. Center for Systemic Peace.

Mavrotas, G, Murshed, SM and Torres, S. 2011: Natural resource dependence and economic performance in the 1970-2000 period. Review of Development Economics 15: 124-138. doi:10. 1111/j. 1467-9361.2010.00597.x.

Mehlum, H, Moene, K and Torvik, R. 2006: Institutions and the resource curse. The Economic Journal 116: 1-20.

Moral-Benito, E. 2013: Likelihood-based estimation of dynamic panels with predetermined regressors. Journal of Business and Economic Statistics 31: 451-472. doi:10.1080/07350015. 2013.818003.

Murphy, RH. 2015: The impact of economic inequality on economic freedom. Cato Journal; Washington 35: 117-131.

Murphy, RH. 2016a: A comment on "measuring economic freedom: a comparison of two major sources". Journal of Private Enterprise 31: 69-75.

Murphy, RH. 2016b: Intergovernmental organisations and economic freedom: wise technocrats or black helicopters? Economic Affairs 36: 145-154. doi:10.1111/ecaf.12181.

Murphy, RH and O'Reilly, C. 2017: Applying panel vector autoregression to institutions, human capital, and output. Working paper. http://papers.ssrn.com/sol3/papers.cfm?abstract_id = 2572059.

Nickell, S. 1981: Biases in dynamic models with fixed effects. Econometrica 49: 1417-1426. doi:10. 2307/1911408.

North, DC. 1990: Institutions, institutional change and economic performance, 59262nd ed. Cambridge: Cambridge University Press.

Omgba, LD. 2009: On the duration of political power in africa the role of oil rents. Comparative Political Studies 42: 416-436. doi:10.1177/0010414008325575.

O'Reilly, C and Murphy, RH. 2016: Do institutions mitigate the risk of natural resource conflicts? Contemporary Economic Policy. doi:10.1111/coep.12207.

O'Reilly, C and Powell, B. 2015: War and the growth of government. European Journal of Political Economy 40(Part A): 31-41. doi:10.1016/j.ejpoleco.2015.10.001.

Oskenbayev, Y, Yilmaz, M and Abdulla, K. 2013: Resource concentration, institutional quality and the natural resource curse. Economic Systems 37: 254-270. doi:10.1016/j.ecosys.2012.11.001.

Pomfret, R. 2011: Exploiting energy and mineral resources in Central Asia, Azerbaijan and Mongolia. Comparative Economic Studies 53: 5-33. doi:10.1057/ces.2010.24.

Ram, R. 2014: Measuring economic freedom: a comparison of two major sources. Applied Economics Letters 21: 852-856. doi:10.1080/13504851.2014.894620.

Robinson, JA, Torvik, R and Verdier, T. 2006: Political foundations of the resource curse. Journal of Development Economics 79: 447-468. doi:10.1016/j.jdeveco.2006.01.008.

Ross, ML. 2015: What have we learned about the resource curse? Annual Review of Political Science 18: 239-259. doi:10.1146/annurev-polisci-052213-040359.

Rothbard, M. 1978: For a new liberty: the libertarian manifesto. Revised ed. San Francisco, CA: Fox, Wilkes.

Sachs, JD and Warner, AM. 2001: The curse of natural resources. European Economic Review, 15th Annual Congress of the European Economic Association 45: 827-838. doi:10.1016/S00142921(01)00125-8.

Sarmidi, T, Law, SH and Jafari, Y. 2014: Resource curse: new evidence on the role of institutions. International Economic Journal 28: 191-206. doi:l0.1080/10168737.2013.787110.

Sheehan, KM and Young, AT. 2015: It's a small world after all: internet access and institutional quality. Contemporary Economic Policy 33: 649-667. doi:10.1111/coep.12095.

Sobel, RS and Coyne, CJ. 2011: Cointegrating institutions: the time-series properties of country institutional measures. Journal of Law and Economics 54: 111-134.

Young, AT and Sheehan, KM. 2014: Foreign aid, institutional quality, and growth. European Journal of Political Economy 36: 195-208. doi:10.1016/j.ejpoleco.2014.08.003.

APPENDIX

See Table 7. Table 7: Determinants of oil discoveries (fixed effects) Method (13) (14) (15) OLS OLS OLS LHS Oil Oil Oil discoveries discoveries discoveries EFW -0.025 (b) -0.015 -0.004 (0.114) (0.018) (0.019) LnRGDPcap lag -- -- -0.001 (0.040) Polity IV lag -- -- 0.002 (0.003) Human capital lag -- -- -0.123 (0.114) Constant 0..303 (c) 0.294 (c) 0.497 (a) (0.071) (0.111) (0.273) Year fixed effects N Y Y Country fixed Y Y Y effects N 993 993 795 Within [R.sup.2] 0.007 0.018 0.022 Clustered standard errors were performed. (a) 90% confidence. (b) 95% confidence. (c) 99% confidence.

(1) See Ross (2015) for a complete literature review.

(2) Extreme bounds analysis, suggested by (Learner, 1983), runs series of regressions with different control variables to test the robustness or fragility of a statistical finding. Kennedy and Tiede (2013) conduct this analysis for 19 different dependent variables and leading to over 18,000 specifications. Searching among alternative dependent variables presents precisely the same statistical issues which extreme bounds analysis is intended to solve.

(3) Recent studies have used fixed effects models for this type of analysis to address concerns of omitted variable bias. Kennedy and Tiede (2013) present random effects estimates as their primary findings despite the result of the Hausman test, which indicates that a fixed effects model is preferred. Their point estimates corresponding to the subcomponents of the EFW index switch signs when estimated by a fixed effects model.

(4) A full description of the oil discoveries data and the exogenous nature of such discoveries can also be found in "Empirical Approach" section.

(5) In panel analysis the ICRG corruption measure is more commonly used than the Corruption Perceptions Index due to the limited time dimension of the Corruption Perceptions Index.

(6) The likelihood of an oil discovery is unaffected the timing of a violent conflict or a lull in violent conflict.

(7) Lei and Michaels note that the use of an indicator variable reduces measurement error that may occur due to revisions in the estimated size of the oil field.

(8) Even the cumulative long run effect calculated from the partial adjustment model does not amount to more than two hundredths of a standard deviation decrease in EFW. This long run effect is calculated by dividing the coefficient estimate corresponding to oil discoveries by one minus the coefficient on the lagged dependent variable.

(9) Results available upon request.

COLIN O'REILLY [1] & RYAN H. MURPHY [2]

[1] Creighton University, 2500 California Plaza, Omaha, NE 68102, USA. E-mail: colinworeiLly@gmail.com

[2] Southern Methodist University, Dallas, TX, USA. Table 1: Descriptive statistics Mean SD Min Max N Economic freedom index 6.10 1.32 2.00 9.11 993 Log of real GDP per capita 8.68 1.22 6.10 12.43 1271 Polity 2 1.44 7.39 -10.00 10.00 1215 Human capital (PWT) 2.12 0.71 1.01 3.70 1167 Giant oil field discoveries 0.13 0.33 0.00 1.00 1738 Table 2: Correlation matrix Economic Log real Polity 2 Human freedom GDP capital index per capita (PWT) Economic freedom index Correlation 1 P value N 993 Log of real GDP per capita Correlation 0.645 1 P value 0 N 960 1271 Polity 2 Correlation 0.4808 0.4243 1 P value 0 0 N 939 1160 1215 Human capital (PWT) Correlation 0.6495 0.7382 0.6177 1 P value 0 0 0 N 925 1167 1090 1167 Giant oil field discoveries Correlation -0.0778 0.1176 -0.097 -0.0079 P value 0.0142 0 0.0007 0.7888 N 993 1271 1215 1167 Giant oil field discoveries Economic freedom index Correlation P value N Log of real GDP per capita Correlation P value N Polity 2 Correlation P value N Human capital (PWT) Correlation P value N Giant oil field discoveries Correlation 1 P value N 1738 Table 3: OLS regressions, overall economic freedom and area scores, 5-year lags LHS (1) (2) (3) EFW Area 1 Area 2 LHS lag 0.536 (c) 0.332 (c) 0.251 (c) (0.039) (0.042) (0.045) Oil lag -0.012 -0.351 (c) 0.021 (0.102) (0.115) (0.136) LnRGDPcap Lag -0.205 (b) -0.327 (b) 0.172 (0.086) (0.139) (0.131) Polity IV lag 0.028 (c) 0.006 0.002 (0.008) (0.010) (0.012) Human capital lag 0.370 0.561 0.367 (0.239) (0.379) (0.423) Constant 3.174 (c) 5.225 (c) 0.646 (0.584) (1.106) (1.062) Year fixed effects Y Y Y Country fixed effects Y Y Y N 758 815 719 Within [R.sup.2] 0.688 0.425 0.353 LHS (4) (5) (6) Area 3 Area 4 Area 5 LHS lag 0.436 (c) 0.464 (c) 0.421 (c) (0.038) (0.035) (0.041) Oil lag 0.020 0.049 -0.072 (0.171) (0.226) (0.084) LnRGDPcap Lag 0.033 -0.517(c) -0.162 (a) (0.211) (0.161) (0.093) Polity IV lag 0.035 (a) 0.042 (c) 0.014 (0.019) (0.016) (0.008) Human capital lag 0.771 0.661 0.117 (0.588) (0.517) (0.213) Constant 1.271 6.028(c) 4.150 (c) (1.658) (1.157) (0.662) Year fixed effects Y Y Y Country fixed effects Y Y Y N 839 777 745 Within [R.sup.2] 0.413 0.566 0.597 Clustered standard errors were performed. (a) 90% confidence. (b) 95% confidence. (c) 99% confidence. Table 4: Quasi-maximum likelihood regressions, overall economic freedom and area scores, 5-year lags LHS (7) (8) (9) (10) EFW Area 1 Area 2 Area 3 LHS lag 0.773 (c) 0.520 (c) 0.447 (c) 1.022 (c) (0.060) (0.053) (0.067) (0.044) Oil lag -0.075 -0.327(c) -0.080 0.064 (0.099) (0.111) (0.138) (0.121) LnRGDPcap lag -0.249(c) -0.404e 0.047 -0.012 (0.078) (0.132) (0.135) (0.109) Polity IV lag 0.024 (c) 0.013 0.002 0.025(c) (0.007) (0.010) (0.009) (0.008) Human capital lag 0.418 0.649 (b) 0.428 0.161 (0.177) (0.313) (0.335) (0.225) Constant 2.687(c) 5.050(c) 1.653 -0.400 (0.867) (1.377) (1.361) (1.158) Year fixed effects Y Y Y Y N 675 703 640 714 LHS (11) (12) Area 4 Area 5 LHS lag 0.620(c) 0.711 (c) (0.051) (0.063) Oil lag -0.036 0.003 (0.227) (0.080) LnRGDPcap lag -0.574(c) -0.211(c) (0.180) (0.071) Polity IV lag 0.034(b) 0.013 (a) (0.015) (0.007) Human capital lag 0.860 (b) 0.299 (a) (0.431) (0.172) Constant 5.652(c) 3.122(c) (1.879) (0.790) Year fixed effects Y Y N 652 672 Clustered standard errors were performed. (a) 90% confidence. (b) 95% confidence. (c) 99% confidence. Table 5: Quasi-maximum likelihood regressions and OLS regressions, Area 1 components, 5-year lag Method (13) (14) (15) OLS OLS OLS LHS Govt Subsides and Govt enter, and consumption transfers investment LHS Lag 0.482c 0.342 (c) 0.317 (c) (0.036) (0.064) (0.032) Oil Lag -0.248 (b) -0.274 (c) -0.412 (0.114) (0.101) (0.296) LnRGDPcap Lag -0.826 (c) -0.188 -0.198 (0.179) (0.155) (0.255) Polity IV Lag -0.006 0.008 0.025 (0.011) (0.011) (0.023) Human Capital Lag 1.813 (c) -0.632 (b) -0.262 (0.464) (0.304) (0.965) Constant 6.917 (c) 7.824 (c) 4.626 (b) (1.305) (1.250) (1.775) Year fixed effects Y Y Y Country fixed effects Y Y Y N 821 656 795 Within [R.sup.2] 0.347 0.204 0.328 Method (16) (17) (18) QML QML QML LHS Govt Subsides and Govt enter, and consumption transfers investment LHS Lag 0.729 (c) 0.541 (c) 0.507 (c) (0.077) (0.083) (0.049) Oil Lag -0.302 (b) -0.273 (b) -0.376 (0.130) (0.106) (0.293) LnRGDPcap Lag -0.556 (c) -0.136 -0.316 (0.173) (0.145) (0.241) Polity IV Lag -0.005 0.015 (a) 0.028 (0.009) (0.009) (0.021) Human Capital Lag 1.709 (c) -0.157 -0.249 (0.368) (0.315) (0.921) Constant 2.295 4.749 (c) 6.128 (c) (1.676) (1.715) (2.465) Year fixed effects Y Y Y Country fixed effects N/A N/A N/A N 707 562 689 Within [R.sup.2] N/A N/A N/A Clustered standard errors were performed. (a) 90% confidence. (b) 95% confidence. (c) 99% confidence. Table 6: QML regressions, 10-year lags, EFW and components LHS (19) (20) (21) (22) EFW Area 1 Govt Subsidies and consumption transfers LHS lag 1.261 (c) 0.299 (b) 0.418 (b) 0.249 (a) (0.134) (0.147) (0.183) (0.150) Oil lag 0.125 -0.082 -0.024 0.046 -0.157 (0.283) (0.285) (0.148) LnRGDPcap -0.745 (c) -0.328 -0.653 -0.198 lag (0.167) (0.357) (0.397) (0.261) Polity IV 0.016 0.019 0.006 0.010 lag (0.012) (0.019) -0.023 (0.012) Human 0.040 -0.267 1.601 (a) -0.402 capital lag -0.55 (0.761) (0.944) -0.531 Constant 4.797 (c) 8.066 (b) 5.374 8.191 (c) (1.808) (3.252) (3.717) (2.397) Year fixed Y Y Y Y effects N 293 291 293 237 LHS (23) Govt enter, and investment LHS lag 0.187 (0.124) Oil lag -0.360 (0.637) LnRGDPcap 0.019 lag (0.666) Polity IV 0.081 lag (0.052) Human -1.440 capital lag (1.679) Constant 7.860 (6.055) Year fixed Y effects N 283 Clustered standard errors were performed. (a) 90% confidence. (b) 95% confidence. (c) 99% confidence.