Natural resources and fiscal performance: Does good governance matter?

Natural resources and fiscal performance: Does good governance matter?

Journal of Macroeconomics 37 (2013) 285–298 Contents lists available at SciVerse ScienceDirect Journal of Macroeconomics journal homepage: www.elsev...

391KB Sizes 4 Downloads 102 Views

Journal of Macroeconomics 37 (2013) 285–298

Contents lists available at SciVerse ScienceDirect

Journal of Macroeconomics journal homepage: www.elsevier.com/locate/jmacro

Natural resources and fiscal performance: Does good governance matter? Amany A. El Anshasy a,b,⇑,1, Marina-Selini Katsaiti a a b

Department of Economics and Finance, College of Business and Economics, United Arab Emirates University, Alain, United Arab Emirates Department of Public Finance, Faculty of Commerce, Alexandria University, Alexandria, Egypt

a r t i c l e

i n f o

Article history: Received 26 July 2011 Accepted 20 May 2013 Available online 13 June 2013 JEL classification: O43 E62 Q38 Keywords: Growth Institutions Fiscal policy Resource abundance

a b s t r a c t Weighing the current world affairs, there seems to be strong association between natural resources, corruption, and bad economic performance. We empirically investigate the interplay between different institutional qualities and fiscal policy, and their effect on resource-abundant economies’ growth. In particular, the study contributes to the existing literature by disentangling the indirect effect of institutions on growth through the ‘‘quality of fiscal performance’’ transmission channel. Using yearly panel data on 79 resource and non-resource countries for the period 1984–2008, we find that the quality of fiscal policy – and not the quantity (government size) – matters to growth in the group of resource-rich countries. We also find that not all types of socio-economic and political institutions impact growth in the same manner. Better governance, stronger democratic institutions, and transparent budgets improve fiscal performance, leading to higher growth rates. Democratic and budget institutions seem to be in effect only through the fiscal channel but not independently. Ó 2013 Elsevier Inc. All rights reserved.

1. Introduction While evidence on long-run fiscal policy growth effects is inconclusive, the recent view that differences in institutional and governance qualities across countries may explain part of the differences in growth performance is better established. Weak governance and institutional infrastructure may have a direct negative effect on growth by lowering productivity. In more shock-prone economies, such as commodity exporters, bad institutions can, in addition to the direct effect, have an indirect effect by undermining the economy’s ability to properly respond to external shocks (Rodrik, 1998). In particular, institutions affect the quality of policy outcomes. Although, this indirect effect can be very significant in resource-abundant economies, it remains under-investigated. This paper contributes to the existing literature by attempting to provide better understanding of the different channels through which fiscal policy may interact with governance qualities in affecting growth in resource-abundant economies. In particular, it presents new empirical evidence on the indirect effects of institutional qualities on growth through the fiscalmanagement transmission channel. The main argument we advance is that the ‘‘quality’’ of windfall management or fiscal performance matters more to growth in the group of resource-abundant economies than fiscal ‘‘quantities’’, represented by

⇑ Corresponding author. Tel.: +971 506603536. E-mail addresses: [email protected] (A.A. El Anshasy), [email protected] (M.-S. Katsaiti). Address: Department of Economics and Finance, Faculty of Business and Economics, United Arab Emirates University, PO Box 17555, Al-Ain, Abu Dhabi, United Arab Emirates. Tel.: +971 3 713 5267 (O); fax: +971 3 762 4384. 1

0164-0704/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jmacro.2013.05.006

286

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298

government size. The magnitude of this quality effect is in turn determined by the quality and strength of a country’s institutions. To this end, this paper empirically attempts to disentangle the direct effect of institutions on growth from that indirect effect through fiscal performance, using a yearly panel of 32 resource-abundant countries and 47 non-resource countries for the period 1984–2008. Weighing the current world affairs reveals that a holy trinity of corruption, lack of accountability, and abundance of natural resources not only may have contributed to the recent political turmoil in the Middle East, but also to the inequality and bad economic performance. It is well documented that resource-abundant economies, on average, have experienced poor growth performance compared to non-resource economies in the past few decades; a phenomenon known as the ’’resource curse’’. Many theories offer potential channels through which this ‘‘curse’’ may occur. One growing strand of research focuses on the institutional and the political incentives channels.2 Linking this institutional approach to fiscal performance, the International Budget Partnership initiative finds in its 2008 and 2010 reports that there is a strong association between oil resources and lack of budget transparency.3 They find oildependent economies to have very poor expenditure control systems. Even when some oil-producers, such as Saudi Arabia, appear to have sound fiscal policies their budgeting process is too opaque and they fail to release key budget information to the public. These economies also score very poorly in terms of linking their short-to-medium term government spending plans to a clear long-run economic plan. This implies that key government spending would likely fail to positively contribute to future growth. The 24 oil-dependent countries surveyed in 2010 had an average Open Budget Index score of just 26 out of 100. Countries rich on mineral resources, however, have had a better profile in terms of the strength and transparency of their budget institutions. No doubt, the interplay between institutions and policies could also exist through monetary, trade, or exchange rate policies. However, fiscal policy in resource-abundant economies is directly linked to the use and allocation of the resource rents. In addition, fiscal policy lies in the heart of the redistributional conflicts that usually arise in the wake of resource booms (Tornell and Lane, 1998, 1999). In such economies, particularly oil-exporting countries, government finance has often been plagued by a highly volatile revenue stream and a weak tax base. Procyclical fiscal policy, high public debt ratios, and (or) disproportionate increases in government spending in the wake of resource booms can thus be signaling weak budget institutions. To the extent that these fiscal policy failures are large, weak institutions and governance in resource-abundant economies can severely hinder growth. Using a battery of tests, we find that in the group of resource-rich economies the quality of fiscal policy – and not the quantity – matters to growth. Low corruption, better governance qualities, stronger democratic and budgetary institutions improve windfall management, leading to higher growth rates. This transmission channel is not warranted in resource-poor economies. The findings are robust to changing the measure of institutions and resource intensity. The main policy implication for resource-abundant economies is that strengthening governance and accountability is a pre-requisite to fully reaping the benefits of the abundant resources and escaping a ‘‘resource curse’’. The paper is structured as follows. We next discuss earlier literature on institutions and growth, particularly in the context of resource-abundant economies. In Section 3 we lay out our empirical methodology. Section 4 describes the data and their sources. The results are discussed in Section 5. Section 6 concludes.

2. Institutions, growth, and resource-abundance The literature on institutions as a deep determinant of growth has been growing for more than a decade. The findings of this literature suggest a positive relation between the quality of institutions and growth (Knack and Keefer, 1995; Acemoglu et al., 2001, 2002; Rodrik et al., 2002; Dollar and Kraay, 2003). Rodrik (1998) makes a direct link between greater exposure to external shocks, the quality of conflict resolution institutions, and growth. Large external shocks usually trigger distributional conflicts; when conflict management institutions are weak, the growth-depressing costs of terms-of-trade shocks are high. Since the seminal work of Sachs and Warner (1995), SW thereafter, a growing body of literature has tried to explain the ‘‘growth deficit’’ of resource-abundant economies. Recent research on the ‘‘resource curse’’ attempts to explore the institutional and political factors. SW dismissed the notion that abundant natural resources negatively affect the quality of institutions, and hence their findings imply that natural resources are not detrimental to growth through the institutional channel. Contrasting SW’s implication, Mehlum et al. (2006), MMT thereafter, argue that ’’institutions may be decisive for how natural resources affect economic growth even if resource abundance has no effect on institutions’’ p. 3. This view is consistent with the above-mentioned findings of Rodrik (1998). Indeed, MMT find that in countries with poor rule of law, natural resources impede growth because entrepreneurial resources shift away from production into unproductive activities. On the contrary, resources are a blessing in countries with ‘‘producer-friendly’’ institutions. Other studies find that there is negative causality that runs from natural resource abundance to institutional qualities; and through this channel to growth (Sala-i-Martin and Subramanian, 2003; Isham et al, 2005). 2 3

See: Wick and Bulte (2009) and Frankel (2010) for recent and comprehensive surveys. http://internationalbudget.org/what-we-do/open-budget-survey/.

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298

287

Collier and Goderis (2007), supporting MMT, find strong evidence that the growth effects of revenue booms work conditional on governance qualities. In the group of good governance countries, the resource rent has a positive effect on growth. In addition, they find that the curse only works through the interaction of institutions with government consumption. Increasing government consumption reduces growth but this negative effect is exacerbated by poor quality institutions. This result lends empirical support to the insights gained from the theoretical work of Robinson et al. (2006) who directly model the effect of institutions on public policies in resource-abundant economies. The model formally shows that ‘‘the incentives politicians face when they confront resource booms map into different policy choices depending on the quality of institutions’’ p. 465. All the above studies provide evidence that institutions matter to resource-abundant countries’ growth but do not address the fiscal channel. On the other end, some studies investigated fiscal policy as a potentially important transition mechanism of oil price shocks in resource-abundant economies; but one shortcoming of these studies is that they do not control for the possible effect of institutions on growth, or on fiscal policy itself (see: El Anshasy, 2012; Fasano and Wang, 2001; Husain et al., 2008; Pieschacon, 2008). Recent economic literature documents a link between fiscal policy outcomes and the quality of institutions. Alesina and Perotti (1999) and Persson and Tabellini (2004) argue that fiscal policy outcomes, particularly the budget balance, are determined by budgetary institutions and constitutional rules. Similarly, Kontopoulos and Perotti (1999) provide evidence that high expenditure and budget deficits usually arise when the fiscal authority is fragmented. Alesina et al. (2008) find that corruption reinforces the procyclicality of fiscal policy, conditional on political institutions. Frankel et al. (2011) show that many developing countries have started experiencing countercyclical policies for the first time during the past decade because of better institutional qualities. In countries that witness trade booms, Tornell and Lane (1998, 1999) show that institutions and the underlying power structure can explain the excessive spending of government revenue windfalls. This is known as the voracity effect; when institutions are fractionalized, powerful groups compete to appropriate a greater share of national wealth amid a temporary windfall gain by exerting pressure on the fiscal authorities to increase public spending that directly benefits their constituencies. Indeed, Stein et al. (1999) find evidence that Latin American countries with certain political traits are more likely to pursue stronger procyclical expenditure policies. They also find that more transparent and hierarchical budgetary procedures can lead to lower deficits and debt. El Anshasy and Bradley (2012) find that, in oil-exporting countries in the longrun, less civil liberty is associated with larger governments. Institutions can also impact government spending composition. Some studies find that corruption can actually favor growth by shifting expenditure composition towards more productive types, despite having a negative effect on productivity (Ghosh and Gregoriou, 2010). On the other hand, Mauro (1998) shows that higher corruption levels crowd out public spending on education, negatively impacting growth. In summing up, one can draw three important conclusions from the preceding discussion which motivated our study. First, institutions do matter to growth, especially in resource-abundant countries. Second, the indirect links between institutions and growth through policy performance remain under-investigated in the existing literature. Third, the quality of fiscal management has not been addressed as a potential channel through which poor governance and institutional qualities may impact growth. 3. Empirical methodology The empirical investigation aims to decompose the effects of institutional and governance qualities into direct and indirect effects through fiscal performance. To this end, we combine a growth model with a fiscal policy equation in a treatmenteffects-like model. This framework is motivated by the objective of capturing a quality effect of fiscal management as opposed to the traditional quantity effect of fiscal aggregates (here government size). Resource economies with weak and fractionalized institutions usually suffer from windfall mismanagement. Strong fiscal procyclicality may thus be signaling such mismanagement.4 In such representation, bad governance is expected to increase the procyclicality of government spending which hinders growth. In addition, we investigate the validity of the ‘‘resource curse’’ hypothesis after controlling for institutions and the fiscal performance channel. As a standard procedure, the growth estimation is augmented with various measures of resource intensity. Therefore, our strategy can be summarized in adding to a standard growth regression, institutional variables, the indirect effect of institutions on growth through the fiscal transmission channel, measures for fiscal performance and resource abundance. 3.1. Fiscal performance characterization: procyclicality Both the Keynesian and the neoclassical models imply that procyclical fiscal policy is suboptimal; what Kaminsky et al. (2005) have dubbed the ‘‘when it rains, it pours’’ phenomenon. For that reason, fiscal procyclicality constitutes a puzzle in 4 It should be noted however, as one referee pointed out, that other signs of mismanagement could be evident in shifting resources from productive uses of public capital to less productive ones. This is better addressed by looking at the composition of public spending not just the pro-cyclicality of spending.

288

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298

developing economies. In particular, in a Keynesian world of sticky prices, the sluggish adjustment of the economy in booms and busts leads to a procyclical output. In such a model, an increase in government consumption during a boom boosts aggregate demand and expands output and vice versa, reinforcing the cycle; hence a counteryclical fiscal policy would be deemed optimal. In a different vein, the neoclassical theory leads to the same conclusion; volatility results in welfare loses, and therefore higher government spending during booms and lower spending during busts enhances such volatility and is considered suboptimal (Ilzetski and Carlos, 2008). In addition, according to Barro’s (1979) tax smoothing hypothesis, an optimal fiscal policy implies constant tax rates over the business cycle. So, good fiscal performance would imply a countercyclical government spending and constant marginal tax rates. Since data on tax rates across countries are not readily available, we take higher procyclicality to signal weaker fiscal performance. In order to address the effect on long-run growth, we include in the outcome (growth) equation a variable which measures the degree of procyclicality, characterizing the quality of fiscal policy management. In measuring such a parameter, Kaminsky et al. (2005) emphasize the importance of using a policy instrument rather than a policy outcome. The intuition is that outcomes do not accurately reveal policy ‘‘choices’’ because they lie outside the policymaker’s direct control and thus are too noisy. They hence suggest the growth in government spending as the most appropriate fiscal instrument. Similarly, Ilzetski and Carlos (2008) argue that the two appropriate fiscal policy instruments to measure procyclicality are government spending and tax rates. On the contrary, tax revenues may not be an appropriate policy instrument because revenues tend to be procyclical by nature. So, even when government spending and tax rates are independent from the business cycle, tax revenues would naturally increase during booms and shrink during recessions, falsely detecting a procyclical fiscal policy. In addition, in the context of resource-rich economies, due to large resource price fluctuations and because most government revenues are driven by the resource sector, the government’s overall balance would also be a misleading measure. Even the non-resource budget balance, which excludes the natural resource revenues, would not yet be appropriate. Kaminsky et al. (2005) argue that it is impossible to determine the procyclicality of fiscal policy using the budget balance (as a percentage of GDP). So, following these studies, we limit our investigation to government consumption spending. We estimate a standard equation of government spending growth on output gap, augmented by measures of institutional qualities, to capture fiscal procyclicality as follows (see: Alesina et al., 2008).5

DGit ¼ c1 þ c2i OUTPUTGAPit þ c3i Iit þ c4i Iit  OUTPUTGAPit þ ki DGit1 þ ai þ eit

ð1Þ

where Git is the government consumption (expressed as GDP share). OUTPUTGAPit is a measure of the business cycle, Iit accounts for institutions, Iit ⁄ OUTPUTGAPit captures the interaction between the business cycle and the quality of institutions, where we expect c3 < 0 and c4 < 0. In the standard neo-Keynesian model, income itself reacts to fiscal policy. Thus an OLS regression of government spending on GDP growth or output gap would suffer from a simultaneity bias. To cope with this issue, following Jaimovich and Panizza (2007), Gali and Perotti (2003), and more recently Alesina et al. (2008), output of country i is instrumented using the output of the region of country i, excluding country i itself. Regions are defined per the World Bank’s classification.6 We then use the Hodrick–Prescott filter to calculate the output gap as the log deviation of GDP (in constant USD) from its HP trend component. We use an index of corruption in the political system, with higher scores indicating less corruption in public office. A positive c2i would indicate procyclical fiscal policy; the larger the coefficient, the greater is the degree of procyclicality and vice versa.7 A negative coefficient, on the other hand, indicates a countercyclical fiscal policy and a larger absolute value of the coefficient signals stronger counter-cyclicality. In order to measure the impact of institutional qualities on fiscal policy, the interaction term between the output gap and institutions is of special importance. According to the political agency model, procyclicality is more likely to persist in countries with higher levels of corruption. In particular, compared to a benevolent social planner, a corrupt government is expected to demonstrate a stronger procyclical policy response. So, if the theory is valid, both the coefficients c3 and c4 are expected to be negative; higher control of corruption should reduce the degree of procyclicality of fiscal policy. So, the marginal total effect of institutional changes on fiscal policy can be measured by (c3 + c4  OUTPUTGAPit). This captures a government’s lower propensity to spend during booms in the presence of better institutional qualities. One argument that can always be raised is that institutions are endogenous. Bad fiscal performance may necessitate and force changes in fiscal and public policy institutions. But, as Alesina and Perotti (1999) note, institutional changes are always costly and hence ‘‘there is a strong ‘‘status quo’’ bias in institutional reforms (p. 15).’’ So, despite the fact that fiscal institutions are not ‘‘set in stone’’, they can be regarded predetermined at least in the short-to-medium run. Therefore, we only instrument for the output gap in the fiscal equation. We estimate Eq. (1) using Pesaran’s (2006) Common Correlated Effects Mean Group Estimator (CCEMG). This method addresses four important issues: (i) cross-sectional dependence, which is usually present across resource-abundant economies often hit by the same shocks and common factors, (ii) unobservable country characteristics which vary over time and across 5 One alternative approach is to measure the fiscal response to the resource cycle (El Anshasy and Bradley, 2012). This approach, however, would miss some important interactions of the resource sector with the rest of the economy (summarize by GDP and the output gap) and would focus on partial changes in resource receipts (which is part GDP movements after all). Therefore, we take a standard approach to measuring procyclicality which also allows comparability across resource and non-resource countries. 6 Regions are defined as follows: High-Income OECD countries, High-Income non OECD countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, and Sub-Saharan Africa. 7 That is under the assumption that the additional nonlinear effect of the business cycle interacted with the quality of institutions (c4i) is not different from zero.

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298

289

countries, (iii) problems of identification, and (iv) the heterogeneity of marginal impacts across groups. So, this technique estimates a unique procyclicality coefficient for each country in the panel. These coefficients can be once again averaged across countries, where different weights may be applied. As shown by Pesaran and Tosetti (2011), the CCEMG methodology is robust to the presence of few ‘strong’ factors, often representing global shocks, and a large number of ‘weak’ factors, which can be thought of as local spillovers. According to Kapetanios et al. (2011) these factors may be nonstationary. So, this method works well even when panel has unit roots. 3.2. Growth estimation The main question of interest in this study is how institutions, both directly and indirectly (through fiscal performance) affect economic growth in resource-abundant economies. For this purpose, we estimate the following growth equation:

y_ ¼ b0 þ b1 Iit þ b2 Ifit þ b3 FPit þ b4 Git þ b5 NRit þ b06 X it þ at þ ui þ eit

ð2Þ

The coefficients of interest are expected to have the following sign:

b1 > 0; b2 > 0; b3 < 0; b4 6 0; b5 < or > 0 where y_ is the growth rate of per capita income, I is the control for institutions and If is the indirect effect of institutions through the fiscal channel, FP denotes fiscal performance, G is a measure of government sending, NR is a measure of natural resource intensity, X is a matrix of control variables often used in growth regressions and b06 is a vector of coefficients for the control variables, at is time specific effects, and ui captures country specific-effects. Eq. (2) decomposes the fiscal effects into (i) FP, a pure fiscal management quality effect (independent from the quality of institutions), measured by the parameter c2i from Eq. (1), (ii) G, a pure fiscal policy quantity effect measured by the changes in government spending (as a percentage of GDP) that are not due to changes in institutional qualities or the business cycle (the residuals from Eq. (1)), and (iii) If, another quality indicator, that measures the changes in the government’s propensity to spend as a result of variations in the institutional qualities, or (c3 + c4  OUTPUTGAPit) from Eq. (1). This postulates that stronger institutions would reduce the procyclicality of fiscal policy and slow down government spending growth. Therefore, higher (absolute) value of c3 and c4 is expected to lead to higher growth rates. This methodology allows disentangling the impact of institutional qualities on growth into two effects; first, a direct effect on growth (b1) conditional on a standard set of control variables, and an indirect effect through fiscal policy performance (b2) that reflects the growth gains associated with the governments’ lower propensity to spend as a result of stronger institutions, as described above. Regarding the selection of the control variables included in the main equation, we follow Sala-i-Martin et al. (2004) who examine the robustness of 67 explanatory variables used in highly cited growth studies and identify18 variables to be the most important growth determinants. Following this literature, we include initial levels of income per capita, investment share and primary school enrollment, an openness measure, price of investment, population growth, and regional and time dummies (Barro, 1991, 1996; Sachs and Warner, 1995; Sala-i-Martin et al., 2004). After we specify our model we proceed by looking into the endogeneity issues. One can again argue that institutions are endogenous to growth. Therefore, we first test for the presence of endogeneity in our main equation using the Augmented Durbin–Wu–Hausman (DWH) test which gives strong evidence of endogeneity; and thus we estimate Eq. (2) employing an IV approach (detailed below). In this regard, we examine two sources of endogeneity. First, we consider a possible feedback effect from growth to institutions. For example, countries growing faster may have more incentives to install institutional reforms. To correct for this form of endogeneity, external instruments are required to estimate a 2SLS model. There has often been controversy in the growth literature about the instruments chosen to correct for the potential endogeneity of institutions (Shaw et al., 2011). In particular, their exogeneity has often been questioned and their strength has caused arguments regarding the robustness and the identification of the parameters of interest. So, we choose to employ geographical variables (longitude, distance to the coastline, and distance to the nearest river) as IV’s for institutions in Eq. (2) mainly due to their indisputably exogenous nature and strong correlation with the corruption index (see Table 1, column 4). The second potential form of endogeneity is when institutional qualities are correlated with the country’s unobservable specific effects. As argued earlier, institutions may be hard to change in the medium-to-short term, but meanwhile their qualities can be correlated with a country’s culture, religion, or legal origins (Acemoglu et al., 2001, 2002). This correlation with the error term renders OLS estimations inconsistent. To deal with this, we use Hausman and Taylor (1981), two-step error-component estimation technique. It preserves the time-invariant variables, and removes the correlation between the included variables and the unobserved country fixed effects using internal instruments. Based on a Hausman specification test, the HT estimation showed superiority over the 2SLS IV estimation. The results of all estimations are detailed in the following section. 4. The Data We use yearly panel data on 79 countries, of which 32 are resource-abundant counties and 47 are resource-poor, for the period 1984-2008. Resource economies are identified when resources constitute more than 2% of GNI. Starting with a panel

290

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298

Table 1 Corruption, fiscal performance, and growth: various estimation techniques. Estimation technique

(1) POOLED

(2a) RE_GLS1

OLS Dependent variable: real per capita GDP growth. Sample: 1984–2008 Corruption Control

Growth of government consumption %GDP: residuals eqn1 Fiscal performance: procyclicality Lagged real per capita GDP growth Real per capita GDP 1980 Investment %GDP 1970 Primary school enrollment 1980 Lagged openness Lagged natural resources– mining %GDP

(3) RE_MLE

(clustered)

0.153 **

Corruption control: the fiscal channel (indirect)

(2b) RE_GLS2

0.119 **

0.143 **

1st-stage

0.151

Dependent Variable

**

(5) Hausman– Taylor 2nd-stage

1.44

0.23 (2.69)** 0.065 (1.74)* 0.017 (0.33) 0.092 (2.16)** 0.161 (4.22)*** 0.253 (2.44)** 0.345 (2.65)** 0.371 (1.21) 0.313 (1.94)* 0.097 (1.16)

(2.15) 0.057 (1.42) 0.02 (0.41) 0.068 (1.74)* 0.154 (2.43)** 0.173 (1.80)* 0.25 (1.68)* 0.289 (0.88) 0.129 (0.85) 0.103

(2.02) 0.056 (1.57) 0.016 (0.33) 0.073 (1.86)* 0.15 (3.95)*** 0.128 (1.52) 0.22 (1.89)* 0.157 (0.59) 0.2 (1.46) 0.087

(2.28) 0.057 (1.00) 0.017 (0.34) 0.075 (1.26) 0.153 (1.85)* 0.181 (2.31)** 0.282 (3.81)*** 0.264 (1.29) 0.267 (2.85)*** 0.084

(2.20) 0.057 (1.61)* 0.025 (0.49) 0.073 (1.86)* 0.157 (4.15)*** 0.184 (2.00)** 0.273 (2.25)** 0.279 (0.97) 0.23 (1.54) 0.083

0.69 (13.78)*** 0.058 (0.76) 0.599 (3.58)*** 0.429 (5.56)*** 0.182

(1.29) 0.06 (1.62)* 0.026 (0.50) 0.076 (1.85)* 0.15 (3.91)*** 0.178 (1.50) 0.27 (2.10)** 0.268 (0.83) 0.238 (1.49) 0.084

(1.08)

(1.16)

(1. 30)

(1.04)

(3.92)***

(1.00)

Ethno-linguistic Fractionalization

1.48 (9.15)*** 0.002 (3.29)*** 0.001 (12.23)*** 0.0001 (1.31)

Longitude of country centroid Distance to nearest inland navigable river Distance to nearest coastline Time dummies Regional dummies No. of observations No. of groups R-squared (within) Regression statistics

(4) 2SLS IV Estimation

Yes Yes 684 32 0.0714 F-stat 1.39*

No No 684 32 0.038 Wald 26.62*

No Yes 684 32 0.07 Wald 623***

Yes Yes 684 32 LR 52.29**

Yes Yes 713 0.704 F-stat 74.33***

No Yes 667 31 0.0716 F-stat 1.38*

No Yes 684 32 Wald 36.35**

Notes: Z statistics is reported in parentheses. Robust Huber-White standard errors are used. Baltagi’s (2005) 2SLS estimator results are reported in column 4. Variables (except corruption scores) are in natural logarithms. * Significance at the 10%. ** Significance at the 5%. *** Significance at the 1%.

of 94 countries, we limit the panel to only those that have at least 16 yearly observations. This restriction is placed on the sample, as detailed in the following section, to arbitrarily allow for at least two or three business cycles in each individual country in the estimation of government spending procyclicality, our indicator of choice for fiscal performance. We use various measures of institutions and governance qualities available from the International Country Risk Guide (ICRG) dataset. We choose this dataset because it has wide country coverage for a reasonably long series. In particular, we use the ICRG index of corruption and bureaucratic qualities individually, and then construct a composite governance index as an average of these two indices and the rule of law index. We also use the Worldwide Governance Indicators (WGI), available from the World Bank Institute. In particular, we use the indices of control of corruption, government effectiveness, and rule of law. Because of the shorter time span of these indicators, we average them over the period 1996 and 2008. For democratic institutions, we use the most credible sources on democracy: the Freedom House Index on political freedom and civil liberties and the Polity IV dataset that measures the strength of democratic institutions. Finally, we use an index that directly measures the transparency of the budgetary process and the quality of the budget institutions: the Open Budget Index (OBI); provided by the International Budget Partnership. The first year for which this index is available is 2006, but for a limited number of countries. In 2008, the coverage almost doubled and the technique improved. We,

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298

291

therefore, use the index for 2008 to rank countries in terms of their budgetary process transparency. We rely on the plausible assumption that the relative ranking of countries today in terms of the strength of their budget institutions reveals important information on their relative engagement in institutional reforms decades ago. In particular, a country that has a higher OBI ranking today must have been engaged in more effective institutional reforms and (or) maintenance in the preceding decades. In choosing a measure of resource abundance, we avoid using the conventional share of resource exports in a country’s total exports. This measure is criticized for its dependence on the structure of the economy. In particular, if two countries have the same GDP size and the same size of natural resource exports, the one that has relatively smaller total exports will appear more resource-dependent than the one that has larger exports. So, we use two different concepts of resource intensity. First, mining as a percentage of GDP, available from the United Nations’ Country National Accounts, provides a broader measure of resource intensity since it measures the contribution of the value added of the mining industry (including minerals and oil) in domestic production. Second, resource rents as a percentage of GNI, available from the World Bank’s WDI, is measured as the resource’s unit price minus unit cost multiplied by production. We use separate measures for mineral resources and for hydrocarbons (oil, gas, and coal), and then we combine them in one measure of total point-source resources rents. Clearly, the concept of resource rent is narrower than the concept of value added. The data on government consumption as a percentage of GDP, real per capita GDP, population growth, investment prices, openness (defined as imports plus exports as a percentage of GDP), gross fixed capital investment, and school enrollment are all available from the World Bank’s WDI. Finally, the distance to nearest river and nearest coastline and the longitude of country centroid are all from Gallup, Mellinger, and Sachs’ GEOGRAPHY DATASETS (2001), available from the Center for International Development at Harvard University. Ethno-linguistic fractionalization is obtained from Montalvo and Reynal-Querol (2005).

5. Results and discussion In this section we lay out our findings and implications for oil and mineral economies benchmarked against a sample of resource-poor economies. Table 1 presents results from the estimation of our growth equation using a variety of techniques. Corruption in public office is used as the indicator of choice for institutional qualities such that higher scores indicate less corruption. In all estimations except the 2SLS, less corruption significantly enhances growth. The indirect effect of institutions through fiscal policy appeared significant after correcting for the endogeneity of corruption. As expected, this effect is robustly positive across estimations. In addition, the coefficient of procyclicality, our choice of fiscal performance conditional on institutional qualities, is robustly significant and negative. It is notable that when endogeneity is corrected for in column 5, both the significance and the magnitude of the coefficient of procyclicality increase. The changes in government spending (what we refer to as the quantity effect) appears insignificant, reiterating previous results in the empirical growth literature that mostly found government size to have a negative or insignificant effect on growth. Our measure here, however, slightly differs from the one commonly used in the literature since it measures the changes in government consumption which are not prompted by the business cycle or an institutional change. Our results reveal that resources do not impede growth once fiscal policy and institutions are controlled for. This supports the view that the ‘‘curse’’ of resources, if any, works through windfall mismanagement and weak socio-economic and political institutions. We see that the sign of the coefficient of mining in GDP is robustly positive across all estimations, despite being insignificant. All subsequent estimations confirm the preliminary results of Table 1. Using a Hausman specification test, we were not able to reject the null hypothesis that the difference between the coefficients of the 2SLS and the HT estimation is systematic, lending support to the validity of using the latter. We see from Table 2 that not all types of socio-economic and political institutions impact growth in the same manner. Corruption control, sound rule of law, and better bureaucratic qualities positively impact growth by raising productivity and improving fiscal performance and windfall management. On the other hand, budgetary and political institutions seem to work through the fiscal transmission channel. So, in all five estimations (columns 6–9), the fiscal transmission channel is significantly positive, indicating that stronger institutions have positive impact on growth due to less spending procyclicality. As expected, the coefficient on procyclicality is significant and has the correct (negative) sign across all estimations. In addition, the direct effect of institutions is only significant and positive in the case of bureaucratic quality and governance. It is also interesting to see that the coefficient of natural resources is negative in only two out of five estimations (column 6 and column 9). In column 7, when corruption, rule of law, and bureaucratic quality are controlled for in one composite governance indicator, the coefficient of resources changes to positive. Similarly, controlling for democracy (in columns 8a and 8b), flips the sign to positive. This has the following implications: first, the ‘‘curse’’ outcome is related to the quality of specific institutions, in particular corruption and political freedom. Once these qualities are controlled for such curse seems to disappear. Second, the impact of resources on growth appears insignificant since the coefficient on lagged mining is insignificant in all estimations. Tables 3–6 present estimation results for which we consider the interaction between institutions and natural resources. In all estimations (columns 10–13) we keep the same controls used in Tables 1 and 2, but add a dummy variable for low quality institutions and an interaction term of such a dummy with a measure for natural resources. We use four different

292

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298

Table 2 Institutional qualities, fiscal performance, and growth: various measures of institutions. Institutional quality measure

(6) Bureaucratic quality

(7) Composite governance index

Dependent variable: real per capita GDP growth Estimation method: Hausman–Taylor error component estimation (Sample: 1984–2008) Measure of institutional quality .017 0.226 (2.86)* (1.86)* Institutions: fiscal channel (indirect) 0.066 0.069 (1.64)*** (1.87)*** Growth of government consumption 0.032 0.015 (%GDP): residuals eqn1 (0.53) (0.29) Fiscal performance: procyclicality 0.116 0.104 (2.48)** (2.44)** Lagged real per capita GDP growth 0.171 0.157 (4.19)*** (4.11)*** Real per capita GDP 1980 0.103 0.215 (1.07) (1.91)* Investment %GDP 1970 0.189 0.328 (1.32) (2.51)** Primary school enrollment 1980 0.459 0.229 (1.39) (0.76) Lagged openness 0.554 0.29 (2.88)** (1.79)* Lagged mining% GDP .022 0.095 (0.23) (1.12) Regional dummies Yes Yes Regional time dummies Yes Yes No. of observations 606 684 No. of groups 32 32 Wald-statistics 36.41*** 34.41 **

(8a) Political freedom & civil liberties

(8b) Democracy

(9) Budget transparency

.047 (0.71) 0.064 (1.71)* 0.016 (0.32) 0.108 (2.52)** 0.157 (4.10)*** 0.027 (0.31) 0.238 (1.71)* 0.086 (0.28) 0.32 (1.98)** 0.043 (0.52) Yes Yes 684 32 31.44**

0.004 (0.24) 0.065 (1.76)* 0.015 (0.29) 0.104 (2.39)** 0.156 (4.08)*** 0.068 (0.81) 0.289 (2.09)** 0.043 (0.14) 0.319 (1.98)** 0.060 (0.70) Yes Yes 684 32 30.99**

0.254 (0.95) 0.199 (2.45)** 0.037 (0.54) 0.296 (2.96)*** 0.167 (3.35)*** 0.086 (0.64) 0.514 (1.89)* 0.63 (1.06) 0.68 (1.98)** 0.013 (0.07) Yes Yes 408 19 27.31 *

Notes: t statistics is reported in parentheses. Robust Huber–White standard errors are used. Variables (except institutional qualities scores) are in natural logarithms. * Significance at the 10%. ** Significance at the 5%. *** Significance at the 1%. Table 3 Natural resources, corruption, fiscal performance, and the resource curse. Different resource intensity measures

(10a) Mining % GDP

(10c) Hydrocarbon rents % GNI

(10d) All point-source resource rents %GNI

Dependent variable: growth of real per capita GDP Estimation method: Hausman–Taylor error component estimation (Sample: 1984–2008) Institutions: corruption control 0.244 0.284 (2.10)** (2.30)** Institutions: the fiscal channel (indirect effect) 0.065 0.074 (1.75)* (1.90)* Fiscal performance: procyclicality 0.097 0.114 (2.27)** (2.45)** Growth of government consumption %GDP: 0.026 0.027 residuals of eqn1 (0.50) (0.52) Lagged resource intensity 0.255 0.018 (2.21)** (0.14) High corruption dummy (ICRG corruption 0.888 0.225 scores < 2.5) (1.55) (0.80) Natural resource measure  high corruption 0.39 0.069 dummy (1.93)** (0.33)

0.265 (2.14)** 0.070 (1.81)* 0.121 (2.65)** 0.032 (0.60) 0.14 (1.45) 0.396 (1.16) 0.152 (1.06)

0.29 (2.27)** 0.069 (1.80)* 0.113 (2.44)** 0.032 (0.60) 0.148 (1.72)* 0.54 (1.30) 0.181 (1.08)

Regional dummies Regional time dummies No. of observations No. of groups Wald statistics

Yes Yes 650 32 39.46***

Yes Yes 650 32 39.53***

Yes Yes 684 32 32.36***

(10b) Mineral rents % GNI

Yes Yes 650 32 38.87***

Note: t statistics is reported in parentheses. Robust Huber–White standard errors are used. All regressions included all control variables and the lagged dependent variable as in Table 1. Variables (except corruption scores) are in natural logarithms. * Significance at the 10%. ** Significance at the 5%. *** Significance at the 1%.

293

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298 Table 4 Natural resources, governance, fiscal performance, and the resource curse. Different resource intensity measures

(11a) Mining % GDP

(11c) Hydrocarbon rents % GNI

(11d) All point-source resource rents %GNI

Dependent variable: growth of real per capita GDP Estimation method: Hausman–Taylor error component estimation (Sample: 1984–2008) Institutions: governance 0.39 0.42 (2.47)** (2.44)** Institutions: the fiscal channel (indirect effect) 0.067 0.074 (1.80)* (1.91)* Fiscal performance: procyclicality 0.102 0.124 (2.40)** (2.71)** Growth of government consumption %GDP: 0.021 0.024 residuals of eqn1 (0.41) (0.46) 0.048 Lagged resource intensity (log) 0.32 (0.36) (2.62)** Bad governance dummy (ICRG governance 0.080 0.53 scores < 2.5) (1.37) (1.02) Natural resource measure  bad governance 0.507 0.20 dummy (2.29)** (1.06)

0.35 (2.16)** 0.074 (1.92)* 0.132 (2.88)** 0.024 (0.46) 0.056 (0.62) 0.42 (1.43) 0.191 (1.08)

0.40 (2.40)** 0.074 (1.92)* 0.126 (2.78)** 0.028 (0.53) 0.154 (1.48) 0.73 (1.55) 0.22 (1.31)

Regional dummies Regional time dummies No. of observations No. of groups Wald statistics

No Yes 650 32 30.03***

No Yes 650 32 39.29***

Yes Yes 684 32 42.48***

(11b) Mineral rents % GNI

No Yes 650 32 39.12***

Note: t statistics is reported in parentheses. Robust Huber–White standard errors are used. All regressions included all control variables and the lagged dependent variable as in Table 1. Variables (except corruption scores) are in natural logarithms. * Significance at the 10%. ** Significance at the 5%. *** Significance at the 1%.

Table 5 Natural resources, democratic institutions, fiscal performance, and the resource curse. Different resource intensity measures

(12a) Mining % GDP

(12c) Hydrocarbon rents % GNI

(12d) All point-source resource rents %GNI

Dependent variable: growth of real per capita GDP Estimation method: Hausman–Taylor error component estimation (Sample: 1984–2008) Institutions: democracy 0.065 0.026 (1.59) (0.58) Institutions: the fiscal channel (indirect 0.058 0.072 effect) (1.68)* (1.85)* Fiscal performance: procyclicality 0.073 0.121 (1.63)* (2.51)** Growth of government consumption %GDP: 0.016 0.022 residuals eqn1 (0.32) (0.42) Lagged resource intensity (log) 0.44 0.043 (2.99)*** (0.30) Non-democracy dummy (Polity scores < 3) 1.24 0.38 (1.49) (0.62) Natural resource measure  non-democracy 0.73 0.078 dummy (3.11)*** (0.27)

0.020 (0.46) 0.069 (1.78)* 0.108 (2.16)** 0.021 (0.39) 0.094 (0.77) 0.73 (1.18) 0.28 (1.44)

0.032 (0.69) 0.069 (1.78)* 0.117 (2.30)** 0.022 (0.41) 0.138 (1.03) 0.91 (1.30) 0.249 (1.23)

Regional dummies Regional time dummies No. of observations No. of groups Wald statistics

Yes Yes 650 32 33.31***

Yes Yes 650 32 32.47***

Yes Yes 684 32 42.40***

(12b) Mineral rents % GNI

Yes Yes 650 32 32.70***

Note: t statistics is reported in parentheses. Robust Huber–White standard errors are used. All regressions included same control variables and the lagged dependent variable as in Tables 1 and 2. Variables (except polity scores) are in natural logarithms. * Significance at the 10%. ** Significance at the 5%. *** Significance at the 1%.

measures of institutional qualities (one in each table) and four measures of natural resources (one in each column). One general conclusion is that all core results were robust to this exercise. Again, the indirect effect of institutions through the fiscal channel is significantly positive; procyclicality coefficient is significantly negative; the changes in government spending are insignificant; and the direct effect of institutions is positive and

294

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298

Table 6 Natural resources, budget institutions, fiscal performance, and the resource curse. Different resource intensity measures

(13a) Mining % GDP

(13c) Hydrocarbon rents % GNI

(13d) All point-source resource rents %GNI

Dependent variable: growth of real per capita GDP Estimation method: Hausman–Taylor error component estimation (Sample: 1984–2008) Institutions: budget openness 0.61 0.24 (0.86) (0.46) Institutions: the fiscal channel (indirect effect) 0.192 0.196 *** (2.34) (2.33)** Fiscal performance: procyclicality 0.242 0.30 (2.01)** (2.93)*** Growth of government consumption %GDP: residuals eqn1 0.037 0.039 (0.54) (0.56) Lagged Resource Intensity (log) 0.319 0.081 (0.80) (0.31) Non transparent budget dummy (open budget index < 35) 2.15 0.33 (0.95) (0.28) Natural resource measure  non-transparent budget dummy 0.48 0.193 (0.92) (0.45)

0.46 (1.13) 0.234 (2.67)** 0.314 (2.81)** 0.043 (0.61) 0.098 (0.51) 0.269 (0.19) 0.20 (0.47)

0.55 (1.23) 0.222 (2.53)** 0.310 (2.56)** 0.0457 (0.64) 0.189 (1.06) 0.20 (0.12) 0.288 (0.68)

Regional dummies Regional time dummies No. of observations No. of groups Wald statistics

Yes Yes 389 19 28.88*

Yes Yes 389 19 29.74*

Yes Yes 408 19 28.27*

(13b) Mineral rents % GNI

Yes Yes 407 19 27.67*

Note: t statistics is reported in parentheses. Robust Huber–White standard errors are used. All regressions included same control variables and the lagged dependent variable as in Tables 1 and 2. Variables are in natural logarithms. * Significance at the 10%. ** Significance at the 5%. *** Significance at the 1%.

significant when corruption and governance qualities are considered. Democratic and budget institutions seem to be in effect only through the fiscal channel but not independently. The sign on natural resource intensity is again positive, except for the case of minerals. This positive effect is significant only in the case of mining industry share in GDP. When this measure is interacted with a dummy for weak institutions, this term appeared significantly negative, implying that resources are a ‘‘blessing’’ only in good governance countries. For countries such as Zimbabwe or Gabon (corruption control < 2.5) a 1% increase in mining share would reduce growth by 0.14 percentage points [the marginal effect of resources (+0.26) minus the extra negative effect in case of poor institutions (0.39)]. In countries with good governance (corruption control > 2.5) such as Canada and Norway, growth increases by 0.26 percentage points. In the case of budget transparency (Table 6), although the sample is significantly smaller due to the limited coverage of the index, we still get the same signs for natural resources. Our results are consistent with the findings of other studies that a curse is very much associated with weak institutions (e.g., Mehlum et al., 2006; Collier and Goderis, 2007). This result however, does not say much about whether natural resources cause poor institutions.

5.1. Robustness checks In examining the indirect effect of institutions through the fiscal transmission channel, we used three constructed indicators from our first fiscal equation to serve as explanatory variables in the outcome regression. We, therefore, check whether results related to this transmission channel are driven by generated regressors. We crudely capture the fiscal-indirect-institutional channel by interacting institutions with observed (actual) procyclicality.8 In particular, we use government consumption growth relative to the growth in real output for actual procyclicality, instrumenting it by its first lag to avoid endogeneity. We also use a different governance index in order to check the robustness of the results to changes in the institutional measures. The KKM Worldwide Governance Indicators, averaged over the period 1996–2008, namely control of corruption, rule of law, and government effectiveness replace the ICRG indicators. The growth equation is thus re-estimated, using the lagged actual procyclicality and a three-way interaction term of actual procyclicality, a governance indicator, and a dummy for poor institutional quality. The results support our main hypothesis that there is interplay between institutions and fiscal 8 It should be emphasized that this interaction term merely captures an association between institutional qualities and procylicality in impacting growth. But, it says nothing about how poor institutions interact with fiscal performance, and weather a causal relationship exists between the two. It also says nothing about the direction of causality, if any. But, again, it serves as a crude check on whether the (total) growth effect of procyclicality can be conditioned on the quality of institutions.

295

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298 Table 7 Growth, institutions, and fiscal performance in resource-poor countries.a Institutional quality measure

Institutions Institutions: the fiscal channel Fiscal performance: procyclicality Growth of government consumption %GDP: residuals eqn1 Lagged mining % GDP Low quality institutions dummy Mining  low quality institutions Regional dummies Regional time dummies No. of observations No. of groups Wald statistics

Resource-poor countries All sample countries

Resource-poor countries Non-democracies (Polity2 < 8)

(14a) Corruption Control

(14b) Governance

(14c) Open Budget

(15a) Corruption Control

(15b) Governance

(15c) Open Budget

0.001 (0.40) 0.0011 (0.10) 0.002 (0.85) 0.057 (3.09)*** 0.0016 (0.37) 0.002 (0.28) 0.0004 (0.08)

0.002 (0.68) 0.0003 (0.03) 0.002 (0.95) 0.058 (3.16)*** 0.0065 (1.38) 0.0023 (0.29) 0.012 (2.15)*

0.053 (0.83) 0.006 (0.50) 0.001 (0.55) 0.019 (1.48) 0.009 (1.38) 0.032 (0.35) 0.003 (0.42)

0.005 (1.66)* 0.005 (0.32) 0.001 (0.96) 0.056 (2.19)*** 0.004 (0.64) 0.003 (0.23) 0.004 (0.50)

0.013 (2.14)** 0.009 (0.56) 0.001 (1.40) 0.058 (2.24)*** 0.001 (0.18) 0.016 (1.28) 0.015 (1.79)*

0.046 (1.87)* 0.009 (0.61) 0.001 (0.51) 0.038 (1.60)* 0.016 (1.23) 0.019 (0.27) 0.016 (0.84)

Yes Yes 1025 47 66.49***

Yes Yes 1018 47 76.11***

Yes Yes 517 23 47.81***

Yes Yes 419 25 51.35***

Yes Yes 419 25 67.33***

Yes Yes 245 15 19.94***

Note: t-statistics is reported in parentheses. Robust Huber–White standard errors are used. All regressions include all control variables in Table 2 and the lagged dependent variable. Variables (except institutional qualities scores) are in natural logarithms. a Mineral and hydrocarbon resources rents do not exceed 2% of GNI, on average, during period 1984–2008. * Significance at the 10% level. ** Significance at the 5% level. *** Significance at the 1% level.

policy in impacting growth. This turned to be robust to using actual observed procyclicality. We find that in countries with poor institutions, more procyclicality reduces growth, and this negative effect is larger the poorer the quality of institutions. Another result that remained robust to this exercise is the negative affect of resources when institutions are poor.9 5.2. Benchmarking against resource-poor countries How do the above-presented results differ from those in resource-poor economies? In particular, how do fiscal policy and institutions interact to impact growth in resource-poor economies as compared to resource countries? To explore this we construct a sample of 47 resource-poor economies in which energy and mineral resources constitute no more than 2% of GNI. We apply the same methodology explained in section 2 to obtain the spending marginal response to the business cycle, the spending marginal response to changes in institutional qualities, and a pure change in government size (the residual). Using the same Hausman–Taylor error component estimation method, and mining (% of GDP) as our resource measure, we estimate models identical to those in 10a, 11a, and 13a with three institutional measures: corruption control, governance, and open budgets. The results are shown in Table 7. In columns 14a, 14b, and 14c we pool all resource-poor countries irrespective of their political regimes. Then, in columns 15a, 15b, and 15c, we consider a subsample of non-democracies to see whether the intuition we get is seriously altered by political system imperfections. The result that we should highlight here is that, unlike in resource economies, the indirect fiscal channel of institutions is not significant in democracies and non-democracies alike. This lends strong support to our hypothesis that in resource-rich economies – in particular – fiscal management is far affected by governance qualities because of resource windfall conflicts and pressures from different interest groups. These post-boom conflicts hurt growth; and this effect is not expected to occur in resource-poor economies. In addition, unlike in resource countries where variations in government size seem insignificant after controlling for the ‘‘fiscal performance’’ channel, we see that in non-resource countries the standard negative effect of government consumption holds; so a larger government size is unproductive in both types of economies, but it is more damaging to growth in non-resource countries. This result is anticipated given the greater role of the government in resource economies. In nonresource economies, however, a larger government size would crowd out more private investment and spending. Lastly, good governance and more budget transparency improve growth performance as in resource-rich economies, and this effect is quite significant in less democratic countries. When political competition is not perfect, strengthening governance qualities and budget institutions has a much significant yield. This last finding is consistent with the findings in Table 5 9

The full tables of results are available from the authors upon request.

296

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298

for resource-rich economies. In conclusion, the direct effect of good governance on growth is positive and similar in both samples. However, the quality of fiscal policy management matters more to growth in resource-rich economies than in resource-poor countries due to the resource windfall management challenges. ‘‘Quantity’’ or government size, on the other hand, matters more in non-resource countries where expanding the government is usually at the expense of crowding out more productive private investments.

6. Conclusion In this paper we attempt to unravel the effect of institutional qualities on the quality of fiscal performance and through this channel on growth. To this end, we decompose the fiscal policy reactions into a procyclicality parameter, the marginal effect of institutional changes on fiscal policy, and a residual that measures changes in government size not due to the business cycle or institutional changes. Using a battery of tests, we find that the quality of fiscal policy – and not the quantity – matters to growth in the group of resource-rich economies. Better governance, stronger democratic institutions, and more transparent budgets improve resource windfall management, leading to higher growth rates. This result is robust to changing the measure of institutions, and resource intensity. We also find that a resource curse only exists under conditions of weak democratic and governance institutions, in line with previous literature. A point worthy of future research – that our study did not address – is the potential impact of institutional and governance qualities on growth through their impact on government spending composition. This has special relevance to resource-rich economies that need to direct more of their depleting resource rents towards the accumulation of human capital and building productive non-resource capacities.

Acknowledgments We are extremely grateful to the United Arab Emirates University for the research grant the first author received to produce this research. We are also grateful to anonymous referees, Abdulnasser Hatemi-J, Ahmed Ismail, Timo Boppart, Jorge Madeira Noqueira, Raimundo Soto, and the participants of the Economics and Finance seminar series at the College of Business and Economics, UAEU, and the 4th Annual Research Symposium in Business and Economics for their many useful comments. We also would like to thank Manar Hamid and Eman Abdulla for valuable research assistance. All errors remain ours.

Appendix A.

Resource countries (32) Sample countries Algeria Australia Bangladesh Botswana Cameroon Canada Cote d’ Ivoire Egypt Gabon Ghana Guinea Hungary India Indonesia Iran Ireland

Resource-poor countries (47) Malaysia Netherlands New Zealand Norway Oman Pakistan Papua New Guinea Philippines Romania South Africa Sudan Syria Trinidad Tunisia Zambia Zimbabwe

Austria Belgium Burkina Faso Bulgaria Brazil Switzerland Costa Rica Germany Denmark Spain Ethiopia France United Kingdom Gambia, The Greece Guatemala

Hong Kong Honduras Haiti Iceland Israel Italy Jordan Japan Kenya Korea, Rep. Lebanon Sri Lanka Luxembourg Morocco Mali Malta

Mozambique Malawi Niger Nicaragua Portugal Paraguay Senegal Singapore Sierra Leone El Salvador Sweden Thailand Turkey Uganda United States

297

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298

Variable

Mean

Std. Dev.

Min

Max

Summary statistics for resource countries (32 groups) Real pc GDP growth 737 Population growth 737 Investment prices 769 Openness 769 Investment % GDP 1970 768 Real per capita SDP 1980 ($US 2005) 768 Secondary education enrollment 1980 763 Government consumption % of GDP 745 Ethno-linguistic fractionalization 745 Longitude of country centroid 769 Latitude (distance from the equator) 769 Distance to nearest navigable river 769 Distance to nearest coastline 769 Polity 2 769 Civil liberties 769 Political Freedom 769 Corruption 768 Governance index 768 Bureaucratic quality 768 Open Budget Index 457 Mining % GDP 769 Mineral rents % GNI 723 Energy rents (oil, NG, coal) MGNI 723

No Obs.

0.009 1.7 77.55 70.02 20.46 7617 37.63 8.00 0.494 3.11 22.57 216.3 237.0 1.760 3.807 3.833 3.092 3.020 2.329 47.11 12.70 1.750 19.76

1.656 1.00 65.80 36.42 11.88 7823 24.50 8.00 0.28 65.34 24.34 298.1 343.4 7.17 1.93 2.214 1.476 1.23 1.159 24.66 12.20 4.846 12.79

11.4 1.00 0.671 4.83 3.11 247.3 3S 7 0.01 0.05 98.3 35.81 11.0 10.9 10 1 1 0 0.33 0 0 0.5 0 0

10.44 6.8 634.7 220.4 50.12 25972.5 163.8 26.11 0.9 134.6 74.70 1466.6 1567.9 10 7 7 6 5.3 4 87 55.01 48.48 112.3

Summary statistics for resource-poor countries Real pc GDP growth Population growth Investment prices Openness Investment % GDP 1970 Real per capita GDP 1980 (SUS 2005) Secondary education enrollment 1930 Government consumption % of GDP Ethno-linguistic fractionalization Longitude of country centroid Latitude (distance from the equator) Distance to nearest navigable river Distance to nearest coastline Polity2 Civil liberties Political freedom Corruption Governance index Bureaucratic quality Open Budget Index Mining % GDP Mineral rents % GNI Energy rents (oil, NG, coal) % GNI

1.78 1.54 45.82 75.54 20.99 9813.27 89.12 15.98 0.38 2.96 25.39 207.57 277.58 4.93 2.96 2.72 3.57 3.34 2.41 48.19 3.59 0.14 0.24

3.89 125 17.05 56.54 6.295 8814.66 27.14 5.81 0.28 54.22 24.03 277.74 307.32 6.16 1.66 1.89 1.41 1.27 1.29 24.09 2.41 0.57 0.73

21.34 3.29 16.25 10.81 763 439.78 16.51 4.99 0.01 112.10 32.87 22.39 22.51 9 1 1 0 0.58 0 3 0.226 0 0

29.72 11.18 92.44 464.32 35.22 28533.63 135.73 43.47 0.93 138.85 64.92 1180.26 1197.49 10 7 7 6 5.3 4 88 23.76 7.21 11.48

(47 groups) 1025 1025 991 1025 1025 1025 1025 1025 965 97S 76 976 975 892 996 996 1025 1018 1018 517 1025 1018 1018

References Acemoglu, D., Johnson, S., Robinson, J., 2001. The colonial origins of comparative development: an empirical investigation. American Economic Review 91 (5), 1369–1401. Acemoglu, D., Johnson, S., Robinson, J., 2002. Reversal of fortune: geography and institutions in the making of the modern world income distribution. Quarterly Journal of Economics 117, 1231–1294. Alesina, A., Perotti, R., 1999. Budget deficits and budget institutions. In: Poterba, J., Hagen, J. (Eds.), Fiscal Institutions and Fiscal Performance. The University of Chicago Press, Chicago, pp. 13–36.

298

A.A. El Anshasy, M.-S. Katsaiti / Journal of Macroeconomics 37 (2013) 285–298

Alesina, A., Tabellini, G., Campante, F., 2008. Why is fiscal policy so often procyclical? Journal of the European Economic Association 6 (5), 1006–1036. Baltagi, B., 2005. Econometric Analysis of Panel Data. Sussex: John Wiley and Sons. Barro, R., 1979. On the determination of the public debt. Journal of Political Economy 64, 93–110. Barro, R., 1991. Economic Growth in a Cross Section of Countries. NBER Working Papers 3120. Barro, R., 1996. Institutions and growth, an introductory essay. Journal of Economic Growth 1 (2), 145–148. Collier, P., Goderis, B., 2007. Commodity Prices, Growth, and the Natural Resource Curse: Reconciling a Conundrum, Centre for the Study of African Economies Working Paper series/2007–15, Oxford University. Dollar, D., Kraay, A., 2003. Institutions, Trade, and Growth: Revisiting the Evidence. World Bank Policy Research Working Paper 3004. Washington DC: WB. El Anshasy, A.A., 2012. Oil revenues, government spending policy, and growth. Public Finance and Management 12 (2), 120–146. El Anshasy, A.A., Bradley, M.D., 2012. Oil prices and the fiscal policy response in oil-exporting countries. Journal of Policy Modeling 34 (5), 605–620. Fasano, U., Wang, Q., 2001. Fiscal Policy and Non-oil Economic Growth: Evidence from GCC Countries, WP/01/195, Washington DC: IMF. Frankel, J., 2010. The Natural Resource Curse: A Survey, NBER Working Paper No. 15836. Frankel, J., Vegh, C., Vuletin, G., 2011. On graduation from fiscal procyclicality. NBER Working Paper No. w17619. Gali, J., Perotti, R., 2003. Fiscal policy and monetary integration in Europe. Economic Policy 18, 533–572. Ghosh, S., Gregoriou, A., 2010. Can corruption favor growth via the composition of government spending? Economics Bulletin 30 (3), 2270–2278. Hausman, X., Taylor, W., 1981. Panel data and unobservable individual effects. Econometrica 49 (6), 1377–1398. Husain, A., Tazhibayeva, K., Ter-Martirosyam, A., 2008. Fiscal Policy and Economic Cycles in Oil-exporting Countries, WP/08/253, Washington DC: IMF. Ilzetski, E., Carlos, V., 2008. Procyclical Fiscal Policy in Developing Countries: Truth or Fiction? NBER Working Paper No. 14191. Isham, J., Prichett, L., Woolcock, M., Busby, G., 2005. The varieties of resource experience: natural resource export structures and the political economy of economic growth. The World Bank Economic Review 19 (2), 141–174. Jaimovich, D., Panizza,U., 2007. Procyclicality or Reverse Causality? InterAmerican Development Bank, Research Department, Working Paper 599. Kaminsky, G., Reinhart, C., Végh, C., 2005. When it rains it pours: procyclical capital flows and macroeconomic policies. NBER Macroeconomics Annual 2004 19, 11–82. Kapetanios, G., Pesaran, M.H., Yamagata, T., 2011. Panels with non-stationary multifactor error structures. Journal of Econometrics 160 (2), 326–348. Knack, S., Keefer, P., 1995. Institutions and economic performance: cross-country tests using alternative measures. Economics and Politics 7, 207–227. Kontopoulos, Y., Perotti, R., 1999. Government fragmentation and fiscal policy outcomes: evidence from OECD countries. In: Poterba, J., Hagen, J. (Eds.), Fiscal Institutions and Fiscal Performance, NBER Conference Report. The University of Chicago Press, Chicago, pp. 81–102. Mauro, P., 1998. Corruption and the composition of government expenditure. Journal of Public Economics 69, 263–279. Mehlum, H., Moene, K., Torvik, R., 2006. Institutions and the resource curse. Economic Journal 116, 1–20. Montalvo, J.G., Reynal-Querol, M., 2005. Ethnic polarization, potentail conict, and civil wars. American Economic Review 95, 796–816. Persson, T., Tabellini, G., 2004. Constitutional rules and fiscal policy outcomes. American Economic Review 94, 25–46. Pesaran, H., 2006. Estimation and inference in large heterogeneous panels with a multifactor error structure. Econometrica, Econometric Society 74 (4), 967–1012. Pesaran, H., Tosetti, E., 2011. Large panels with common factors and spatial correlation. Journal of Econometrics 161 (2), 182–202. Pieschacon, A., 2008. Oil prices and Fiscal Policy in Small Open Economies, Ph.D. Dissertation, Duke University. Robinson, J., Torvik, R., Verdier, T., 2006. The political foundations of the resource curse. Journal of Development Economics 79, 447–468. Rodrik, D., 1998. Where Did All the Growth Go? External Shocks, Social Conflicts, and Growth Collapses, NBER Working Paper No. 6350. Rodrik, D., Subramanian, A., Trebbi, F., 2002. Institutions Rule: The Primacy of Institutions over Integration and Geography in Economic Development, IMF Working Paper No. 02/189. Sachs, J., Warner, A., 1995. Natural Resource Abundance and Economic Growth. NBER Working Paper 5398. Cambridge, MA. Sala-i-Martin, X., Subramanian, A., 2003. Addressing the Natural Resource Curse: An Illustration from Nigeria, WP/03/139, Washington DC: IMF. Sala-I-Martin, X., Doppelhofer, G., Miller, R., 2004. Determinants of long-term growth: a Bayesian averaging of classical estimates (BACE) approach. American Economic Review 94 (4), 813–835. Shaw, P., Katsaiti, M., Jurgilas, M., 2011. Corruption and growth under weak identification. Economic Inquiry 49 (1), 264–275. Stein, E., Talvi, E., Grisanti, A., 1999. Institutional arrangements and fiscal performance: the Latin American experience. In: Poterba, J., Hagen, J. (Eds.), Fiscal Institutions and Fiscal Performance. The University of Chicago Press, Chicago, pp. 103–134. Tornell, A., Lane, P.R., 1998. Are windfalls a curse? A non-representative agent model of the current account. Journal of International Economics 44, 83–112. Tornell, A., Lane, P.R., 1999. The voracity effect. American Economic Review 89, 22–46. Wick, K., Bulte, E., 2009. The curse of natural resources. Annual Review of Resource Economics 1, 139–156.