Governance and Economic Development in MENA Countries: Does Oil Affect the Presence of a Virtuous Circle?

Abstract

While most of the studies attest the presence of a significant virtuous circle between governance and economic development, this is the first article that investigates its sustainability on a sample of 23 Middle East and North Africa (MENA) countries covering the period 1996–2005 and stressing salient features and major challenges currently facing the MENA Region, namely, oil, growth of the working age population, enrollment in secondary education, high reliance on fixed investment and presence of religious fractionalization. It studies the significance of a country's maturity and the embedded institutional learning-by-doing phenomenon (as reflected in the number of years of a country's independence) as a predictor of the quality of governance. The results of the Generalized Method of Moments show that governance reforms remain the most significant challenge facing the countries of the MENA Region, notably in countries relying on oil, a point source product. Among the six governance indicators, voice and accountability, government effectiveness, and control of corruption exhibit the highest economic impact on economic development. Finally, sound governance is a mere reflection of a learning-by-doing process and is endogenous to “greater maturity” of countries.

KEYWORDS:

• economic development
• governance
• institutions
• MENA region
• oil

Notes

Note. Averages of each dependent and independent variables over the period 1996–2005 are taken into consideration, except for independence where we have taken the number of years of independence in 2005. Lebanon and Libya are excluded from the First Stage Regression 1 for data unavailability. Results of the First Stage Regression 2 (not reported in this table) are insensitive to the removal of data relative to Lebanon and Libya. Estimation results of first regression 2 are non sensitive to the use of the standardized level of economic development as a dependent variable. Results are White robust-heteroskedasticity estimates for the variance–covariance matrix and are reported in the above table. P-values: *less than 0.1; **less than 0.05; ***less than 0.01.

Note. Valid sample size is 230 observations; 23 cross-sections and 10 time-series observations over the period 1996–2005. Std GDP is the z-score of GDP_c.

Note. Valid sample size is 230 observations; 23 cross-sections and 10 time-series observations over the period 1996–2005.

Note. Valid sample size is 230 observations; 23 cross-sections and 10 time-series observations over the period 1996–2005.

Note. Valid sample size is 210 observations; 21 cross-sections and 10 time-series observations over the period 1996–2005. Lebanon and Libya are excluded from the sample for data unavailability. POP is the growth rate of working age population plus 5% (on account of technological progress and capital depreciation). Std GDP_c is the z-score of GDP_c. Log Predicted Trade Share is used as an instrument for the Overall Quality of Governance (i.e., simple average of the six dimensions of governance indicators). GFCF, EDU and POP and Oil are assumed to be exogenous. In all models convergence is achieved after two coefficient iterations. Standard errors of coefficients are reported between parentheses. P-values: *less than 0.1; **less than 0.05; ***less than 0.01.

Note. Valid sample size is 210 observations; 21 cross-sections and 10 time-series observations over the period 1996–2005. Lebanon and Libya are excluded from the sample for data unavailability. POP is the growth rate of working age population plus 5% (on account of technological progress and capital depreciation). Log Predicted Trade Share is used as an instrument for each of the governance indicators. GFCF, EDU and POP and Oil are assumed to be exogenous. Standard errors of coefficients are reported between parentheses. P-values: *less than 0.1; **less than 0.05; ***less than 0.01.

Note. Valid sample size is 230 observations; 23 cross-sections and 10 time-series observations over the period 1996–2005. Historical infant mortality in 1980 is used as instrumental variable for the level of economic development. Religious Fractionalization (RELGFRACT), Independence (INDEP) and OIL Dummy are exogenous variables. We reprocessed the estimation on sample of 21 cross-sections after the omission of Lebanon and Libya. The results (not reported within this paper) were not significantly and materially sensitive. Standard errors of coefficients are reported between parentheses. P-values: *less than 0.1; **less than 0.05; ***less than 0.01.

Unless otherwise mentioned, the definition of the MENA Region followed in this article covers Algeria, Bahrain, Djibouti, Egypt, Iran, Iraq, Israel, Jordan, Kuwait, Lebanon, Libya, Mauritania, Morocco, Oman, Qatar, Saudi Arabia, Somalia, Sudan, Syrian Arab Republic, Turkey, Tunisia, United Arab Emirates, and Yemen. Following Makdisi, Fattah, & Weil. (2005), MENA oil-exporting countries are: Algeria, Bahrain, Iran, Iraq, Kuwait, Libya, Oman, Qatar, Saudi Arabia, and United Arab Emirates.

In light of the absence of a consensual agreement on a single and precise definition of governance, literature uses interchangeably the terms governance, institutions and institutional quality (Kaufmann & Kraay, 2007).

At a very early stage of the Industrial Revolution (i.e., in 1820), now-developed countries were suffering from the absence of a universal suffrage (even for men) and were lacking effective formal institutions. Several institutional developments accompanied the acceleration of the industrialization process (e.g., by 1875 some of the now-developed countries achieved universal male suffrage, instituted patent laws, and introduced the concept generalized limited liability). When the process of industrialization reached its maturity stage (i.e., by 1913) further institutional enhancements shaped the history of now-developed countries: secret ballots were introduced in France, bureaucratic professionalism emerged in the United States of America, compulsory auditing was introduced in the United Kingdom. Withal, the institutional quality of now-developed countries was lower than what is expected from currently developing ones. By 1913, the United States of America reached a comparable level of economic development to the one of Mexico in the contemporary period (i.e., the per capita GDP of the United States in 1913 was around 5301 international Geary-Khamis dollars, while the per capita of Mexico at the beginning of the 1990s was around 5900 international Geary-Khamis dollars). However, the United States was institutionally less developed with the presence of formal disenfranchised women, blacks and many other minorities, the absence of a bankruptcy law and regulation on child labor and many other features.

Among the critiques addressed: (1) governance indicators cannot be compared over time as they are scaled to have the same global averages in every period; (2) governance cannot be compared across countries and over time since different sources are used in the estimation of the governance indicators; and (3) the underlying imprecision of the governance indicators does not permit meaningful comparisons of governance over time and across countries.

Kaufmann et al. (2007b) provide convincing counter-arguments to the above stated critiques: (1) Since the world averages of the individual sources of governance do not vary over time, the relative changes in country scores on the aggregate governance indicators reflect absolute changes making it appropriate to maintain the assumption that the governance indicators have the same mean in each period; (2) the aggregation model used (i.e., the unobserved component model) enables to place different underlying data sources into common units providing the basis for comparisons across countries although different underlying definitional sources are used (e.g., the aggregation methodology extracts the common component between corruption in judiciary and corruption in procurement defining, therefore one implicit definition of corruption labeled control of corruption); (3) The presence of margins of error in the governance indicators provide guidance as to which observed differences or changes in point estimates are likely to be meaningful and which are not and, therefore, enhances the ability to make comparisons across countries and over time.

The correlation between the explanatory variable and the error term results from: (1) endogeneity; (2) omitted variables; and (3) measurement errors in the explanatory variables.

Observations on the same individuals in two different periods are correlated, but observations on two different individual are not.

As Hall and Jones state, following Kamarck (1976), the direct relationship between the distance from equator and the level of economic development is not but through the prevalence of disease and the presence of a highly variable rainfall and inferior soil quality.

Kaufmann and Kraay (2002) used the information on the accuracy of the governance indicators (i.e., the standard errors of point estimates) and the correlation between the error terms of the structural equations studying the issue of causality (i.e., the impact of governance on per capita income and the impact of per capita income on governance while controlling for other observed factors) to identify the causal effects running from per capita income to governance.

We do not test the weakness of the initial level of GDP per capita since we think it does have a direct influence on the current governance quality.

The Pearson correlation coefficient between the instrument and economic development (−0.8) is much higher than the one between historical infant mortality rates and governance (−0.5) adding further evidence for the validity of employing the historical infant mortality rate as an instrument.

The bias relative to the ordinary least squares generally is well approximated by the inverse of the First Stage F-Statistic.

The ADF assumes the presence of an individual unit root under the null hypothesis.

In 14 cases out of 23 the unit-root null cannot be rejected.

In 17 cases out of 23 the unit-root null cannot be rejected.

In 17 out of 23 cases the unit-root cannot be rejected.

In 12 cases out of 23 the unit-root null cannot be rejected.