Journal of Policy Modeling 35 (2013) 669–683

Available online at www.sciencedirect.com

Economic growth in MENA countries: Is thereconvergence of per-capita GDPs?

M. Simona Andreano a, Lucio Laureti b, Paolo Postiglione c,∗a Universitas Mercatorum, Via Appia Pignatelli 62, 00178 Rome, Italy

b LUM “Jean Monnet University” SS 100 Casamassima (BA) and Jean Monnet Permanent Course – EuropeanCommission, Italy

c “G. d’Annunzio” University of Chieti-Pescara, Department of Economic Studies, Viale Pindaro 42,65127 Pescara, Italy

Received 15 December 2011; received in revised form 3 July 2012; accepted 15 October 2012Available online 13 March 2013

Abstract

In the last years a central issue in economic growth debate has been represented by the convergenceproblem. Many empirical economists have noticed that per-capita GDPs of poor regions tend to converge tothose of the richer ones. This tendency is more evident in the nineties when the globalization phenomenonwas born. In this paper we use a conditional β-convergence approach to evaluate the economic growth ofthe Middle East and North Africa (MENA) countries. In particular, we use a set of state, environmental,and economic covariates as conditioning variables of the model. The MENA region is daily at the center ofeconomic and political debate, and this stylized fact represents a further source of interest. Our data set isconstituted by 26 countries, and ranges from 1950 to 2007.© 2013 Society for Policy Modeling. Published by Elsevier Inc. All rights reserved.

JEL classification: C13; O47; O57; F11

Keywords: β-Convergence; MENA countries; Determinants of growth

1. Introduction

There are very large differences in per-capita GDPs across countries today. The richer countriesshow a per-capita GDP more than thirty times larger than that of the poorest countries in terms

∗ Corresponding author. Tel.: +39 08545083229.E-mail addresses: s.andreano@unimercatorum.it (M.S. Andreano), laureti@lum.it (L. Laureti), postigli@unich.it

(P. Postiglione).

0161-8938/$ – see front matter © 2013 Society for Policy Modeling. Published by Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.jpolmod.2013.02.005

670 M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683

of Purchasing Power Parity (PPP) adjusted dollars.1 For example, in 2011 per-capita GDP in theUnited States (US) was $48,442 (valued at current international dollars, World Bank source),while it was $8442 in China, $3650 in India, $2532 in Nigeria, and much lower in some otherAfrican countries such as Chad, Ethiopia, and Mali. The gap is obviously larger when there isno PPP adjustment. It is worth noticing that high-income levels generally reflect high standardsof living. In fact, per-capita GDP is usually used as a proxy for the quality of life in differentcountries, but we are aware that material wealth is only one of many aspects of life that enhanceeconomic well being. For example, recent estimates suggest that longevity has been a quantita-tively important component in welfare in the US during the twentieth century (Nordhaus, 2003).So far, understanding the motivation of the presence of these persistent economic differencesamong countries represents one of the most important challenges facing social sciences.

During the last years, the analysis of economic growth has become increasingly popular inthe macroeconomic literature (Abramovitz, 1986; Barro & Sala-i-Martin, 1995). Many empiricaleconomists, in agreement with Solow’s neoclassical growth model (1956), have observed that per-capita GDPs of poor regions grow more quickly than those of the rich ones, in other words poorcountries tend to finally catch up rich ones (Barro & Sala-i-Martin, 1992). This phenomenon,known in literature as economic convergence, implies a long run tendency to equalization ofper-capita GDPs.2 The assessment of this empirical tendency represents a matter of primaryrelevance for policy makers (Islam, 2003). According to the classification originally proposedby Galor (1996), three different definitions of economic convergence can be identified: absoluteconvergence, conditional convergence, and convergence clubs. Absolute convergence is reachedwhen all economies converge toward the same steady-state (in terms of per-capita GDP growthrates). However, the steady-state may depend on features specific to each economy, in which caseconvergence will still take place, but not necessarily at the same levels. This is the case whenper-capita GDP depends on a series of determinants such as, for example, factor endowment orinstitutions, which can vary from one economy to another even in the long run. Convergence isthen said to be conditional. Finally, the concept of convergence clubs is linked to the existenceof multiple, locally stable, steady-state equilibrium to which economies with similar characteris-tics converge (Durlauf & Johnson, 1995). Recently, the interest of empirical researcher focuseson the investigation of the phenomenon at the regional level, namely the analysis of economicconvergence on intra-national scales. More recent studies introduce a spatial dimension into theformulation of the problem, see, for instance Rey and Montouri (1999) for an introduction ofthe problem, and Postiglione, Benedetti, and Lafratta (2010) and Postiglione, Andreano, andBenedetti (in press) for some very recent contributions to the debate.

Alternative definitions, as those based on the concept of stochastic convergence, have also beenintroduced in the literature (Evans & Karras, 1996). To overcome some problems linked with theanalysis of economic convergence, such as endogeneity, heterogeneity, and omitted variables,other techniques, like panel data (Islam, 1995; Laureti & Postiglione, 2005), and probabilitytransition matrices (Quah, 1997), have often been used.

The purpose of the present paper is to analyze the economic growth in the Middle East and NorthAfrica (MENA) countries, with particular emphasis on the convergence process in terms of long-term trend of per-capita GDPs. The economy of the region has been heavily influenced by peculiar

1 Per-capita GDP based on purchasing power parity is gross domestic product converted to international dollars usingpurchasing power parity rates. An international dollar has the same purchasing power over GDP as the U.S. dollar has inthe United States. For an application to Mediterranean countries see Laureti (2001).

2 See Laureti (2008) for an analysis of economic convergence in Mediterranean countries.

M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683 671

factors, such as energy sources, and demographic and institutional characteristics. Furthermore,this thematic paper aims to evidence the specific determinants of the growth process.

The term MENA covers a wide geographical area, extending from Morocco to Iran, includingthe majority of both the Middle Eastern and Maghreb countries. Following the definition ofWorld Bank, the MENA is: an economically heterogeneous region that includes both the oil-richeconomies in the Gulf and countries that are resource-scarce in relation to population, such asEgypt, Morocco, and Yemen. For the complete description of the countries in our sample data seeSection 3.3

Over the last fifteen years, the growth performance of the MENA region as a whole, despite itsnatural resources richness, has been unsatisfactory and not in line with other developing countries.In comparison with other regions in the world, growth rates in the MENA countries have beenremarkably volatile and lower than that of the poor-performing regions such as Sub-SaharanAfrica (SSA). This volatility is only partly due to political and social instability, to the wars or tothe marked fluctuations in oil prices that have characterized the history over the last century.

Besides, the area is subject to a peculiar process of development, which probably has no equalsworldwide. An example is the significant inter-regional migration flows, the consistent populationgrowth, the policy mis-management, and, finally, the strong interdependence between politics onone side, and the economic and social spheres on the other side.

Only a few empirical studies have dealt with the MENA region, largely due to lack of data.Abu-Qarn and Abu-Bader (2007) analyzed a period ranging from 1960 to 1998, and observed thataccumulation of capital seems to be the major determinant of economic growth. Adams and Page(2003) used cross-country data to analyze trends in poverty, inequality, and economic growth inthe MENA region. The analysis showed that international migration/remittances and public sectoremployment had a statistically significant impact on reducing the level and depth of poverty inthe MENA region. Ben Naceur and Ghazouani (2007) highlighted the idea of not significant rela-tionship between banking and stock market development and growth. Arouri, Youssef, M’henni,and Rault (2012) investigated the relationship between carbon dioxide emissions, energy con-sumption, and real GDP for 12 MENA countries over the period 1981–2005. Finally, Guetatand Serranito (2007) tested the convergence hypothesis in the MENA region using unit roots inpanel data following Evans and Karras (1996) methodology. They concluded that the conditionalconvergence is not rejected for the majority of the MENA countries.

In a previous paper, Andreano and Savio (2012) analyzed the absolute β-convergence forthe MENA countries. They observed that the large and heterogeneous region of MENA hasexperienced over the last sixty years a process of weak not significant convergence, with apparentinequalities in recent period. The consequent social and political tensions still constitute an elementof risk for the region, as evidenced by the recent events. In other words, the hypothesis of absoluteβ-convergence does not seem to be confirmed by the data.

Compared to other studies, our analysis uses an enlarged definition of the MENA countriesand covers a more extensive sample period ranging from 1950 to 2007.

In the light of the recent events that characterized the Arabic spring, we re-examines theSummer-Heston data set in order to identify the conditioning factors of the growth in the area.The aim of the paper is to achieve a better understanding of the heterogeneous, interrelatedcharacteristics of the growth, ranging from economic and social, to demographic and governancefactors. The objective is to provide a proper interpretation of how these aspects, interacting with

3 Note that often the studies on the MENA countries are added and overlap with those of the Mediterranean countries.

672 M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683

each other and at the same time with the initial conditions of development, have led to a given pathof growth in the area. These main determinants are analyzed in terms of conditional β-convergence.

The layout of the paper is the following. Section 2 is devoted to a review of the main conceptsof β-convergence. A description of data and some statistical measure are given in Section 3.Furthermore, Section 3 presents the empirical analysis of conditional β-convergence approachfor our data sample. Finally, Section 4 concludes the paper and outlines the future research agenda.

2. A review of economic convergence approach

The economic convergence in per-capita GDPs across countries has been largely analyzedfrom both the theoretical and the empirical point of view (Barro & Sala-i-Martin, 1992). Themost popular approaches in the quantitative measurement of convergence are those based on theconcepts of �-convergence and β-convergence (Barro & Sala-i-Martin, 1995).

The σ-convergence approach is based on the study of the time trend of the variance of thelogarithms of per-capita GDP. If there is a decreasing long-term trend, then countries converge toa common growth rate, and so �-convergence is satisfied. This approach is not justified by anyeconomic theory and, furthermore, the variance of logarithms is insensible to permutations, it doesnot allow discriminating between different geographical situations. However, the σ-convergenceis a widely used measure. For example, for some application in European Union see Monfort(2008).

So, the β-convergence approach has been considered the more convincing under the theoret-ical viewpoint, as well as the more appealing, since it leads to a quantification of the speed ofconvergence. The concept of β-convergence is directly related to the neoclassical Solow-Swanexogenous growth theory (Solow, 1956; Swan, 1956), assuming exogenous saving rates and aproduction function based on decreasing productivity of capital and constant returns. Accordingto this model we can write that:

Y (t) = F (K(t); L(t))

∂k

∂t= sf (k) − pk

(1)

where Y(t) is the total production at time t, F(.) is a production function, homogenous of degreeone, K is the stock of physical capital, L is the labor force, k is the per-capita capital, ∂k/∂t is thederivative of k with respect to time t, s is constant saving rate, f(k) is the per-capita production,and p is the population’s growth rate.

On this basis Barro and Sala-i-Martin (1992), in order to measure absolute β-convergence,suggest the following statistical model:

gi = α + βqi + εi (2)

where gi is the average growth rate of per-capita GDP across the time period under investiga-tion, α is the intercept, qi = ln(yi0) is the natural logarithm of the initial level of per-capita GDP,β = − (1 − e−λT)/T, λ is the speed of convergence which measures how fast economies will con-verge toward the steady state,4 T is the spanned time interval, and εi is the error term and is theerror term which is assumed to be normally distributed (0, σ2

ε ). In a cross-section of economies,

4 λ = −ln(Tβ+1)

T.

M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683 673

there is absolute β-convergence if we find a negative relation between the growth rate of per-capitaGDP and its initial level, i.e. if β is negative and statistically different from 0. Note that throughthe absolute approach, the per-capita GDP growth rate is described as a function of only onecovariate (i.e. the natural logarithm of the initial level of per-capita GDP).

The use of absolute β-convergence approach does not seem to be realistic, since it is implausibleto consider different economies identical for some structural characteristics as, for example: savingrates, population growth rates, preferences (Sala-i-Martin, 1996). The idea that economic growthis a composite function of a great number of interrelated factors has led some economists todevelop the idea of conditional economic convergence. The conditional approach is coherentwith the neoclassical framework, but it concerns the tendency of a cross-section of countries toconverge to their own steady states as a function of a number of conditioning variables: in thiscase, economies are considered different in their structural features. Conditional convergence isestimated on the basis of a multivariate regression analysis where per- capita GDP growth rateis measured with respect to the natural logarithm of the initial level of per-capita GDP and a setof other explanatory variables. In this case, the statistical model for the analysis of conditionalβ-convergence for each economy i is expressed as:

gi = αi + βqi + π1xi1 + . . . + πhxih + εi (4)

where x1, . . ., xh are some exogenous variables.A satisfactory conditional β-convergence model strongly depends on the choice of an appro-

priate list of conditioning variables. In this paper, we use two systems of explicative variables. Thefirst set is defined as state variables, and mainly affects the structure of an economic system. Inthis class we can include the stock of physical and human capital, measured in terms of educationand health level. The second group is outlined as control and environmental variables, and isgenerally influenced by specific factors that are managed by government or by private economicagents. The most commonly used are: government consumption (as share of GDP), private and/orpublic investment (as share of GDP), the degree of openness, fertility and migration rates, politicalinstability and security, and the tax system.The physical capital data are not always available, andwhen available are of poor reliability, because they depend on arbitrary assumptions about thedepreciation rate, the stock of initial capital, and the subsequent investment flows. As an alterna-tive to the use of physical capital, we assume that, for a given education level, a higher level ofinitial per-capita GDP reflects a somewhat higher capital stock.5 In this way, the effect of physicalcapital can be incorporated into the coefficient of β-convergence. Following these assumptionsthe convergence Eq. (4) can generally be better re-specified as:

gi = F (qi, hi, Ωi) (5)

where qi is the initial level of per-capita GDP, hi is the per-capita human capital measured in termsof education and health level, and Ωi is the vector of control and environmental covariates.

Finally, it is worth noticing that, although the concepts of σ and β-convergence are not identical,it is possible to demonstrate that the two formulation of convergence are linked each other. Inparticular, we can say that β-convergence is a necessary condition for obtaining �-convergence,in other words if there is no β-convergence there cannot be σ-convergence (Sala-i-Martin, 1996).

5 This result was also obtained by Leontief in the well-known “Leontief Paradox”. See Salvatore (2010) for details.

674 M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683

Table 1The comparative growth process of the MENA countries.

Growing countries Not growing countries

Country Growth GDP’50 (ranking) Country Growth GDP’50 (ranking)

Oman 0.049 7383 (17) Bahrain 0.015 9377 (3)Tunisia 0.032 7451 (14) Algeria 0.014 7961 (12)Libya 0.030 8165 (9) Iraq 0.014 7670 (13)UAE 0.029 9223 (5) Sudan 0.012 7035 (19)Israel 0.028 8536 (8) Comoros 0.009 6933 (20)Turkey 0.027 7423 (15) Jordan 0.008 8070 (10)Egypt 0.027 7149 (18) Eritrea 0.006 6025 (25)Yemen 0.024 5655 (26) Qatar 0.004 11,186 (2)Palestine 0.024 7916 (21) Lebanon −0.007 9349 (4)Iran 0.022 7986 (11) Djibouti −0.008 8789 (6)Mauritania 0.022 6496 (24 Somalia −0.011 6752 (23)Syria 0.022 6764 (22) Kuwait −0.014 11,428 (1)Morocco 0.021 7390 (16)Saudi A. 0.020 8774 (7)

3. The empirical analysis

There is not a commonly accepted definition of the MENA area. The definitions of the Leagueof Arab States (LAS) and the Economic and Social Commission for Western Asia of the UnitedNations (UN-ESCWA) coincide and include 22 member states. The definition of the World Bankexcludes the Comoros Islands, Mauritania, Somalia, and Sudan, but includes Iran, Israel andMalta, for a total of 21 countries. Other authors refer to more distinctly operational definitions.For instance, Nugent and Pesaran (2007) follow an enlarged definition, including 25 countriesand excluding the Comoros Islands and Malta. The definition used here is operationally extendedto include more observations, but excludes Malta for obvious reasons of geography, culture, andlanguage. In this paper the 26 MENA countries are: Mauritania, Sudan, Djibouti, Somalia, Yemen,Comoros Island, Egypt, Lebanon, Syria, Palestine, Jordan, Israel, Iraq, Morocco, Algeria, Tunisia,Libya, Saudi Arabia, Kuwait, Bahrain, Oman, United Arab Emirates (UAE), Qatar, Turkey, Iran,and Eritrea.

One of the main stylized facts of the area highlighted by several authors is the exceptionalgrowth experienced in the period 1950–1980, stimulated by the exploration, discovery, and pro-duction of gas and oil. The growth rate significantly reduced in subsequent years, in particularduring the period 1980–1995, when the area experienced a relative stagnation.

Data used in this paper, per-capita GDP6 expressed in PPP (Heston, Summers, & Aten, 2009),show a global average rate of growth of 0.81% for the whole sample, while in the years 1950–1980is around 1.75%.

Table 1 shows for each country the average growth in the period under investigation andthe starting level of GDP in 1950. The ranking position of each country in 1950 is reported inparenthesis. The countries are divided into two groups: the first (referred to as growing countries,on the left of Table 1) includes zones with the growth over the average (equal to 0.016) and the

6 The data for West Bank and Gaza are from Maddison (2006), updated to 2007 with information available from thePalestinian Central Bureau of Statistics (PCBS).

M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683 675

Table 2Estimation of the absolute �- convergence model.

Variable Estimates (p-value)

Constant 0.0472(0.00)

ln(GDP1950) −0.0045(0.11)

λ (speed of convergence) 0.521%

Statistics or Tests

AIC −5.64HQa −5.92R2 0.150Jarque–Bera Test 1.313(p-value) (0.52)Breusch-Pagan Test 0.019(p-value) (0.98)

a Akaike information criteria (AIC) and Hannan-Quinn criteria (HQ) are criteria for model selection. The aim is to findthe model with the lowest value of the selected information criterion.

second (referred to as not growing countries, on the right of Table 1) constituted by the otherregions. The heterogeneity of the growth rates is evident from the statistics reported in Table 1.Some of the least developed countries (as Somalia, Eritrea, Comoros Islands, and Sudan) areconfined to the bottom in the rankings in 1950, and showed very low, even negative, growthrates in the period; while others countries starting from the top (as Libya, UAE, Israel, and Iran)continued to have growth rates above the average. Egypt, Yemen, Palestine, Mauritania, Syria, and,on the opposite side, most countries of the Gulf, Lebanon, Jordan, and Djibouti, showed instead adevelopment in line with the hypothesis of absolute convergence.Table 1 also emphasizes that theclassification among rich and poor countries, implicitly assumed in the neoclassical approach, isnot confirmed by our sample data, and a more realistic classification between globalized and notglobalized countries might be better considered (Bhandari & Heshmati, 2007).

The hypothesis of absolute β-convergence fails badly for our cross-country data. Table 2summarizes the main estimation results of the absolute β-convergence model for the 26 MENAcountries in the period under investigation.

The hypothesis of absolute convergence can not be accepted, since the β coefficient has a low,not significant value (i.e. β = −0.0045), R2 is only 0.15, with a half-life7 of 150 years. Note thatthe Jarque–Bera (JB) and Breusch–Pagan (BP) tests do not reject the assumptions of normalityand homoscedasticity of the errors, respectively.

The idea that economic growth is a composite function of a great number of interrelatedfactors was highlighted in Section 2, and, consequently, here we try to investigate the role of somecovariates in the identification of β-conditional convergence in the area. Table 3 describes thevariables used in the present paper.

The variables considered in our economic model are consistent with important empirical studieson growth (Barro & Sala-i-Martin, 1995; Caselli, Esquivel, & Lefort, 1996, among others).

7 The half-life is defined as: the time it takes for the gap between steady state (potential level of per-capita GDP) andper-capita GDP gap to decrease by half.

676 M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683

Table 3Covariates and their determination.

Covariate Description Covariate Description

stockcapa ln(y0) openness (export +import)/GDPlifeexp Life expectation at birth spillover Average of GDP of neighbors

countriesprimed Primary completion rate natresources Value added of agriculture and

mining sector on the total valueadded

seconded Progression to secondary school inflat Inflation ratefertility Fertility rate volatility Standard deviation of natural

logarithm of per-capita GDPmigration Migration rate termsoftrade Export price index/import price indexscale Population between 15 and 65 years patents Number of patents applicationskg Public expenditure/GDP wgi 1st indicator of governanceki Investment/GDP fwi 2nd indicator of governance

a Note that the name of the variables will be denoted in italic.

The conditioning covariates used in this paper are classified into two groups: state and envi-ronmental and control variables (see Section 1).

The state variables are essentially: the natural logarithm of the starting value of per-capitaGDP (ln(y0)) and some measures of human capital.

Remember that here ln(y0) is also considered as proxy of capital stock. Problems concerningthe availability and the reliability of capital stock data in the MENA countries, the assumptionsabout depreciation, the definition of initial capital stock, and the determination of investmentflows in the analyzed period are all problems that suggest the use of a proxy for the capital stock.Following the consideration that, for a given educational level, higher initial per-capita GDPvalues are related to higher capital stock values, the effect of capital stock can be incorporatedinto the β-convergence coefficient (Barro & Sala-i-Martin, 1995).

Human capital is measured through lifeexp, primed, and seconded. These three variables mainlyaffect the basic structure of the economic system, and should all have a positive effect on growth,since they reflect better life conditions. For some interesting remarks about the impact of humancapital on economic convergence see Cohen and Soto (2007).

The environmental and control variables describe the economic, social, and politic character-istics of a country. The reasons for the inclusion of these covariates in our model are clarified inthe following.

The demographic covariates (fertility, migration, and scale) are often considered in the eco-nomic growth models (Prskawetz & Lindh, 2007). The dimension of the population, expressedhere through the scale variable, has a positive impact on the growth, as, for example, the costsof innovations, the new technologies, and the introduction of new products are minor in case oflarge economies. The scale variable can also be seen as measure of international openness, aslarger countries tend to be less open because internal trade offers a large market that can substituteeffectively international trade (Barro & Sala-i-Martin, 1995).

The idea that openness, kg, and ki are among the most important determinants of economicgrowth (see Laureti & Postiglione, 2005; Irmen & Kuehnel, 2009; Yanikkaya, 2003; among others)leads our choice to include these variables as covariates of the β-convergence model. Note thatopenness, kg, and ki are expressed in terms of a percentage of the GDP and ki is the empiricalmeasure of the effect of the saving rate in the neoclassical growth model.

M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683 677

The geographic distribution of world income is not uniform. On the other side, we observethat wealthy countries tend to be located close to other wealthy countries and that fast growingeconomies are also geographically clustered. Furthermore, Chua (1993) found that a countrycould benefit from an increase in production of neighboring countries: improvements in the levelof technological knowledge, managerial talent, and capital flows have a positive impact on thegrowth. In order to take into account these effects we include a covariate that measures spilloveramong neighboring countries. In this case spillover is computed as a weighted mean of GDP ofthe neighboring countries, with weights equal to the populations. Kremer (1993) proposed theworld population as scale effect of spillover; however if the transmission of ideas is stronger inclose neighbors, the appropriate scale may be the population of neighboring countries.

To evaluate the impact of agriculture and mining sector on the economic growth of the MENAcountries we have defined the variable natresources as the ratio of the value added (VA) ofagriculture and mining sectors, on the total VA. For a deepening discussion on the potentialconnections among agriculture and economic growth see Mundlak (2000).

One of the fundamental objectives of macroeconomic policies is to sustain high economicgrowth together with low inflation. Inflation can lead to uncertainty about the future profitabilityof investment projects. This usually leads to more conservative investment strategies, ultimatelyleading to lower levels of economic growth. However, there has been considerable debate on thenature of the inflation and growth relationship (Temple, 2000). These considerations certain applyfor the MENA countries, and for these reasons we included inflat variable (i.e. the inflation rate)in our convergence model.

The MENA area is characterized by a high level of volatility, due to many factors, as oil pricefluctuations, climatic conditions, remittances and capital movements, political instability andregional conflicts. Therefore, its effect on the growth rate should be negative. Furthermore, Barroand Sala-i-Martin (1995) highlighted that the cross-sectional dispersion of ln(GDP) is sensitiveto shocks that have a common influence on sub-groups of countries or regions. The omission ofsuch shocks from the regression will tend to bias the estimates of β. Therefore, they introduceda random variable that represents the economy-wide disturbance and volatility in the economicconvergence model. We measure the volatility through the standard deviation of natural logarithmof per-capita GDP.

Besides, it is very important to investigate whether the economic growth in the MENA countriescan be also ascribed to improving terms of trade, which is assumed to change the competitiveenvironments of different countries (Singh, 2010). Therefore, in this study the terms of trade isused as an explanatory variable to see whether the countries that absorb more foreign trade havegreater economic performance than those trading less. Movements in the terms of trade dependprimarily on world conditions and would therefore be largely exogenous with respect to economicgrowth of individual countries.

It is often believed that inventions and innovations play a significant role in economic growth.What accurately measures inventions and innovations is a matter of debate, however, most stud-ies use patent data (Schneider, 2005). We follow this approach, and use the number of patentsapplications (i.e. patents variable) in our model.

Finally, two synthetic indicators of governance, wgi and fwi are computed. wgi is estimatedthrough an average of six different indicators derived from the World Bank: the effectivenessof institutions, the reliability of institutions, the political stability, the regulations quality, thenormative quality, and the corruption control. The governance indicators reflect the statisticalcompilation of responses on the quality of governance given by a large number of enterprise,citizen, and expert survey respondents in industrial and developing countries, as reported by a

678 M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683

Table 4Conditional covariates and their statistics.

Covariate Mean Dev. St. Corr. Covariate Mean Dev. St. Corr.

lifeexp 3.8 0.2 −0.01 openness 82.1 34.4 0.06primed 66.8 20.8 0.33 spillover 7.7 0.6 0.09seconded 83.5 14.0 0.21 natresources 0.5 0.1 0.06fertility 1.9 0.1 −0.09 inflat 2.1 2.1 −0.37migration 3.5 10.7 −0.25 volatility 0.4 0.2 0.57scale 6.6 1.9 0.31 termsoftrade 1.0 0.1 −0.17kg 21.2 13.7 −0.18 patents 8569.9 1547.0 0.45ki 17.4 8.6 0.15 wgi −0.5 0.7 0.27

fiw 5.2 1.2 0.17

Table 5Groups of the MENA countries.

Group Countries

Less Developed countries Mauritania, Sudan, Djibouti, Somalia, Yemen, Comoros IslandMashreq Egypt, Lebanon, Syria, Palestine, Jordan, Israel, IraqMaghreb Morocco, Algeria, Tunisia, LibyaGulf countries Saudi Arabia, Kuwait, Bahrain, Oman, UAE, QatarOther Turkey, Iran, Eritrea

number of survey institutes, non-governmental organizations, and international organizations. Onthe other hand, fwi is calculated through an average of two indicators about the political rightand the civil liberties levels. The expected signs of these variables are clearly positive: improvedgovernance fosters a higher average rate of development in the country. For a discussion of theuse of democracy, corruption, and rule of law variables see Mauro (1995).

Note that all data employed in the paper are from the United Nations and the World Bank,except fwi that is taken from Freedom in the World8 and the GDPs from Heston et al. (2009).

The basic statistics of the potential explanatory variables in terms of average growth rate overthe entire sample, and their correlation with the dependent variable are presented in Table 4. Thecorrelations have overall the expected signs, except for the volatility of GDP.

Public expenditure represents a key variable in the model: higher levels of kg lead to a lowerlevel of steady state, therefore to a lower rate of growth. The variable can be also seen as a proxyof political corruption and mis-managed administration, and may reflect the negative effects ofnon-productive expenditure and taxation. This is confirmed by the positive relationship betweenthe growth rate and the indicators of governance (e.g. a higher quality of governance raises thevalue of competition and this determines an increase in the average rate of development).

Structural differences in the MENA countries could also suggest the use of spatial dummies toidentify geographically distinct behaviors. The 26 MENA countries have therefore been dividedinto five groups according to their economic, social, religious, and territorial characteristics (seeTable 5). In the assignment of the countries through the five different groups we follow the defini-tion given by the United Nations (UN-ESCWA).9 The possibility that the growth is consequence of

8 Freedom in the World is an American organization established in 1973 that ranks every nation according to theirpolitical freedoms and civil rights.

9 Egypt is indifferently classified as Maghreb or Mashreg country.

M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683 679

Table 6Estimation of conditional �-convergence model.

Variable Estimates(p-value)

Constant 0.062609(0.00)

ln(GDP1950) −0.010(0.00)

primed 0.000238(0.00)

kg −0.000228(0.05)

openness 8.8e−05

(0.00)patents 1.7e−07

(0.08)volatility 0.022609

(0.00)wgi 0.006991

(0.00)λ (speed of convergence) 1.495%

Statistics or Test

AIC −10.0607HQ −9.94924R2 0.886F(7, 18) for significance of the model 19.97(p-value) (0.00)Jarque–Bera test 3.3626(p-value) (0.19)Breusch–Pagan test 0.6843(p-value) (0.73)

being located in a particular region will be investigated in the following by inserting four dummyvariables in the basic model and checking the significance of this choice through an appropriatetest (see Table 7).

The identification and estimation of the final model is obtained through a general to specificapproach (Hendry, 1995) that imposes F-tests zero-restrictions on the parameters. This proce-dure involves the formulation of a general unrestricted model consistent with the data, and theapplication of a testing down process, eliminating variables with coefficients that are not statisti-cally significant, leading to a simpler specific congruent model that involves competing models.However, the number of variables is too large with respect to the number of countries (e.g. obser-vations) and the all-inclusive regression is therefore not computationally feasible. We thereforeconsider two different starting models with combinations of similar conditioning variables andapply on each one the reduction procedure.The estimates of the final resulting model are givenin Table 6, which provides the inclusion of six conditioning variables and a constant term. Thevariables are: primed, kg, volatility, patents, openness, and wgi.

All the parameters of the final equation, with the exclusion of the volatility, have the expectedsign, and are statistically significant at least at a 8% level. The R2 is close to 0.90 (i.e. 0.8806),and the effect of the six conditioning variables on the goodness of fit appears highly relevant. The

680 M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683

Table 7Tests on the single variables.

Variable t-Value F(1; 17) p-Value

ki 0.605 0.366 0.55inflat −0.612 0.374 0.55spillover 0.163 0.026 0.87scale 0.101 0.010 0.92fertility −0.705 0.497 0.49lifeexp 0.389 0.151 0.70migration −1.13 1.271 0.27fiw 0.327 0.107 0.75seconded −0.603 0.364 0.55natresource 0.170 0.029 0.87termsoftrade 0.603 0.364 0.55spatial dummies −0.956 0.507a 0.73

−0.616−1.110−0.341

a Restrictions on spatial dummies are considered globally and therefore the test is F(4; 14).

improvement of the model specification is also confirmed by the information criterions AIC andHQ, which decrease from −5.64 and −5.92 to −10.06 and −9.95 respectively. Finally, also forthe augmented model, the non-normality and the heteroscedasticity does not seem to be supportedby data (i.e. JB and BP tests are not significant).

The lack of a significant relationship between termsoftrade and the growth is not a new elementin the literature. Gamo et al. (1997) and Makdisi, Fattah, and Liman (2007) observed the sameresult in many MENA countries. Besides, results show also that ki has no influence on long-term growth, notwithstanding the high level of investment recorded in the period. Many authors(Ben Naceur & Ghazouani, 2007; O’Toole & Tarp, 2012) noted, however, that the area suffersfrom an endemic lack of capital efficiency, due partly to the presence of public projects, devotedto low productivity investments. This is further confirmed by a negative relationship betweenthe growth rate and kg. Protectionism and lack of integration in international market contributeto restrain competitiveness and efficiency. The low capital efficiency can also be explained bythe low expansion of infrastructures and the absence of political and institutional support toprivate business. This interpretation seems to be confirmed by the final estimated model, wheregovernance variables and R&D (i.e. patents) are significant.

In the estimation and the choice of the variables, we evidenced that the inclusion of all potentialvariables in one regression was not practicable. However we are aware that the significance of onevariable can depend on which other variables are included in the regression model. To answer thequestion which variables are really correlated with the growth, the robustness of the final estimatedmodel should be checked. Levine and Renelt (1992) and Barro and Sala-i-Martin (1995) proposedifferent procedures. One simple and heuristic robustness analysis is to re-introduce one by onein the final specification the variables excluded, to verify the validity of their previous exclusion.In our case, the t-test and F-test (see Table 7) confirm the validity and the robustness of the finalestimated equation and the non-significance of all the omitted variables.

The spatial dummies are all not significant, suggesting that the growth does not have a territorialcharacteristic, and that the hypothesis of convergence clubs related to the possibility of multiplesteady-state equilibrium does not seem to be supported by data (see Durlauf & Johnson, 1995;Postiglione, Benedetti, et al., 2010; Postiglione, Andreano, et al., in press). However, further

M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683 681

analysis is needed in order to confirm the results obtained. The high significance of the con-stant term suggests that technical progress, assumed exogenous in the model, is not completelyexplained in this framework.

Some doubts remain on the unexpected positive relationship between volatility and the growthrate. However, this could be explained through the Galton’s fallacy: the observation that heightsin a family tend to regress toward the mean across generations (a property analogous to ourconvergence concept for per-capita GDP) does not imply that the dispersion of heights across thefull population tends to reduce over time. In the same way, the volatility of economic and socialphenomena can be positively related to its mean (Quah, 1997). However, other proxy variablescould be considered in the analysis to better understand the volatility of the MENA area andcapture the different shocks sources.

The coefficient of convergence, despite the introduction of variables directly related to humancapital and technological progress, seems to indicate a process of catching-up that develops in arelatively long time interval: it takes about 50 years to reduce by half the initial gap among thecountries. Our estimated model confirms the hypothesis of conditional convergence, though theprocess of convergence is slow comparing to other developing countries.

4. Concluding remarks

Globalization is changing the political and economic world with different effects in differentregions. The gap between rich and poor countries has widely changed over the past two decades,and long-term convergence can be generally verified in empirical analyses.

This paper attempts to empirically answer the question of whether there is convergence in percapita output across MENA countries. The empirical analysis of the natural logarithm of per-capita GDPs for 26 MENA countries over the last 60 years strongly confirms the hypothesis ofconditional convergence.

The analysis enabled us to identify the main variables on which a careful and prudent pol-icy intervention at regional level should be based. In fact, the long-term growth in this highlyheterogeneous area is the result of a set of socio-economic, technological, and governance factors.

In our empirical analysis, the degree of international openness and the government interven-tion and expenditure are important economic control variables. The improvement of governancefactors, such as actions to reduce corruption, the greater reliability and efficiency of government,political stability and violence reduction, play a role in stimulating the long-run behavior and mov-ing up the development path of the steady-state. Technological development and human capitalare both highly relevant for the growth.

One important extension of our work concerns the analysis of the impact of the space dimen-sion on the convergence process, here verified only with ex-ante identified dummies for the fivegeographic areas. In this respect, a more in-depth analysis is needed in order to confirm theseresults.

Given the prominent role of technological progress in the conditional convergence analysis,different proxy might be used in the empirical analysis, in order to better evaluate its conditioningimpact on the growth.

References

Abramovitz, M. (1986). Catching up, forging ahead and falling behind. The Journal of Economic History, 46(2),385–406.

682 M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683

Abu-Qarn, A. S., & Abu-Bader, S. (2007). Sources of growth revisited: Evidence from selected MENA countries. WorldDevelopment, 35(5), 752–771.

Adams, R. H., Jr., & Page, J. (2003). Poverty, inequality and growth in selected middle East and North Africa Countries,1980–2000. World Development, 31(12), 2027–2048.

Andreano, M. S., & Savio, G. (2012). La Convergenza Economica nei paesi del Medio Oriente e Nord Africa: FattoriCaratterizzanti e Peculiarità dell’Area. Rivista di Statistica ed Economia del Territorio, 2, 95–119 (in Italian).

Arouri, M. E. H., Youssef, A. B., M’henni, H., & Rault, C. (2012). Energy consumption, economic growth and CO2

emissions in Middle East and North African countries. Energy Policy, 45, 342–349.Barro, R. J., & Sala-i-Martin, X. (1992). Convergence. Journal of Political Economy, 100(2), 223–251.Barro, R. J., & Sala-i-Martin, X. (1995). Economic growth. New York: McGraw-Hill.Ben Naceur, S., & Ghazouani, S. (2007). Stock markets, banks, and economic growth: Empirical evidence from the MENA

region. Research in International Business and Finance, 21, 297–315.Bhandari, A. K., & Heshmati, A. (2007). Measurement of globalization and its variations among countries, regions and

over time. In A. Heshmati, & A. Tausch (Eds.), Roadmap to Bangalore? Globalization, the EU’s Lisbon process andthe structures of global inequality. Nova Science Publishers.

Caselli, F., Esquivel, G., & Lefort, F. (1996). Reopening the convergence debate: A new look at cross country growthempirics. Journal of Economic Growth, 1(3), 363–390.

Chua, H. B. (1993). Regional spillovers and economic growth. Center Discussion Paper No. 700, Economic GrowthCenter, Yale University.

Cohen, D., & Soto, M. (2007). Growth and human capital: Good data, good results. Journal of Economic Growth, 12(1),51–76.

Durlauf, S. N., & Johnson, P. A. (1995). Multiple regimes and cross-country behaviour. Journal of Applied Econometrics,10, 365–384.

Evans, P., & Karras, G. (1996). Do economies converge? Evidence from a panel of U. S. states. The Review of Economicsand Statistics, 384–388.

Galor, O. (1996). Convergence? Inference from theoretical models. The Economic Journal, 106, 1056–1069.Gamo, P. A., Fedelino, A., & Horvitz, S. P. (1997). Globalization and growth prospects in Arab Countries. Working Paper

WP/79/127, International Monetary Fund.Guetat, I., & Serranito, F. (2007). Income convergence within the MENA countries: A panel unit root approach. The

Quarterly Review of Economics and Finance, 46, 685–706.Hendry, D. F. (1995). Dynamic Econometrics. Oxford: Oxford University Press.Heston, A., Summers, R., & Aten, B. (2009). Penn World Table Version 6.3. Center for International Comparisons of

Production, Income and Prices at the University of Pennsylvania, August 2009.Irmen, A., & Kuehnel, J. (2009). Productive government expenditure and economic growth. Journal of Economic Surveys,

23, 692–733.Islam, N. (1995). Growth empirics: A panel data approach. The Quarterly Journal of Economics, 110(4), 1127–1170.Islam, N. (2003). What have we learnt from the convergence debate? Journal of Economic Surveys, 17, 309–362.Kremer, M. (1993). Population growth and technological change: One Million B.C. to 1990. Quarterly Journal of

Economics, 108(3), 681–716.Laureti, L. (2001). Purchasing power parity and integration: An empirical evidence in the Mediterranean countries. Journal

of Policy Modeling, 23(7), 731–751.Laureti, L. (2008). Economia dello sviluppo e dell’integrazione Euromediterranea. Franco Angeli Editore (in Italian).Laureti, L., & Postiglione, P. (2005). The effects of capital inflows on the economic growth in the Med Area. Journal of

Policy Modeling, 27, 839–851.Levine, R., & Renelt, D. (1992). A sensitivity analysis of cross-country growth regressions. American Economic Review,

82(4), 942–963.Maddison, A. (2006). The world economy: Volume 1 – A millennial perspective and volume 2 – Historical statistics.

OECD Development Centre.Makdisi, S., Fattah, Z., & Liman, I. (2007). Determinants of economic growth in the MENA Countries, explaining growth

in the Middle East. In J. B. Nugent, & M. H. Pesaran (Eds.), Contributions to economic analysis, vol. 278 (pp. 31–60).Amsterdam: Elsevier.

Mauro, P. (1995). Corruption and growth. Quarterly Journal of Economics, 110(3), 681–712.Monfort, P. (2008). Convergence of EU regions. Measures and evolution. Working Paper n. 1, Directorate-General for

Regional Policy.Mundlak, Y. (Ed.). (2000). Agriculture and economic growth: Theory and measurement. Cambridge, MA: Harvard

University Press.

M.S. Andreano et al. / Journal of Policy Modeling 35 (2013) 669–683 683

Nordhaus, W. D. (2003). The health of nations: The contribution of improved health to living standards. In K. M. Murphy,& R. H. Topel (Eds.), Measuring the gains from medical research: An economic approach. The University of ChicagoPress, 1999.

Nugent, J. B., & Pesaran, M. H. (2007). Introduction: Explaining growth in the Middle East, in explaining growth inthe Middle East. In J. B. Nugent, & M. H. Pesaran (Eds.), Contributions to economic analysis, vol. 278 (pp. 1–27).Amsterdam: Elsevier.

O’Toole, C. M., & Tarp, F. (2012). Corruption and efficiency of capital investment in developing countries. Working Paper27, United Nations University WIDER.

Postiglione, P., Andreano, M. S., & Benedetti, R. Using constrained optimization for the identification of convergenceclubs. Computational Economics, Springer. ISSN: 0927-7099, http://dx.doi.org/10.1007/s10614-012-9325-z, in press.

Postiglione, P., Benedetti, R., & Lafratta G. (2010). A regression tree algorithm for the identification of convergence clubs.Computational Statistics and Data Analysis, 54(11), 2776–2785.

Prskawetz, A., & Lindh, T. (Eds.). (2007). The Relationship between demographic change and economic growth in theEU. Research Report 32, Vienna Institute of Demography and Institute for Futures Studies.

Quah, D. (1997). Empirics for growth and distribution: Stratification, polarization, and convergence clubs. Journal ofEconomic Growth, 2, 27–59.

Rey, S. J., & Montouri, U. S. (1999). Regional income convergence: A spatial econometric perspective. Regional Studies,33(2), 143–156.

Sala-i-Martin, X. (1996). Regional cohesion: Evidence and theories of regional growth and convergence. EuropeanEconomic Review, 40, 1325–1352.

Salvatore, D. (2010). International economics (10th ed.). Wiley.Schneider, P. H. (2005). International trade, economic growth and intellectual property rights: A panel data study of

developed and developing countries. Journal of Development Economics, 78, 529–547.Singh, T. (2010). Does international trade cause economic growth? A survey. World Economy, 33, 1517–1564.Solow, R. M. (1956). A contribution to the theory of economic growth. Quarterly Journal of Economics, 70, 65–94.Swan, T. (1956). Economic growth and capital accumulation. Economic Record, 32(2), 334–361.Temple, J. (2000). Inflation and growth: Stories short and tall. Journal of Economic Surveys, 14(4), 395–426.Yanikkaya, H. (2003). Trade openness and economic growth: A cross-country empirical investigation. Journal of Devel-

opment Economics, 72(1), 57–89.