Does Globalisation Promote Economic Output in Sub-Saharan Africa? Evidence from Bootstrap ARDL Model

Abstract

This study aims to investigate whether globalisation promotes economic output in Sub-Saharan African countries in both the short run and the long run. Based on the latest version of the KOF globalisation index, we employ a newly developed bootstrap autoregressive distributed lag model to analyse this question. Compared to the traditional autoregressive distributed lag model, which ignores the degenerate cases, the new approach could avoid spurious cointegration. Results show that globalisation and economic output are positively correlated for most Sub-Saharan African countries, while the causal effect cannot be concluded except for a couple of exceptions. This finding implies that globalisation cannot guarantee an increase in economic output in the long run for most Sub-Saharan African countries. The Granger causality test shows that globalisation leads to economic output for Burundi, Gabon, Rwanda, Senegal and Zambia in the short run. Conversely, economic output leads to globalisation for Burkina Faso, Cameroon, Ghana, Kenya and Senegal. For Senegal, globalisation and economic output mutually determine each other and therefore form a positive spiral development path. Policymakers should be aware of the specific features of different economies in making sound globalisation policies to avoid the underlying adverse effects of global integration.

Keywords

Globalisation economic output bootstrap autoregressive distributed lag Granger causality Sub-Saharan Africa

Introduction

Globalisation is believed to play an essential role in stimulating economic growth, particularly for developing and transition economies (Gurgul and Lach, 2014). Globalisation is a process whereby all artificial barriers hindering the free flow of goods, services and production factors in the global market are removed (Ayomitunde et al., 2020). The process of globalisation has various consequences for economic and financial variables, including trade openness, diversification of exports, and foreign direct investments (FDIs), and all these variables can significantly affect the level of economic output, particularly in poor economies (Buysse et al., 2018). We have to realise, however, that globalisation brings not only a chance, but also new challenges and risks, such as increased competition. As Stiglitz (2004) claimed, globalisation should be well managed. Otherwise, it will restrict job creation and lead to a risky economic system which may not spur economic output. Therefore, an investigation of the nexus between globalisation and economic output is necessary for developing countries to explore the benefits while minimising the downside risks of globalisation. Sub-Saharan African (SSA) countries offer an interesting research background for their relatively low economic status on average, and they are generally not well integrated into the global economy. Thus, SSA countries might be able to use globalisation-facilitating policies to increase economic output and vice versa.

China has experienced impressive growth in the past decades since its opening-up policy, which provides a model of development for other developing countries. Meanwhile, some countries in Africa are still experiencing poverty, inequality and terrorism (Asongu and Biekpe, 2018). It is a critical issue to figure out whether they can further improve their economic status through globalisation. Since the US 45th president Donald Trump came to power, trade protectionism and de-globalisation are emerging across the world. In contrast, China proposed the Belt and Road Initiative (BRI) for deepening its economic transition and seeking to transfer its massive equipment and technology overseas. The initiative aims to further integrate China into the global economy through trade, investment, infrastructure construction, connectivity and other development projects (Chen, 2016). Given the competing situation of the two giant economies, an investigation into the nexus between globalisation and economic output in African regions would be an instructive reference for Africa’s further development and also for the Chinese government in terms of the potential challenge in exploring cooperation with African countries. If globalisation indeed stimulates growth in African countries, the BRI would be an excellent chance for them to diversify their trade and benefit from FDI.

Studies generally use proxy variables, such as trade flows and FDI, to represent globalisation (Carkovic and Levine, 2005; Dollar and Kraay, 2004; Greenaway et al., 1999). It can be anticipated that, because of using different proxy variables, there is no consensus on the relationship between globalisation and economic output. In addition, Kacowicz (1999) points out that globalisation should be a complex intensification which consists of various aspects, including economic, social, political and cultural relations. In the seminal work by Dreher (2006a), he firstly quantified the globalisation from economic, social and political aspects and concluded that globalisation could promote economic output. Recently, Gygli et al. (2019) provided a revised version of the KOF Globalisation Index, which includes more variables and is more robust than other globalisation indices. Based on the first version of the KOF Globalisation Index, Dreher (2006b) employs a dynamic panel model to examine the impact of globalisation on taxes and policies. He finds that globalisation only affects capital taxation. Dreher also emphasises the role of social integration which could effectively influence social policies. He further points out that the endogeneity among globalisation and other economic indicators should be specified beforehand. Otherwise, the accuracy of the empirical results will be affected. Since then, the nexus between globalisation and other macroeconomic variables has been widely studied (Kandil et al., 2017; Shahbaz et al., 2018).

Dreher and Gaston (2008) visit the impact of globalisation on income inequality in Organisation for Economic Co-operation and Development (OECD) countries with the consideration of the synthetic index of globalisation and also its components, namely, economic, social and political globalisation. They find that globalisation would exacerbate income inequality in OECD countries. Dreher et al. (2008) shed new light on the relationship between globalisation and government expenditures with detailed categories through a panel data model. They conclude that globalisation does not affect government expenditure significantly. Chang and Lee (2010) reveal the significant unidirectional causality from globalisation to economic output exists in the long run, while only weak causality is confirmed in the short run. McMillan and Rodrik (2011) argue that globalisation promotes specialisation through increasing competition which drives production factors to more productive sectors. Gurgul and Lach (2014) investigate the relationship between globalisation and economic output and confirm the role of globalisation in stimulating economic output. Kandil et al. (2015) investigate the interaction between globalisation and financial development by using panel cointegration and causality analysis. They find that globalisation helps mobilise economic growth, but does not help financial development as it helps increase access to external financing. Potrafke (2015) reviews nearly all of the previous studies related to the impact of globalisation on macroeconomic indicators and finds that almost all of the relevant papers used panel data models to explore the relationship (Chang and Lee, 2010; Chang et al., 2011; Gurgul and Lach, 2014; Sakyi, 2011). Panel data models are more likely to support economic issues, while more countries and economic variables are considered. However, panel data models tend to ignore the differences across countries. For example, Aluko et al. (2020) examine the nexus between FDI and globalisation in Africa. They find evidence of unidirectional causality from social and political globalisation to FDI while, in the case of economic globalisation, the direction of causality moves from FDI. However, they also find that there are substantial variations in terms of the causal relations across countries, which necessitates the time-series analysis at the country level. Overall, the relationship between globalisation and economic output could be categorised by the following four hypotheses: (a) output hypothesis: globalisation causes economic output; (b) conversation hypothesis: economic output causes globalisation; (c) neutrality hypothesis: no causality between globalisation and economic output; (d) feedback hypothesis: bidirectional causality between globalisation and economic output.

This study adds to the literature in the following ways. First, we adopt a newly proposed bootstrap autoregressive distributed lag (ARDL) test for cointegration and causality (McNown et al., 2018). To the best of our knowledge, this study is the first to apply this method to investigate the nexus between globalisation and economic output in SSA countries. The new method is more powerful than the traditional ARDL model (Pesaran et al., 2001) in the following aspects:

With the new test on the lagged level(s) of the independent variable(s), we can better identify the cointegration, non-cointegration or a degenerate case.

The bootstrap ARDL test could eliminate inconclusive inferences with the bounds test.

The bootstrap ARDL model allows for the endogeneity and feedback that may exist among the variables.

A traditional Granger causality test could be implemented based on the model to identify the causal relationship further.

Second, we use the recently developed composite index of globalisation by Gygli et al. (2019) which is overtly more robust compared to prior indexes of globalisation (e.g. Dreher’s (2006) KOF Globalisation Index; the Maastricht Globalisation Index, and the DHL Global Connectedness Index, among others). Last but not least, equipped with the new approach and the most updated data, we investigate both the long-run and the short-run nexus between globalisation and economic output in the context of SSA which is rarely studied in the literature. The only exception, as far as we know, is Le (2016) who implements a panel cointegration test on SSA countries to examine the nexus between energy, output, openness and financial development. However, as stated before, panel data models tend to ignore the differences across countries. To the best of our knowledge, this study is the first paper to explore the relationship between globalisation and economic output in the context of SSA countries using time series analyses, which could provide more individual characteristics across SSA countries.

The rest of the paper is organised as follows: the second section introduces the ARDL test for cointegration and causality;. the third section presents the datasets; the fourth section reports the empirical results; and the final section concludes the paper.

Econometric methodology

A bi-variate ARDL model is considered as follows,

y_{t} = c + \sum_{i = 1}^{p_{i}^{y}} α'_{i} y_{t - i} + \sum_{j = 1}^{p_{j}^{x}} β'_{j} x_{t - j} + \sum_{s = 1}^{l} δ'_{s} D u m m y_{t, s} + u_{t}

(1)

where $y_{t}$ is the dependent variable and $x_{t}$ is the independent variable. $D u m m y_{t, s}$ is the $s_{t h}$ structural break which could be detected through the procedure of Bai and Perron (2003). $α_{i}$ , $β_{j}$ and $δ_{s}$ are coefficients of the lags of $y_{t}$ , $x_{t}$ . and the dummy variables, respectively. $u_{t}$ is the i.i.d. disturbance term with zero mean and a finite variance $σ^{2}$ . $i$ and $j$ are an index of lags $i = 1, 2, \dots, p_{i}^{y}$ ; $j = 1, 2, \dots, p_{j}^{x}$ , which are selected by the Akaike Information Criterion (AIC) in an unrestricted vector autoregressive (VAR) model suggested by Pesaran et al. (2001). Equation (1) could be re-parameterised as the following error correction model (ECM),

Δ y_{t} = c + φ' y_{t - 1} + ω' x_{t - 1} + \sum_{i = 1}^{p_{i}^{y} - 1} ρ'_{i} Δ y_{t - i} + \sum_{j = 1}^{p_{j}^{x} - 1} θ'_{j} Δ x_{t - j} + \sum_{s = 1}^{l} δ'_{s} D u m m y_{t, s} + ε_{t}

(2)

where $φ = - (1 - \sum_{i = 1}^{p_{i}^{y}} α_{i})$ and $ω = \sum_{j = 0}^{p_{j}^{x}} β_{j}$ , parameters $ρ_{i}$ and $θ_{j}$ could be obtained through the results of equation (1). Commonly, the coefficient $ω$ is used to describe the long-run impact of $x_{t}$ on $y_{t}$ , and the coefficients $θ_{j}$ . is used to capture the effect of $x_{t}$ on $y_{t}$ in the short-run. To verify the validity of estimations on equation (2), some diagnostic statistics are used. We use $L M_{s c}$ , $L M_{r r}$ and $L M_{h e t}$ to denote Lagrange multiplier (LM) tests for serial correlation, functional form and heteroscedasticity in equation (2), respectively. Pesaran et al. (2001) suggest identifying the cointegration by jointly using an F-test rejecting $H_{0} : φ = ω = 0$ and a t-test on the lagged dependent variable $y_{t - 1}$ rejecting $H_{0} : φ = 0$ . Goh et al. (2017a), McNown et al. (2018) and Sam et al. (2019) all suggest using an additional t-test on the lagged independent variable $x_{t - 1}$ , $H_{0} : ω = 0$ , to complement the existing F-test and t-test on dependent variables (Pesaran et al., 2001).

The F-test on all lagged variables and t-tests on both dependent and independent variables are employed to distinguish between cointegration, non-cointegration and degenerate cases (McNown et al., 2018; Pesaran et al., 2001; Sam et al., 2019). The degenerate case #1 occurs when the F-test on all lagged variables is significant, the t-test on the lagged independent variable is significant and the t-test on the lagged dependent variable is insignificant. The degenerate case #2 happens when the F-test on all lagged variables is significant, the t-test on the lagged dependent variable is significant and the t-test on a lagged independent variable is insignificant. In the study by Pesaran et al. (2001), they provided the critical value for the overall F-test and the t-test for the degenerate case #2, but they did not provide the critical value for the test on the degenerate case #1. Hence, the method proposed by Pesaran et al. (2001) does not allow testing on the degenerate case #1. In this case, to rule out the degenerate case #1, the integrated order for the dependent variable has to be I(1) process. However, traditional unit root tests like the augmented Dickey–Fuller (ADF) test and Phillips–Peron (PP) test are notorious for their lower testing power and neglecting the breaks (Cai et al., 2018a; Perron, 1989).¹

McNown et al. (2018) further develop the bootstrap ARDL test to solve these problems by providing an additional t-test on the degenerate case #1, which provides a better insight into the cointegration status of the model.² Since then, several studies on a diverse range of topics applied this approach to obtain a more reliable cointegration relationship. For example, Nawaz et al. (2019) explore the link between financial development and economic growth in Pakistan for the period 1972–2017. The bootstrapping ARDL bounds testing approach is employed to examine the cointegration between the factors of production. Ghazouani et al. (2020) empirically investigate the trade–energy–growth nexus in Asia-Pacific countries using the bootstrap ARDL approach. Goh and McNown (2020) apply the extension of ARDL model to identify the cointegration relationship between macroeconomic variables and demographic variables in Japan. Shahbaz et al. (2020) examine the cointegration between carbon emissions and its determinant by using the bootstrapping ARDL model to reveal the role of technological innovations in China. Caglar (2020) also applies the bootstrap ARDL approach to investigate the existence of cointegration between growth and emissions. Clearly, the most recent studies prefer to use this approach to obtain a more reliable cointegration relationship.

Moreover, the causal nexus running from one variable to the other could be tested by standard Granger causality tests (Cai et al., 2018b; Goh et al., 2017b). If long-run cointegration exists between $y_{t}$ and $x_{t}$ when $y_{t}$ serves as the dependent variable, the Granger causality test with the null that $x_{t}$ does not Granger cause $y_{t}$ (for simplicity $x_{t} \neq > y_{t}$ ) should include the lagged level and difference of $x_{t}$ .³ That is, we test the null hypothesis $H_{0} : \hat{ω}' > 0; \hat{θ}'_{j} = 0 (j = 1, 2, \dots, p_{j}^{x})$ . However, if cointegration does not exist between $y_{t}$ and $x_{t}$ when $y_{t}$ . serves as the dependent variable, then the Granger causality with the null that $x_{t}$ does not Granger cause $y_{t}$ (for simplicity $x_{t} \neq > y_{t}$ ) should only include lagged difference of $x_{t}$ . We test the null hypothesis $H_{0} : \hat{θ}'_{j} = 0 (j = 1, 2, \dots, p_{j}^{x})$ .

Empirical specifications and datasets

To investigate the cointegration and the causal nexus between economic output and globalisation, the bi-variate ARDL model could be specified as follows:

G D P_{t} = c + \sum_{i = 1}^{p_{i}^{y}} α'_{i} G D P_{t - i} + \sum_{j = 1}^{p_{j}^{x}} β'_{j} G L O_{t - j} + \sum_{s = 1}^{l} δ'_{s} D u m m y_{t, s} + u_{t}

(3)

G L O_{t} = c + \sum_{e = 1}^{p_{e}^{y}} α'_{e} G L O_{t - e} + \sum_{f = 1}^{p_{f}^{x}} β'_{f} G D P_{t - f} + \sum_{k = 1}^{l} δ'_{k} D u m m y_{t, k} + v_{t}

(4)

where we call equation (3) the gross domestic product (GDP) equation and equation (4) the globalisation (GLO) equation. The real GDP per capita and the level of globalisation serve as dependent variables in equations (3) and (4), respectively. Following Pesaran et al. (2001), we use the AIC criterion to select the optimal order. We use the multiple structural breaking test (Bai and Perron, 2003) to create the dummy variable for the structural breaking date in each equation. That is, the long-run equilibrium only exists when the F statistics on all variables and the t statistic on both the independent and the dependent variable are significant. As mentioned earlier, the model proposed by Pesaran et al. (2001) omits the degenerate case#1. We apply a bootstrap procedure with 5000 replications to generate critical values for all three tests.

The globalisation index could be withdrawn from the revised version of the KOF Globalization Index (Gygli et al., 2019).⁴ In this paper, we focus only on the synthetic globalisation index rather than its components. Because we need to estimate 2-way equations for 30 countries, respectively, we will have massive results and lose our focus if we further discuss the subdimensions of globalisation. We use real GDP per capita to represent the economic output. The SSA countries include 30 countries, namely, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Central African Republic, Chad, Democratic Republic of the Congo (Congo Dem. Rep.), Republic of the Congo (Congo Rep.), Cote d’Ivoire, Gabon, Gambia, Ghana, Kenya, Lesotho, Madagascar, Malawi, Mali, Mauritania, Niger, Nigeria, Rwanda, Senegal, Seychelles, Sierra Leone, South Africa, Sudan, Eswatini, Togo and Zambia. Since time-series analysis requires a long period of data series, we have to drop those countries with the severe problem of missing data. The 30 countries included in our study have relatively complete time series data and are representative enough to discuss the nexus between globalisation and economic output in the SSA area. The sample covers the period from 1970 to 2017. The summary of descriptive statistics is listed in Table 1.

Table 1.

Descriptive statistics.

		Mean	Median	Maximum	Minimum	SD	Skewness	Kurtosis
Benin	GDP	909.83	870.89	1166.76	745.78	118.77	0.58	−0.80
Benin	GLO	36.40	35.08	52.90	21.70	9.31	0.17	−1.20
Botswana	GDP	4263.54	4322.95	7864.25	744.92	2125.40	0.03	−1.10
Botswana	GLO	47.15	43.81	58.30	38.30	6.43	0.52	−1.34
Burkina Faso	GDP	470.66	403.47	771.25	303.97	138.68	0.74	−0.72
Burkina Faso	GLO	35.36	31.48	54.00	25.30	9.46	0.77	−0.82
Burundi	GDP	264.01	255.68	336.58	214.14	37.01	0.54	−1.06
Burundi	GLO	25.16	23.59	41.40	14.30	8.65	0.60	−0.90
Cameroon	GDP	1284.17	1265.77	1830.98	930.99	225.29	0.57	−0.17
Cameroon	GLO	39.95	37.25	51.50	30.40	6.13	0.51	−0.98
Central African Republic	GDP	488.72	481.08	643.47	334.44	85.47	0.15	−0.90
Central African Republic	GLO	30.43	29.14	39.00	23.60	4.28	0.77	−0.70
Chad	GDP	617.94	557.11	961.38	406.14	166.77	0.80	−0.78
Chad	GLO	29.72	27.47	41.10	20.40	6.79	0.32	−1.48
Congo Dem. Rep.	GDP	602.85	427.77	1134.30	276.06	292.22	0.43	−1.39
Congo Dem. Rep.	GLO	32.72	30.88	46.20	20.00	7.33	0.50	−0.79
Congo Rep.	GDP	2566.73	2567.74	3413.04	1678.85	414.09	−0.29	−0.31
Congo Rep.	GLO	42.08	41.78	54.80	32.40	6.28	0.35	−0.84
Cote d’Ivoire	GDP	1596.76	1451.32	2471.02	1132.55	373.25	0.81	−0.50
Cote d’Ivoire	GLO	43.96	41.05	55.40	34.30	6.33	0.35	−1.30
Gabon	GDP	10821.36	10473.80	19581.67	7205.91	2177.62	−0.20	−0.56
Gabon	GLO	48.31	47.81	54.70	39.10	4.24	−0.14	−0.79
Gambia	GDP	813.15	822.61	903.54	702.15	47.89	1.24	0.63
Gambia	GLO	39.81	35.95	54.60	29.60	8.22	0.55	−1.36
Ghana	GDP	1035.26	940.41	1739.47	693.95	275.41	0.71	2.28
Ghana	GLO	45.31	44.33	61.50	29.20	9.38	0.11	−1.32
Kenya	GDP	877.03	856.59	1156.43	598.46	98.85	0.65	−0.59
Kenya	GLO	44.80	44.93	57.00	32.00	7.73	0.20	−1.30
Lesotho	GDP	778.36	724.52	1421.74	349.60	305.26	1.17	0.11
Lesotho	GLO	41.46	41.07	53.90	30.00	7.13	0.09	−1.01
Madagascar	GDP	565.12	503.63	863.34	428.21	121.77	0.84	−0.10
Madagascar	GLO	34.86	30.85	50.80	25.00	9.10	0.66	−1.13
Malawi	GDP	392.30	381.88	512.65	293.16	56.42	0.24	−1.26
Malawi	GLO	35.75	31.60	50.20	26.30	7.81	0.50	−1.24
Mali	GDP	540.50	504.84	766.48	355.06	123.14	0.54	0.36
Mali	GLO	37.64	38.21	49.40	25.50	7.92	−0.06	−1.38
Mauritania	GDP	1662.32	1645.72	2004.32	1403.20	134.82	1.14	0.43
Mauritania	GLO	38.83	36.21	52.50	27.60	7.49	0.60	−0.85
Niger	GDP	542.36	508.27	837.02	426.68	108.57	0.40	−1.14
Niger	GLO	31.99	29.49	47.60	19.30	8.12	0.53	−0.79
Nigeria	GDP	1798.11	1766.49	2563.90	1324.30	397.40	1.24	0.77
Nigeria	GLO	44.82	43.20	57.40	30.40	8.01	0.14	−1.19
Rwanda	GDP	451.89	414.23	802.75	220.54	137.27	0.74	0.50
Rwanda	GLO	30.12	25.74	52.40	18.90	10.64	1.03	−0.34
Senegal	GDP	1177.09	1146.60	1495.34	1002.41	109.14	0.29	−0.87
Senegal	GLO	47.12	45.03	61.40	31.70	8.89	0.20	−0.99
Seychelles	GDP	8141.42	8177.25	14028.72	3511.85	2855.78	−0.22	−0.88
Seychelles	GLO	47.87	42.85	62.50	39.80	7.83	0.71	−1.11
Sierra Leone	GDP	416.91	431.67	567.83	272.99	73.36	0.38	−0.91
Sierra Leone	GLO	31.48	29.00	46.70	22.20	7.85	0.90	−0.62
South Africa	GDP	6501.05	6384.35	7582.70	5517.53	622.02	1.20	0.14
South Africa	GLO	50.75	41.29	70.60	36.70	13.64	0.41	−1.71
Sudan	GDP	1052.93	880.22	1900.09	691.98	362.84	0.23	−1.23
Sudan	GLO	33.70	30.77	43.80	24.50	6.51	0.41	−1.48
Eswatini	GDP	2818.62	2860.77	4711.69	1225.81	1101.07	1.76	5.20
Eswatini	GLO	40.23	39.25	48.90	31.60	5.06	−0.02	−1.19
Togo	GDP	589.90	589.25	733.46	445.26	58.94	0.08	0.08
Togo	GLO	41.03	39.60	54.90	25.80	8.31	0.03	−0.82
Zambia	GDP	1258.61	1214.41	1660.76	906.58	256.39	0.22	−1.47
Zambia	GLO	45.09	43.41	57.60	33.40	7.70	0.17	−1.37

GDP: gross domestic product, GLO: globalisation.

The descriptive statistics show that 14 countries’ means of real GDP per capita are over $1000. They are Botswana, Cameroon, Congo Rep., Cote d’Ivoire, Gabon, Ghana, Mauritania, Nigeria, Senegal, Seychelles, South Africa, Sudan, Eswatini and Zambia. Notably, the real income per capita of Gabon even exceeds $10,000. Meanwhile, the globalisation indices are over 40 in Botswana, Congo Rep., Cote d’Ivoire, Gabon, Ghana, Kenya, Lesotho, Nigeria, Senegal, Seychelles, South Africa, Eswatini, Togo and Zambia. These figures indicate that most SSA countries with higher income also have a higher level of globalisation. In addition, apart from Gabon, Mali and Eswatini, globalisation indices are positively skewed for all SSA countries. GDP in these countries is also positively skewed except Congo Rep., Gabon and Seychelles. The most striking case in SSA countries is Gabon, which benefited from abundant petroleum and foreign private investment. Moreover, as the most globalised country and a member of Brazil, Russia, India, China, and South Africa (BRICS), South Africa’s real income per capita is also impressive (over $6000) in comparison to other SSA countries.

Empirical findings

The ARDL bounds test allows for time series to be either I(0) or I(1) but not integration order higher than 1. In addition, Goh et al. (2017a) point out that the bootstrap procedure ensures the correct inference for degenerate case #1. To figure out the integrated order of each variable, we apply the unit root test proposed by Ng and Perron (2001). The results are presented in Table 2. The M-class statistics MZa, MZt, MSB and MPT perform much better than the traditional univariate unit root tests. In brief, we could conclude that all variables for most SSA countries are I(1) process, which satisfies the requirements of the ARDL bounds test. Moreover, the bootstrap ARDL test proposed by McNown et al. (2018) relaxes such restrictions on the integrated order of the variables in the model. In the bootstrap procedure, an additional t-test on the dependent variable can rule out the degenerate case #1.

Table 2.

Results for Ng-Perron unit root test.

		Level				First difference
		MZa	MZt	MSB	MPT	MZa	MZt	MSB	MPT
Benin	GDP	1.33	0.79	0.59***	30.59***	-21.24***	-3.22***	0.15	1.30
Benin	GLO	1.08	1.02	0.95***	64.45***	−9.02***	−2.12**	0.24	2.72
Botswana	GDP	0.73	0.68	0.94***	58.43***	−2.54	−0.98	0.39***	8.95***
Botswana	GLO	0.66	0.48	0.73***	37.62***	−22.70***	−3.35***	0.15	1.15
Burkina Faso	GDP	2.25	2.95***	1.31***	144.12***	−22.76***	−3.36***	0.15	1.13
Burkina Faso	GLO	1.91	2.36**	1.24***	121.61***	−21.61***	−3.29***	0.15	1.14
Burundi	GDP	−1.32	−0.61	0.46***	13.47***	−22.44***	−3.33***	0.15	1.15
Burundi	GLO	1.25	1.13	0.90***	60.78***	−3.53	−1.33	0.38***	6.94***
Cameroon	GDP	−23.33***	−3.35***	0.14	1.27	−5.82	−1.71	0.29***	4.21***
Cameroon	GLO	1.09	0.72	0.67***	35.19***	−20.97***	−3.23***	0.15	1.19
Central African Republic	GDP	−2.39	−0.84	0.35***	8.88***	−22.82***	−3.37***	0.15	1.10
Central African Republic	GLO	1.18	0.71	0.60***	30.53***	−22.61***	−3.15***	0.14	1.80
Chad	GDP	−2.04	−0.90	0.44***	10.90***	−21.75***	−3.29***	0.15	1.15
Chad	GLO	0.98	0.79	0.81***	47.34***	−10.20**	−2.25**	0.22	2.45
Congo Dem. Rep.	GDP	−1.54	−0.77	0.50***	13.80***	−9.15**	−2.14**	0.23	2.68
Congo Dem. Rep.	GLO	1.39	1.45	1.05***	81.44***	−4.43	−1.41	0.32***	5.67***
Congo Rep.	GDP	−3.79	−1.29	0.34***	6.52***	−16.62***	−2.83***	0.17	1.69
Congo Rep.	GLO	1.37	0.94	0.69***	38.70***	−22.98***	−3.39***	0.15	1.08
Cote d’Ivoire	GDP	−2.13	−1.02	0.48***	11.43***	−6.17***	−1.74***	0.28	4.03
Cote d’Ivoire	GLO	1.54	1.44	0.94***	68.27***	−22.34*	−3.32*	0.15	1.17
Gabon	GDP	−3.25	−1.27	0.39***	7.53***	−19.86***	−3.14***	0.16	1.28
Gabon	GLO	−0.59	−0.30	0.51***	17.47***	−22.83***	−3.38***	0.15	1.08
Gambia	GDP	−5.41	−1.64	0.30***	4.55***	−22.83***	−3.36***	0.15	1.14
Gambia	GLO	1.02	0.89	0.87***	54.61***	−22.30***	−3.33***	0.15	1.13
Ghana	GDP	−0.04	−0.02	0.49***	18.41***	−19.14***	−3.07***	0.16	1.37
Ghana	GLO	1.33	1.37	1.03***	77.95***	−22.51***	−3.36***	0.15	1.09
Kenya	GDP	1.37	0.74	0.54***	26.66***	−3.02	−1.18	0.39***	8.00***
Kenya	GLO	1.18	1.45	1.23***	105.36***	−11.97**	−2.37**	0.20	2.33
Lesotho	GDP	1.63	1.78	1.09***	92.36***	−22.93***	−3.36***	0.15	1.16
Lesotho	GLO	0.69	0.71	1.04***	69.53***	−9.77**	−1.89*	0.19	3.68
Madagascar	GDP	0.08	0.07	0.88***	45.75***	−22.96***	−3.39***	0.15	1.08
Madagascar	GLO	1.24	1.26	1.02***	75.13***	−22.64***	−3.36***	0.15	1.08
Malawi	GDP	0.55	0.27	0.48***	20.12***	−5.59	−1.64	0.29***	4.49***
Malawi	GLO	0.56	0.33	0.58***	26.20***	−359.58***	−13.41***	0.04	0.07
Mali	GDP	1.05	0.78	0.74***	41.27***	−22.34***	−3.34***	0.15	1.10
Mali	GLO	1.02	1.20	1.18***	94.46***	−22.10***	−3.32***	0.15	1.11
Mauritania	GDP	−3.05	−1.22	0.40***	8.01***	−22.11***	−3.32***	0.15	1.12
Mauritania	GLO	1.20	0.92	0.77***	45.40***	−2.92	−1.16	0.40***	8.26***
Niger	GDP	−0.92	−0.60	0.66***	22.49***	−22.49***	−3.35***	0.15	1.09
Niger	GLO	1.53	1.39	0.91***	64.65***	−21.09***	−3.24***	0.15	1.20
Nigeria	GDP	−2.01	−0.81	0.40***	10.34***	−15.51***	−2.74***	0.18	1.73
Nigeria	GLO	0.84	0.75	0.89***	55.19***	−21.42***	−3.26***	0.15	1.17
Rwanda	GDP	−1.46	−0.48	0.33***	10.02***	−22.04***	−3.32***	0.15	1.11
Rwanda	GLO	2.19	2.93***	1.34***	147.97***	−2.25	−1.05	0.47***	10.82***
Senegal	GDP	−0.84	−0.33	0.40***	12.71***	−7.25***	−1.78***	0.25	3.81
Senegal	GLO	1.00	0.86	0.86***	52.95***	−22.68***	−3.37***	0.15	1.08
Seychelles	GDP	1.38	1.27	0.92***	64.30***	−19.33***	−3.10***	0.16	1.31
Seychelles	GLO	0.84	0.53	0.64***	31.47***	−21.16***	−3.24***	0.15	1.20
Sierra Leone	GDP	−3.01	−1.22	0.40***	8.13***	−22.97***	−3.39***	0.15	1.07
Sierra Leone	GLO	0.87	0.55	0.63***	31.26***	−55.61***	−5.27***	0.09	0.45
South Africa	GDP	−2.82	−0.98	0.35***	8.10***	−18.49***	−3.04***	0.16	1.34
South Africa	GLO	0.07	0.04	0.61***	25.62***	−16.37***	−2.85***	0.17	1.53
Sudan	GDP	1.97	1.49	0.76***	50.36***	−20.07***	−3.17***	0.16	1.23
Sudan	GLO	1.00	1.09	1.09***	81.46***	−22.18***	−3.32***	0.15	1.13
Eswatini	GDP	1.13	1.13	1.00***	71.69***	−17.51***	−2.92***	0.17	1.53
Eswatini	GLO	1.37	1.31	0.96***	69.01***	−22.88***	−3.37***	0.15	1.12
Togo	GDP	−5.87	−1.67	0.28***	4.32***	−22.86***	−3.37***	0.15	1.11
Togo	GLO	0.85	0.78	0.92***	58.14***	−14.16***	−2.65***	0.19	1.76
Zambia	GDP	−5.50	−1.62	0.29***	4.56***	−4.53	−1.49	0.33***	5.43***
Zambia	GLO	1.20	1.23	1.03***	76.28***	−38.94***	−4.40***	0.11	0.67

Note: ***, ** and * represent 1%, 5% and 10% significant level, respectively. The lag order is determined by SIC criterion.

GDP: gross domestic product, GLO: globalisation, SIC: Schwarz information criterion.

We first estimate equation (3), the so-called GDP equation, and the results are reported in Table 3. As mentioned earlier, the structural breaks are detected through the procedure of Bai and Perron (2003). The optimal lag is determined by the AIC criterion in an unrestricted VAR model (Pesaran et al., 2001).

Table 3.

Estimations on gross domestic product (GDP equation).

	Benin	Botswana	Burkina Faso	Burundi	Cameroon	Central African Republic	Chad	Congo Dem. Rep.	Congo Rep.	Cote d’Ivoire	Gabon	Gambia	Ghana
C	1.0538	0.1715	0.5096*	0.5883*	0.7222*	1.6856*	0.5300	0.1927	0.8774**	0.3368	4.7883***	1.8771***	−0.1058
Y (–1)	−0.2226	−0.0567***	−0.2421*	−0.0826	−0.1185**	−0.1801*	−0.1061	−0.0272	−0.1156**	−0.0530	−0.2602***	−0.2538**	−0.0524
X (–1)	0.1327**	0.0867	0.2810*	−0.0451**	0.0353	−0.1722	0.0474	−0.0095	0.0083	0.0128	−0.6098***	−0.0481	0.1251
DY (–1)	0.0053	0.1695	−0.0729	0.0055	0.4677**	−0.0018	0.2101	0.5124***	0.4653***	0.4088**	0.1248	0.0710*	0.2531
DY (–2)	−0.0619				−0.0313		−0.0347	0.0903	0.0255	0.0006
DY (–3)	−0.0648						0.1926			0.1582
DX (–1)	0.0361	−0.2893	0.0270	0.4895**	−0.1874	0.4144	−0.0937	0.2260	0.0324	0.4317	−0.1903	0.1693	0.1542
DX (–2)	−0.2641*				0.0679		0.0251	0.1194	0.0431	−0.3068
DX (–3)	−0.0927						−0.1005			0.2244
TB1	1980	1980	1977	1987	1981	1990	1979	1993	1981	1977	1978	1979	1981
TB2	1997	1988		1995	1992		1986	2000	2000	1984			1994
TB3		2011					2004			1991			2006
TB4										2011
R²	0.3283	0.4971	0.1688	0.2888	0.4254	0.1163	0.4931	0.6583	0.3979	0.4242	0.5130	0.2322	0.3509
Adjusted R²	0.1247	0.4044	0.0649	0.1793	0.2978	0.0059	0.3189	0.5824	0.2641	0.2013	0.4522	0.1362	0.2313
DW Stat.	2.1894	2.1543	2.0492	2.0726	2.0021	2.0720	2.3350	1.6308	2.1216	1.6320	2.1755	2.0771	1.9550
LM_sc	1.6622	2.8506	0.9218	0.8706	4.4504	1.9020	1.3284	2.1786	1.8818	3.4882	2.2832	3.1372	0.0454
LM_rr	2.7265	3.4779	0.1519	0.2449	0.7005	7.3486	0.6597	0.0001	2.7959	2.4795	3.3627	3.8600	0.6518
LM_het	1.2056	0.8182	1.9977	0.7832	1.3277	2.2899	0.4288	2.4064	1.1902	0.9447	1.9063	0.4339	1.0158
F	2.4599	5.6999**	2.2147	2.3626	3.0925	1.8603	1.3197	2.1539	2.6284	1.7701	9.8721***	4.4475	4.4574*
t _dep	−1.6602	−2.8252**	−1.8688*	−1.5629	−2.4818	−1.9205	−1.4004	−1.2685	−2.1081	−1.0385	−3.4243***	−2.6594	−1.4701
t _ind	2.0428^**	1.0156	1.9789*	−2.1017	0.6616	−1.3989	0.5204	−0.1691	0.1313	0.1458	−3.9124***	−1.7742	2.9553***
	Kenya	Lesotho	Madagascar	Malawi	Mali	Mauritania	Niger	Nigeria	Rwanda	Senegal	Seychelles	Sierra Leone	South Africa
C	0.0214	0.0297	0.4537	0.3947	0.6211**	1.1427	0.3173	0.1630	0.5847***	0.1680	−0.0020	0.2970	0.8726*
Y (–1)	−0.0250	−0.0356	−0.0722	−0.1202	−0.2126**	−0.1684*	−0.0777	−0.0484	−0.1877***	−0.0793	−0.0148	−0.0753	−0.1193*
X (–1)	0.0408	0.0671	−0.0025	0.0928	0.2023*	0.0272	0.0473	0.0526	0.1743***	0.1017***	0.0404	0.0465	0.0462**
DY (–1)	0.4821***	0.1515	−0.1134	0.0576	−0.0470	−0.1949	−0.0827	0.2120	−0.1814***	−0.1471	0.2054	−0.0257	0.3825**
DY (–2)	−0.1671	−0.5274***	−0.0149	0.2191				0.1312		0.1003			−0.0582
DY (–3)		0.0512		0.0612
DX (–1)	−0.0980	−0.2998	0.0421	−0.1804	−0.2400	0.0912	0.0248	−0.1453	0.2816	0.0069***	−0.3364	0.0066	−0.1263
DX (–2)	0.0254	−0.3335	−0.0192	−0.1114				0.2009		0.9466			−0.0655
DX (–3)		0.0645		−0.1559
TB1	1993	1981	1981	1985	2001	1990	1977	1983	1994	1992	1978	1992	1991
TB2	2011		1991	1995		2005	1984	2002			1986	2011	2000
TB3				2007			1991
TB4							2004
R²	0.3288	0.4115	0.3596	0.3423	0.2033	0.2100	0.4491	0.3331	0.8829	0.3881	0.2367	0.2442	0.3499
Adjusted R²	0.1796	0.2557	0.2173	0.1162	0.1038	0.0885	0.3300	0.1849	0.8683	0.2723	0.1192	0.1279	0.2054
DW stat.	1.9296	1.9786	2.2533	2.2333	2.0332	2.1500	2.0988	2.0848	1.5650	1.7017	2.0534	2.0290	2.0425
LM_sc	0.1815	3.5823	1.5742	1.8224	0.1885	3.4567	0.2849	0.6036	2.2331	1.6069	0.1321	0.0445	2.2142
LM_rr	0.1927	1.4515	0.4932	1.0514	1.9498	2.2615	0.1543	1.5372	3.5472	1.2130	0.2278	2.7250	0.9845
LM_het	0.9778	2.9795	0.3396	1.3994	2.5475	0.8621	1.9698	1.5299	1.5164	3.7270	1.5129	1.1703	1.4813
F	1.2816	0.0923	1.9274	1.2799	2.3408	2.7051	4.7726*	0.8693	9.8073***	5.2703	0.0936	1.4055	2.9341
t _dep	−0.4380	−0.3539	−1.3424	−1.2158	−2.1566	−1.8172	−1.3215	−1.1646	−4.0841***	−1.1947	−0.3487	−1.3269	−1.9654
t _ind	1.4448	0.3018	−0.0604	1.5995***	1.9825	0.7086	1.0269	0.9949	4.3443***	3.2447*	0.4326	0.9667	2.3755
	Sudan	Eswatini	Togo	Zambia
C	0.1778	−0.3001	1.0203	−0.2903
Y (–1)	−0.1406**	−0.0899*	−0.1487	−0.0205
X (–1)	0.2292**	0.2793	−0.0189	0.1131**
DY (–1)	0.4158***	0.1946	0.0483	−0.0733
DY (–2)				0.2398
DY (–3)				0.0074
DX (–1)	−0.0965	−0.1894	−0.0111	−0.1482
DX (–2)				0.3608
DX (–3)				0.3808
TB1	1980	1987	1983	1977
TB2	2011	2011	2011	1992
TB3				2000
TB4				2009
R²	0.3198	0.1650	0.1200	0.5459
Adjusted R²	0.2152	0.0366	−0.0154	0.3701
DW Stat.	1.8479	2.0017	1.9542	1.9416
LM_sc	0.5699	0.0018	0.0958	2.4447
LM_rr	1.4497	0.1713	0.4475	2.6815
LM_het	1.8609	2.5414	0.5599	0.6605
F	2.8715	1.4636	1.0254	2.4731
t _dep	−2.2859	−1.7069	−1.3718	−0.7032
t _ind	2.3663	1.6012	−0.3699	2.0043

Note: ***, ** and * denote 1%, 5% and 10% significant levels. TB1, TB2, TB3, TB4 denote structural breaks detected by Bai and Perron (2003). LM_sc, LM_rr, LM_het are used to investigate the serial correlation, functional form and heteroscedasticity in each equation based on the Lagrange Multiplier (LM) test. DW Stat. refers to the Durbin-Watson statistic. F test is utilised to test on all lagged variables, t_dep test is utilised on testing on the lagged dependent variables, t_ind is utilised to test on the lagged independent variable. Cointegration occurs when F, t_dep and t_ind statistics are all significant. Degenerate case #1 occurs when F and t_dep statistics are significant. Degenerate case #2 occurs when F and t_ind statistics are significant. To generate the critical values, the bootstrap procedure is used with 5000 replications. The optimal lag is determined by the Akaike Information Criterion (AIC) criterion by allowing the maximum lag to be 3 by referring the sample size.

We build a series of LM tests, including LM_sr, LM_rr and LM_het, for each equation to detect issues of serial correlation, misspecification and heteroscedasticity. Moreover, the coefficients of lagged independent variables are used to describe their long-run impacts. Likewise, the coefficients of the lagged difference of independent variables capture the short-run impacts (Shahbaz et al., 2016, 2018).

Results show that, in the long run, globalisation would positively affect economic output in Benin, Botswana, Burkina Faso, Cameroon, Chad, Congo, Rep, Cote d’Ivoire, Ghana, Kenya, Lesotho, Madagascar, Malawi, Mali, Mauritania, Niger, Nigeria, Rwanda, Senegal, Seychelles, Sierra Leone, South Africa, Sudan, Eswatini and Zambia by considering the coefficients of lagged variables (namely X). In contrast, globalisation would negatively affect economic output for the rest of the countries in the long run. For the short-run effects evidenced by the coefficients for the lagged difference of independent variables (namely DX), the results differ from each case due to different lag structures. In brief, globalisation positively affects economic output in Burkina Faso, Burundi, Central African Republic, Congo Dem. Rep., Congo Rep., Gambia, Ghana, Mauritania, Niger, Rwanda, Senegal and Sierra Leone. This result implies that globalisation has heterogeneous effects on economic output across SSA countries. Statistics including LM_sr, LM_rr and LM_het are used here to demonstrate the validity of our estimations. We found no serial correlation and heteroscedasticity for each country. Apart from this, the LM_rr statistics indicate the functional forms are suitable for corresponding equations. We follow McNown et al. (2018) to examine the cointegration in the GDP equation by ensuring all the statistics are significant. In addition, the degenerate case #1 occurs when $F$ and $t_{i n d}$ are rejected. The degenerate case #2 occurs when the $F$ and $t_{d e p}$ are rejected. According to these criteria, only Gabon and Rwanda have such long-run dynamic equilibrium with cointegration. Furthermore, the degenerate cases mentioned above occur in Botswana and Ghana. In fact, any ignorance of the degenerate cases can lead to spurious cointegration (Goh et al., 2017a). If the degenerate case #1 or #2 occurs, there is no cointegration for that particular variable in the corresponding equation.

Similarly, estimations for equation (4) are reported in Table 4. In the long-run estimations evidenced by the coefficients of lagged X, economic output positively affects globalisation in Botswana, Burkina Faso, Congo Rep., Ghana, Lesotho, Malawi, Mali, Niger, Nigeria, Senegal, Seychelles, Sudan and Eswatini. For other countries, economic output exerts negative impacts on globalisation in the long run. Moreover, the short-run effects of economic output on globalisation, indicated by the coefficients of the lagged difference of independent variables, are positive in Burundi, Cameroon, Chad, Gabon, Gambia, Madagascar, Mauritania, Nigeria, Seychelles, Sierra Leone and Sudan. In contrast, economic output negatively affects globalisation in Botswana, Burkina Faso, Central African Republic, Ghana, Mali, Niger, Rwanda, Eswatini and Togo in the short run. For other countries, the sign of the coefficients on different Xs are mixed, which implies mixed results. LM_sr, LM_rr and LM_het statistics demonstrate that the specifications are valid. In addition, we find cointegration for the GLO equation in Burkina Faso and Ghana, indicating that economic output can determine globalisation only in these two countries. We only observe degenerate cases for GLO equations in Cameroon.

Table 4.

Estimations on globalisation (GLO) equation.

	Benin	Botswana	Burkina Faso	Burundi	Cameroon	Central African Republic	Chad	Congo Dem. Rep.	Congo Rep	Cote d’Ivoire	Gabon	Gambia	Ghana
C	0.1606	0.1208	−0.6079	0.2860	0.5401***	0.2669	0.2456	0.4326*	−0.0155	0.1716	0.4516	0.2899	−0.1104
Y (–1)	−0.0058	−0.0454	−0.2366	−0.0120	0.0219	−0.0485	0.0077	−0.0703*	−0.0269	−0.0093	−0.1042	−0.0028	−0.0863**
X (–1)	−0.0175	0.0074	0.2390	−0.0416	−0.0853***	−0.0151	−0.0400	−0.0282*	0.0163	−0.0181	−0.0048	−0.0404	0.0655**
DY (–1)	−0.0999	0.0770	0.0000	0.1062	−0.1008	−0.0161	0.0984	−0.0867	0.0116	0.1846	0.0254	0.0684	0.0210
DY (–2)	−0.3925**				−0.3327***		−0.5329***	0.1228	−0.2711*	−0.0468
DY (–3)	0.1778						−0.0463			0.1074
DX (–1)	0.0868	−0.0164	−0.1887	0.1033	0.0801	−0.0250	0.0166	−0.1146	0.0571	−0.0649	0.0082	0.1762	−0.0291
DX (–2)	−0.0316				0.1379		0.1023	0.1637	−0.1203	0.0390
DX (–3)	−0.0496						0.1515			0.0132
TB1	1983	2000	1990	1987	1978	1979	1977	1984	1982	1980	1978	1987	1978
TB2	1991	2010	1998	2000	1992	2005	1993	2007	1993	1998	1998	1999	1993
TB3				2007					2004	2007		2006	2000
TB4
R²	0.3316	0.3123	0.5142	0.5203	0.5496	0.3583	0.3670	0.4397	0.2545	0.6366	0.3047	0.1770	0.2461
Adjusted R²	0.1291	0.2065	0.4394	0.4320	0.4496	0.2596	0.1752	0.3152	0.0629	0.5116	0.1977	0.0254	0.1072
DW Stat.	2.0046	1.9490	1.8456	1.8569	2.0647	1.7397	2.1158	2.5063	2.1692	2.0645	2.0805	1.9306	2.2900
LM_sc	0.7527	0.0190	0.3995	0.2842	0.0764	0.0003	0.6468	3.7262	1.1609	0.5583	0.3801	0.2780	10.6696
LM_rr	0.3009	0.0022	0.0203	2.8273	8.3445	3.2647	2.1844	3.1156	0.3837	0.0313	0.5632	1.4847	1.1767
LM_het	0.4442	1.1665	0.4325	0.4571	0.9951	0.9925	1.2435	0.9232	0.2566	1.2482	0.5604	0.2051	1.0194
F	0.1854	0.2983	4.2661**	0.9695	5.4530**	0.6077	0.4130	1.7089	0.2059	0.6864	1.4067	0.0897	3.8522*
t _dep	−0.0815	−0.7466	−2.0547*	−0.9647	0.7540	−1.0344	0.1055	−1.7240	−0.6182	−0.2719	−1.6059	−0.1013	−2.6131**
t _ind	−0.1196	0.4991	2.2804**	−1.3529*	−3.3002***	−0.4052	−0.6628	−1.8061**	0.4307	−0.8917	−0.1512	−0.4229	2.3329**
	Kenya	Lesotho	Madagascar	Malawi	Mali	Mauritania	Niger	Nigeria	Rwanda	Senegal	Seychelles	Sierra Leone	South Africa
C	0.5111*	0.1921	0.5777	−0.2903	−0.1959	0.3365	−0.0179	0.2071	−0.0254	−0.0526	−0.0407	0.2340	0.9070*
Y (–1)	0.0145	−0.1478	−0.0436	−0.0142	−0.1348**	−0.0225	−0.0094	−0.0645*	0.0282	−0.0377	−0.0566	−0.0135	0.0272
X (–1)	−0.0816	0.0558	−0.0643	0.0598	0.1112*	−0.0324	0.0109	0.0063	−0.0082	0.0298	0.0294*	−0.0298	−0.1145*
DY (–1)	−0.2873*	−0.1552	−0.1718	0.0998	0.0141	−0.0898	−0.1863	−0.1574	0.0479	0.0043	0.2502*	−0.0112	0.3096*
DY (–2)	−0.1511	0.3042	−0.0937	−0.3712**				0.0442		0.1085			0.0726
DY (–3)		−0.0186		−0.1891
DX (–1)	−0.1684	−0.1066	0.0960	0.1005	−0.1127	0.0391	−0.1046	0.2407**	−0.0602	−0.4045**	0.0235	0.0563	0.0496
DX (–2)	0.2140**	−0.0138	0.1774	−0.2802				0.0227		0.2901			−0.1904
DX (–3)		0.0048		0.2106
TB1	1983	1981	1986	1985	1981	1979	1977	1978	1986	1978	1989	1979	1997
TB2	1993	2000	1997	1999	1993	2003	1994	1991	1994	1995	1999	1986
TB3	2005		2004				2004	1999	2001		2008	2004
TB4			2011
R²	0.4686	0.3636	0.4590	0.4913	0.2105	0.0357	0.2144	0.3332	0.2172	0.3533	0.4078	0.3169	0.5169
Adjusted R²	0.3320	0.1708	0.2998	0.3371	0.0890	−0.1127	0.0697	0.1617	0.0730	0.2096	0.2987	0.1911	0.4255
DW Stat.	2.3450	2.0144	1.9883	2.1029	2.1752	2.1509	1.9631	2.1293	1.7990	1.8138	2.0893	2.1160	1.8876
LM_sc	2.0719	2.0098	0.1402	1.9911	2.2339	1.9985	0.0111	8.0187	1.6750	1.4403	0.3094	0.4250	1.8399
LM_rr	7.7120	0.6599	0.2547	3.6816	6.5941	0.3127	3.4286	0.2898	0.7486	5.1388	7.5018	0.0288	0.3945
LM_het	0.7829	0.7470	0.4919	2.5251	1.8472	0.5962	0.2690	0.3330	1.1604	1.5283	0.6079	0.2833	3.1185
F	1.7105	0.9324	1.2098	0.2494	2.7553	0.2460	0.1893	1.8841	1.0144	0.5151	1.8956	0.3134	1.6635
t _dep	0.6086	−1.0672	−1.3468	−0.2445	−2.2653*	−0.6797	−0.2432	−1.8805	0.9713	−1.0089	−1.6510	−0.3871	1.3570
t _ind	−1.6819	0.8916	−1.5242	0.6190	1.9490*	−0.3970	0.2220	0.2427	−0.2445	0.3702	1.9403	−0.7011	−1.8161**
	Sudan	Eswatini	Togo	Zambia
C	0.0625	0.3217**	0.6835	0.2089
Y (–1)	−0.0301	−0.2210**	−0.0674**	0.0034
X (–1)	0.0072	0.0641**	−0.0663	−0.0287
DY (–1)	0.1850	−0.0609	0.2794*	−0.3415**
DY (–2)				−0.3723**
DY (–3)				−0.1243
DX (–1)	0.0451	−0.0613	−0.0603	−0.0312
DX (–2)				−0.0204
DX (–3)				0.1117
TB1	1980	1978	1979	1985
TB2	2001	1993	1993	1996
TB3		2000	2004	2005
TB4
R²	0.1778	0.2407	0.2844	0.5084
Adjusted R²	0.0513	0.1008	0.1526	0.3394
DW Stat.	1.9276	1.9874	1.8727	2.4036
LM_sc	0.1035	0.0494	0.5282	1.9888
LM_rr	1.2727	1.1577	1.3779	3.9314
LM_het	0.6460	0.2259	0.3060	0.5690
F	0.4256	2.9985	2.7276	1.3042
t _dep	−0.5864	−2.4486*	−2.3255*	0.0984
t _ind	0.2201	2.3573*	−1.0582	−1.5622***

Note: ***, ** and * denote 1%, 5% and 10% significant levels. TB1, TB2, TB3, TB4, TB5 denote structural breaks detected by Bai and Perron (2003). LM_s_c, LM_rr, LM_het are used to investigate the serial correlation, functional form and heteroscedasticity in each equation based on the Lagrange Multiplier (LM) test. DW Stat. refers to the Durbin-Watson statistic. F test is utilised to test on the all lagged variables, t_dep test is utilised on testing on the lagged dependent variables, t_ind is utilised to test on the lagged independent variable. Cointegration occurs when F, t_dep and t_ind statistics are all significant. Degenerate case #1 occurs when F and t_dep statistics are significant. Degenerate case #2 occurs when F and t_ind statistics are significant. To generate the critical values, the bootstrap procedure is used with 5000 replications. The optimal lag is determined by the Akaike Information Criterion (AIC) criterion by allowing the maximum lag to be 3 by referring the sample size.

Although several cointegration relationships are identified, the Granger causality test built upon the ARDL model could be implemented for further investigation. As suggested earlier, we should test on both the lagged level and the difference of variables in the GDP equation of Gabon and Rwanda, and the GLO equation of Burkina Faso and Ghana by imposing the restriction $H_{0} : \hat{ω}' > 0; \hat{θ}'_{j} = 0 (j = 1, 2, \dots, p_{j}^{x})$ . For other cases, we only test the coefficients of the lagged differences of the variables by imposing the restriction $H_{0} : \hat{θ}'_{j} = 0 (j = 1, 2, \dots, p_{j}^{x})$ . For simplicity, we use the chi-square statistics to test the non-Granger causality null hypothesis. The results are reported in Table 5. Clearly, the null hypothesis that GLO does not Granger cause GDP is rejected in Burundi, Gabon, Rwanda, Senegal and Zambia. For Gabon and Rwanda, globalisation accounts for economic outputs in the long run, while for Burundi, Senegal and Zambia, globalisation accounts for economic outputs in the short run. Similarly, the null hypothesis that GDP does not Granger cause GLO is rejected in Burkina Faso, Cameroon, Ghana, Kenya and Senegal. Therefore, economic outputs can lead to globalisation in the long run for Burkina Faso and Ghana. In contrast, for Cameroon, Kenya and Senegal, economic outputs account for globalisation in the short run. In addition, bidirectional causality is evident in Burkina Faso and Senegal.

Table 5.

Results for Granger causality test.

		Chi-square statistics	P-value
Benin	GLO≠>GDP	4.1179	0.249
Benin	GDP≠>GLO	0.3445	0.951
Botswana	GLO≠>GDP	1.7797	0.182
Botswana	GDP≠>GLO	0.0252	0.874
Burkina Faso	GLO≠>GDP	0.0266	0.871
Burkina Faso	GDP≠>GLO	5.5409*	0.063
Burundi	GLO≠>GDP	6.1565**	0.013
Burundi	GDP≠>GLO	1.5021	0.220
Cameroon	GLO≠>GDP	0.9064	0.636
Cameroon	GDP≠>GLO	6.1089**	0.047
Central African Republic	GLO≠>GDP	1.3311	0.249
Central African Republic	GDP≠>GLO	0.1698	0.680
Chad	GLO≠>GDP	0.2478	0.970
Chad	GDP≠>GLO	2.6764	0.444
Congo Dem. Rep.	GLO≠>GDP	1.7706	0.413
Congo Dem. Rep.	GDP≠>GLO	1.9704	0.373
Congo Rep.	GLO≠>GDP	0.0586	0.971
Congo Rep.	GDP≠>GLO	1.1997	0.549
Cote d’Ivoire	GLO≠>GDP	2.9975	0.392
Cote d’Ivoire	GDP≠>GLO	1.1129	0.774
Gabon	GLO≠>GDP	16.5511***	0.000
Gabon	GDP≠>GLO	0.0243	0.876
Gambia	GLO≠>GDP	1.3049	0.253
Gambia	GDP≠>GLO	1.3457	0.246
Ghana	GLO≠>GDP	0.6868	0.407
Ghana	GDP≠>GLO	5.4537*	0.065
Kenya	GLO≠>GDP	0.4895	0.783
Kenya	GDP≠>GLO	4.6739*	0.097
Lesotho	GLO≠>GDP	2.0806	0.556
Lesotho	GDP≠>GLO	1.7490	0.626
Madagascar	GLO≠>GDP	0.0797	0.961
Madagascar	GDP≠>GLO	2.1242	0.346
Malawi	GLO≠>GDP	2.8439	0.416
Malawi	GDP≠>GLO	4.9505	0.176
Mali	GLO≠>GDP	0.8437	0.358
Mali	GDP≠>GLO	1.4419	0.230
Mauritania	GLO≠>GDP	0.2698	0.604
Mauritania	GDP≠>GLO	0.0889	0.766
Niger	GLO≠>GDP	0.0203	0.887
Niger	GDP≠>GLO	0.8812	0.348
Nigeria	GLO≠>GDP	1.1424	0.565
Nigeria	GDP≠>GLO	6.7069**	0.035
Rwanda	GLO≠>GDP	33.4902***	0.000
Rwanda	GDP≠>GLO	2.0880	0.149
Senegal	GLO≠>GDP	8.3786**	0.015
Senegal	GDP≠>GLO	9.8645***	0.007
Seychelles	GLO≠>GDP	0.8260	0.363
Seychelles	GDP≠>GLO	0.2193	0.640
Sierra Leone	GLO≠>GDP	0.0011	0.973
Sierra Leone	GDP≠>GLO	0.2594	0.611
South Africa	GLO≠>GDP	0.9602	0.619
South Africa	GDP≠>GLO	1.2669	0.531
Sudan	GLO≠>GDP	0.1125	0.737
Sudan	GDP≠>GLO	0.3456	0.557
Swaziland	GLO≠>GDP	0.3988	0.528
Swaziland	GDP≠>GLO	0.5862	0.444
Togo	GLO≠>GDP	0.0021	0.963
Togo	GDP≠>GLO	0.3754	0.540
Zambia	GLO≠>GDP	7.2617*	0.064
Zambia	GDP≠>GLO	1.1482	0.765

Note: ***, ** and * represent 1%, 5% and 10% significant level, respectively. A≠>B refers to the hypothesis that A Granger does not cause B.

GDP: gross domestic product, GLO: globalisation.

Although the bootstrap procedure could obtain correct inference (Goh et al., 2017b) and verify the robustness of estimations, we further implement the cumulative sum of recursive residuals (CUSUM) and CUSUM square tests to check the stability for each equation (Brown et al., 1975). Shahbaz et al. (2016) also point out that model misspecification would result in biased coefficients which may weaken the explanatory power of results. Additionally, the critical values and the CUSUM values are closely linked to each other, which could be used to examine the efficiency of the bootstrap procedure. Shahbaz et al. (2016) suggest that if the expected value of residuals is zero, we could conclude that non-rejection of the consistency of the estimations. Figure 1 shows both CUSUM and CUSUM square statistics of both GDP and GLO equations for all countries. We find that most CUSUM and CUSUM square statistics are within the 5% significant bounds, which indicates the stability of our estimations.

Figure 1.

Plots of CUSUM and CUSUM square.

Conclusions and policy implications

Globalisation is essential for developing and transitioning economies, which brings not only chances, but also new challenges and risks. With the emerging of trade protectionism and de-globalisation as well as China’s ambitious BRI plan, whether globalisation promotes economic output in developing areas like SSA countries awaits a thorough investigation. This study adds to the literature by focusing on the SSA countries to explore the relationship between globalisation and economic output using time-series analysis for the first time. Compared to studies using panel data models, time-series analysis could provide more details at the country level and could identify the long-run nexus between variables. In addition, this study uses the latest version of the KOF Globalisation Index, which is more robust than other indices used before. Moreover, this study adopts a newly proposed bootstrap ARDL test for more reliable identification of cointegration and causality. This study contributes to the theory on the nexus between globalisation and economic output in areas that are developing and not well integrated into the global economy.

The empirical results show that only the GDP equations for Gabon and Rwanda are cointegrated in the long run. Moreover, we also rule out the degenerate case #1 for Ghana and the degenerate case #2 for Botswana, which could avoid the spurious cointegration. For the GLO equation, we only find cointegration for Burkina Faso and Ghana. We also rule out the degenerate case #1 for Cameroon. Although several cointegration relationships are found in different countries, the Granger causality test is implemented. For simplicity, GLO causes GDP for Burundi, Gabon, Rwanda, Senegal and Zambia. GDP causes GLO for Burkina Faso, Cameroon, Ghana, Kenya and Senegal. We can conclude that causality varies across countries.

Therefore, for 24 out of 30 SSA countries, there is a positive relationship between globalisation and economic output in the long run. Apart from Gabon and Rwanda, however, we do not observe the causal effect from globalisation to economic output. Conversely, economic output positively affects globalisation in 13 out of 30 SSA countries in the long run, but the causal effect is only observed in Burkina Faso and Ghana. Results from the Granger causality test also show that the causal effects either from globalisation to economic output or the opposite way only exist in several countries and are not a widespread phenomenon. Thus, globalisation cannot increase economic output for all SSA countries, which could empirically verify that globalisation should be well managed. Otherwise, it will restrict job creation and lead to a risky economic system which may not spur economic output. It seems that globalisation and economic output mutually determine each other and therefore form a positive spiral development path. From this point of view, engagement with China’s BRI is an excellent chance for SSA since it will not only increase its globalisation level, but also is beneficial to the improvement in infrastructure which is one of the fundamental factors of economic output.

Indeed, these findings offer important policy implications. Our evidence highlights that making policies regarding globalisation–output nexus should be treated with caution as ‘one size does not fit all’, given the various causal relationship across SSA countries. The specific features of different economies should be realised by relevant authorities to make sound globalisation policy to avoid the adverse effects of global integration. For countries where globalisation drives economic output in a unidirectional fashion, policies should be targeted at deepening the level of globalisation in corresponding countries. Greater openness to the global economy will contribute to the economic development of these countries. However, for countries where economic output drives globalisation in a unidirectional causal link, policies targeted at promoting economic performance should be pursued before further opening-up. Once the size or quality of the economy reaches a certain level, globalisation follows. For countries where bidirectional causality is documented, there is a positive spiral development path for them. Then, policies should be strengthened to maintain this virtuous circle.

This study also has some limitations which could be addressed in future studies. First, due to the unavailability of data, we are not able to cover all SSA countries in our analysis. If all countries have enough quality data, future studies could examine the spatial effects of globalisation on economic performance in the SSA area: in other words, whether the globalisation of one country improves or inhibits the adjacent countries’ economic performance and vice versa. Second, the causal relationship between globalisation and economic output may be conditioned on a certain time framework, namely, in the short, intermediate or long run. It would be interesting for future research to examine underlying time-varying causality relationships. Further studies could also focus on the indirect effects of globalisation on economic output, through pathways such as FDI, trade, communications and cultural interactions for non-output hypothesis supporters.

Footnotes

Authors’ note

Huiping Dong is now affiliated with University of Science and Technology of China, Hefei, China and Yifei Cai is now affiliated with The University of Western Australia, Perth, Australia.

Funding

The authors disclosed receipt of the following financial support for the research, authorship and/or publication of this article: This study is supported by National Natural Science Foundation of China (grant number 72004052).

ORCID iD

Xing Shi

Notes

Author biography

Huiping Dong is a Research Associate at the University of Science and Technology of China. She obtains her master degree of management from the University of Science and Technology of China. Her research interest falls in the area of international economics, policy evaluation and time series analysis.

Yifei Cai is a PhD student at the University of Western Australia. His research interest falls in the time series analysis, energy and environmental economics.

Xing Shi is a Professor in economics at the Hefei University of Technology. He obtains his PhD from the University of Western Australia. His research interest falls in the economics of innovation, international economics and policy evaluation.

References

Aluko

Ibrahim

Atagbuzia

(2020) On the causal nexus between FDI and globalisation: Evidence from Africa. Journal of International Trade & Economic Development 1–21. DOI: 10.1080/09638199.2020.1823460.

Asongu

Biekpe

(2018) Globalization and terror in Africa. International Economics 156: 86–97.

Ayomitunde

George

Ojo

, et al. (2020) Globalization and economic growth: Evidence from European countries. European Financial and Accounting Journal 2020(1): 67–82.

Bai

Perron

(2003) Computation and analysis of multiple structural change models. Journal of Applied Econometrics 18(1): 1–22.

Brown

Durbin

Evans

(1975) Techniques for testing the constancy of regression relationships over time. Journal of the Royal Statistical Society, Series B (Methodological) 37(2): 149–163.

Buysse

Can

Gozgor

(2018) Globalisation outcomes and the real output in the sub-Saharan Africa LICs: A cointegration analysis. Economic research – Ekonomska istraživanja 31(1): 338–351.

Caglar

(2020) The importance of renewable energy consumption and FDI inflows in reducing environmental degradation: Bootstrap ARDL bound test in selected 9 countries. Journal of Cleaner Production 264: 121663.

Cai

Chang

Inglesi-Lotz

(2018a) Asymmetric persistence in convergence for carbon dioxide emissions based on quantile unit root test with Fourier function. Energy 161: 470–481.

Cai

Sam

Chang

(2018b) Nexus between clean energy consumption, economic growth and CO₂ emissions. Journal of Cleaner Production 182: 1001–1011.

10.

Carkovic

Levine

(2005) Does foreign direct investment accelerate economic growth? In: Moran

Graham

lomström

(eds) Does Foreign Direct Investment Promote Development? Institute for International Economics and Center for Global Development, Washington, DC, pp. 195–220.

11.

Carrion-i-Silvestre

Del Barrio-Castro

López-Bazo

(2005) Breaking the panels: An application to the GDP per capita. Econometrics Journal 8(2): 159–175.

12.

Chang

Lee

(2010) Globalization and output: A political economy analysis for OECD countries. Global Economic Review 39(2): 151–173.

13.

Chang

Lee

Hsieh

(2011) Globalization, real output and multiple structural breaks. Global Economic Review 40(4): 421–444.

14.

Chen

(2016) China’s ‘One Belt, One Road’ initiative and its implications for Sino-African investment relations. Transnational Corporations Review 8(3): 178–182.

15.

Dollar

Kraay

(2004) Trade, growth, and poverty. The Economic Journal 114(493): F22–F49.

16.

Dreher

(2006a) Does globalization affect growth? Evidence from a new index of globalization. Applied Economics 38(10): 1091–1110.

17.

Dreher

(2006b) The influence of globalization on taxes and social policy: An empirical analysis for OECD countries. European Journal of Political Economy 22(1): 179–201.

18.

Dreher

Gaston

(2008) Has globalization increased inequality? Review of International Economics 16(3): 516–536.

19.

Dreher

Sturm

Ursprung

(2008) The impact of globalization on the composition of government expenditures: Evidence from panel data. Public Choice 134(3–4): 263–292.

20.

Enders

Lee

(2012) A unit root test using a Fourier series to approximate smooth breaks. Oxford Bulletin of Economics and Statistics 74(4): 574–599.

21.

Ghazouani

Boukhatem

Sam

(2020) Causal interactions between trade openness, renewable electricity consumption, and economic growth in Asia-Pacific countries: Fresh evidence from a bootstrap ARDL approach. Renewable and Sustainable Energy Reviews 133: 110094.

22.

Goh

McNown

(2020) Macroeconomic implications of population aging: Evidence from Japan. Journal of Asian Economics 68: 101198.

23.

Goh

Sam

McNown

(2017a) Re-examining foreign direct investment, exports, and economic growth in asian economies using a bootstrap ARDL test for cointegration. Journal of Asian Economics 51: 12–22.

24.

Goh

Yong

Lau

, et al. (2017b) Bootstrap ARDL on energy-growth relationship for 22 OECD countries. Applied Economics Letters 24(20): 1464–1467.

25.

Greenaway

Morgan

Wright

(1999) Exports, export composition and growth. Journal of International Trade & Economic Development 8(1): 41–51.

26.

Gurgul

Lach

(2014) Globalization and output: Evidence from two decades of transition in CEE. Economic Modelling 36: 99–107.

27.

Gygli

Haelg

Potrafke

, et al. (2019) The KOF globalisation index – revisited. Review of International Organizations 14(3): 543–574.

28.

Kacowicz

(1999) Regionalization, globalization, and nationalism: Convergent, divergent, or overlapping? Alternatives 24(4): 527–556.

29.

Kandil

Shahbaz

Mahalik

, et al. (2017) The drivers of economic growth in China and India: globalization or financial development? International Journal of Development Issues 16(1): 54–84.

30.

Kandil

Shahbaz

Nasreen

(2015) The interaction between globalisation and financial development: new evidence from panel cointegration and causality analysis. Empirical Economics 49(4): 1317–1339.

31.

(2016) Dynamics between energy, output, openness and financial development in sub-Saharan African countries. Applied Economics 48(10): 914–933.

32.

McMillan

Rodrik

(2011) Globalization, structural change and productivity growth. NBER Working Papers 17143. Cambridge, Massachusetts: National Bureau of Economic Research. Available at: https://www.nber.org/system/files/working_papers/w17143/w17143.pdf

33.

McNown

Sam

Goh

(2018) Bootstrapping the autoregressive distributed lag test for cointegration. Applied Economics 50(13): 1509–1521.

34.

Nawaz

Lahiani

Roubaud

(2019) Natural resources as blessings and finance-growth nexus: A bootstrap ARDL approach in an emerging economy. Resources Policy 60: 277–287.

35.

Perron

(2001) Lag length selection and the construction of unit root tests with good size and power. Econometrica 69(6): 1519–1554.

36.

Perron

(1989) The great crash, the oil price shock, and the unit root hypothesis. Econometrica 57(6): 1361–1401.

37.

Pesaran

Shin

Smith

(2001) Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics 16(3): 289–326.

38.

Potrafke

(2015) The evidence on globalisation. World Economy 38(3): 509–552.

39.

Sakyi

(2011) Trade openness, foreign aid and output in post-liberalisation Ghana: An application of ARDL bounds test. Journal of Economics and International Finance 3(3): 146.

40.

Sam

McNown

Goh

(2019) An augmented autoregressive distributed lag bounds test for cointegration. Economic Modelling 80: 130–141.

41.

Shahbaz

Mallick

Mahalik

, et al. (2016) The role of globalization on the recent evolution of energy demand in India: Implications for sustainable development. Energy Economics 55: 52–68.

42.

Shahbaz

Raghutla

Song

, et al. (2020) Public-private partnerships investment in energy as new determinant of CO₂ emissions: The role of technological innovations in China. Energy Economics 86: 104664.

43.

Shahbaz

Shahzad

SJH

Alam

, et al. (2018) Globalisation, economic growth and energy consumption in the BRICS region: The importance of asymmetries. Journal of International Trade & Economic Development 27(8): 985–1009.

44.

Stiglitz

(2004) Globalization and growth in emerging markets. Journal of Policy Modeling 26(4): 465–484.