Economic growth, trade and capital flows: A causal analysis of post-liberalised South Africa

Abstract

Development Finance and EconometricsThis paper investigates the causal relationships between trade, capital inflows and economic growth in post-liberalised South Africa over the period from 1995 to 2011. The results show that economic growth in South Africa is driven primarily by trade and fixed investment rather than by capital inflows. However, the relationship between economic growth and imports is bidirectional, and thus economic growth in South Africa is associated to a greater extent with the export-led growth hypothesis than the import-led growth hypothesis. In addition, the results find in favour of growth-led FDI rather than FDI-led growth, and that portfolio inflows rather than FDI are integrated into the country's trade-led growth dynamics.

Keywords:

• causality
• economic growth
• exports
• imports
• liberalisation

JEL Classifications:

• C32
• F16
• F21
• F36
• F43
• O11
• O40

Notes

Note: The ADF unit root tests include a maximum of four lags chosen on the basis of the Akaike Information Criterion (AIC). ***, ** and * represent significance at the 1%, 5% and 10% level, respectively.

Notes: The [k + d(max)]th order level VAR was estimated with d(max) = 1 for the order of integration and lag length selection of k = 2. Reported estimates are asymptotic Wald statistics. Values in italics are p-values. ***, ** and * represent significance at the 1%, 5% and 10% level, respectively.

See Fedderke (2010) for a description of South Africa's economic growth dynamics.

Based on narrow unemployment rate data obtained from Statistics South Africa.

The reasons commonly cited for the country's high unemployment rate mainly relate to labour market distortions, including labour mispricing and inflexibility (Barker 2003; Edwards and Golub 2003; Fedderke 2005; Burger and Woolard 2005), labour saving technological change (Fedderke, Shin and Vaze 2003; Dunne and Edwards 2006), skills market mismatches and sectoral changes of demand (Burger and Woolard 2005; Pauw, Oosthuizen, and Van Der Westhuizen 2006; Bhorat and Oosthuizen 2005; Banerjee et al. 2006), insufficient absorption in the informal sector (Rodrik 2008), the impact of HIV/AIDS (Arndt and Lewis 2000; Laubscher, Smit, and Visagie 2001; Booysen, Geldenhuys, and Marinkov 2003) and the impact of trade liberalization (Bell and Cattaneo 1997; Nattrass 1998; Bhorat 1999; Birdi, Dunne, and Watson 2001).

See Giles and Williams (2000a, 2000b) for a comprehensive review of the export-led growth literature.

Cross-country evidence in support of the ELGH has been reported for developed countries (Marin 1992), Africa (Fosu 1990; Ukpolo 1994; Njikam 2003), Latin America (Van den Berg and Schmidt 1994), Asia (Rahman and Mustafa 1997; Doğanlar and Fisunoğlu 1999; Ekanayake 1999) and the Middle East (Riezman, Summers, and Whiteman 1996). Other studies report a predominantly insignificant long-run relationship in Asia (Jin 1995; Islam 1998; Shan and Sun 1998), among the ASEAN countries (Ahmad and Harnhirun 1996) and the Middle East (Al-Yousif 1997; Abu-Qarn and Abu-Bader 2004). Country-specific evidence in favour of the ELGH include Bangladesh (Begum and Shamsuddin 1998), Canada (Awokuse 2003), Chile (Siliverstovs and Herzer 2006), Ireland (Stilianos 2000), Italy (Federici and Marconi 2002), Namibia (Jordaan and Eita 2007), Pakistan (Shirazi and Manap 2004), Paraguay (Richards 2001) and Turkey (Taban and Aktar 2008). Little evidence is found for Australia (Shan and Sun 1998; Moosa 1999; Iyer, Rambaldi, and Tang 2009), Canada (Henriques and Sadorsky 1996), Columbia (Amin Gutiérrez de Piñeres and Ferrantino 1999) and India (Sharma and Panagiotidis 2003), and bidirectional causality is found for China (Shan and Sun 1998), Slovenia (Beko 2003), South Korea (Awokuse 2005) and Taiwan (Biswal and Dhawan 1998).

Country-specific evidence in favour of the ILGH include China (Liu, Song, and Romilly 1997), Czech Republic and Poland (Awokuse 2007), Japan (Lawrence and Weinstein 1999), Korea (Kim, Lim, and Park 2009), Malaysia (Baharumshah and Rashid 1999), Mexico (Iscan 1998), Singapore (Khalid and Cheng 1997), Thailand (Damooei and Tavakoli 2006) and the USA (Lawrence 1999), while bidirectional causality is reported for Nigeria (Deme 2002) and Turkey (Ugur 2008).

Cross-country studies that do not focus on endogeneity have found evidence in favour of the FDI-led growth hypothesis in Latin America (De Gregorio 1992) and among large samples of low, middle and high income countries (Blomstrom, Lipsey, and Zejan 1994; Ram and Zhang 2002), while country-specific studies have reported a causal relationship between FDI and economic growth for the United Kingdom (Blake and Pain 1994), Portugal (Cabral 1995), Ireland (Barry and Bradley 1997), Brazil (Oliveira 2001), China (Xiaohui, Burridge, and Sinclair 2002), Mexico (Ramirez 2000), Nigeria (Okodua 2009), Philippines and Thailand (Bende-Nabende et al. 2003) and Singapore and Taiwan (Ng 2006). In contrast, cross-country studies that control for endogeneity have found that growth drives FDI (Hertzer, Klasen, and Nowak-Lehmann 2008), that only a small group of countries have FDI-led growth (Duttaray, Dutt, and Mukhopadhyay 2008) or in the case of sub-Saharan Africa, that domestic investment is more significantly associated with economic growth than FDI (Adams 2009). Country-specific studies that take endogeneity into account have similarly found that growth leads to FDI in Brazil (de Mello 1997), India (Chakraborty and Basu 2002), Malaysia (Ang 2008) and South Korea (Ng 2006), while bidirectional causality has been reported in China (Ng 2006; Zhang 1999, 2002).

According to Collier, Hoeffler, and Pattillo (2001), 40% of the wealth of residents in Africa and the Middle East is held outside the host countries while for Latin America this proportion is 10%. With regards to South Africa, Mohamed and Finnoff (2004) calculate that between 1980 and 2000, South African flight capital amounted to 6.6% of GDP each year and thus was an impediment to the country's development.

These data are graphically presented in Appendix A.

Data points are considered as outliers if they last for one quarter and demonstrate the greatest positive or negative magnitude in the series. If outliers are too close together to use a 5-year window period, the next window period is used instead. The dates of the corrected outliers are listed in Appendix B and mainly relate to the Anglo American-De Beers unwinding in 2001 and the global financial crisis in 2008.

The level VAR also includes zero-one dummy variables in the first quarter of 1999, the fourth quarter of 2006 and the third quarter of 2008.

The Eviews 6 software used to conduct the analysis requires that the lag-augmented VAR is first re-specified as a seemingly unrelated regression (SUR) system in accordance with Rambaldi and Doran (1996) before the Wald tests can be undertaken.