The potential of African trade integration – Panel data evidence for the COMESA-EAC-SADC Tripartite

ABSTRACT

The COMESA-EAC-SADC Tripartite FTA, formed in 2011, is supposed to be a milestone towards Africa's continental trade integration. This study analyzes the impact of regional integration among the Tripartite countries on their bilateral imports before that date to evaluate the latest integration efforts. We estimate an extended gravity model on a large panel of 51 African countries using yearly observations from 1995 to 2010. We proxy existing formal trade barriers by sample average tariff data on imports from the world as well as indicator variables for the membership in regional FTAs. We consider different estimation techniques and discuss distinct sets of fixed effects. The PPML regression results indicate that remaining tariffs are significantly negatively correlated with imports throughout the preferred multiplicative models. An FTA status does not show a clear-cut import enhancing effect. In the specifications that control for country-year effects, the EAC coefficient is positively correlated with imports, and the COMESA and SADC FTA membership show a positive relation to imports within some reduced-sample robustness checks.

KEYWORDS:

• Free trade agreement
• tariff barrier
• panel data
• Poisson pseudo-maximum likelihood
• high-dimensional fixed effects
• Africa

JEL CLASSIFICATIONS:

• F13
• F14
• F15
• C23

Acknowledgments

We thank Michael Bleaney, Matthias Busse, Bernd Kempa and Philipp Dybowski, and the participants of the ETSG 16th Annual Conference 2014, the EABEW 1st Annual Conference 2015, the International Conference on Globalization and Development 2015, and Annual Conference of the Research Group on Development Economics of the Verein für Socialpolitik 2015, the Biennial Conference of the Economic Society of South Africa 2015, the PEGNet Conference 2015 and the EEFS Annual Conference 2018 for valuable comments on an earlier version of this paper.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Trade creation effects are much more important than trade diversion in Africa (see also Musila 2005). We do not consider diversion effects, but refer to Magee (2008) for an extensive empirical analysis in this strand of literature.

2. See Table 6 for a list of current and past member states of the three regional economic communities and the corresponding FTAs.

3. A Regional trade agreement (RTA) can notify one of the following agreement types: Preferential trade arrangement, free trade area, customs union, common market and economic unions (see also Frankel 1997, 12–17).

4. The complete elimination of tariffs when being in an FTA poses a simplifying assumption that holds true for the EAC, but for the COMESA FTA and the SADC FTA still many exceptions exist.

5. See also Table 5 in the appendix for summary statistics.

6. Since the estimation is computationally demanding and any time-varying country specific explanatory variables are discarded from the analysis, an alternative would be to include interval-effects, such as described in Gylfason, Martínez-Zarzoso, and Wijkman (2015).

7. The estimation is implemented in the Stata module ppml. The authors offer supplementary material on the webpage: http://personal.lse.ac.uk/tenreyro/lgw.html. For computational reasons we use the Stata command PPML_PANEL_SG written by Zyklin (2017) (see https://ideas.repec.org/c/boc/bocode/s458249.html) whenever we include country-year effects.

8. Introducing also indicator variables for the other African FTAs (of ECOWAS, Cemac and PanArab) into our loglinear models results in insignificant coefficient estimates for joining those FTAs and all other results are robust to this inclusion.

9. Introducing additional indicator variables for the other African FTAs (of ECOWAS, Cemac and PanArab) and re-estimating the four PPML models results in a higher EAC coefficient estimate in the model with time-invariant country-specific effects presented above (${0.687}^{\ast \ast }$; vs. Table 2, col. (1)). In the specification with time-varying effects, all three estimates are significant (COMESA ${0.372}^{\ast \ast \ast }$, EAC ${1.078}^{\ast \ast \ast }$, SADC ${0.303}^{\ast \ast }$; vs. Table 2, col. (2)). The results become closer to the ones obtained from the FE estimation. The minimal, if any, time variation of the other FTA indicators act like country-pair dummies for selected trading partners. This is confirmed by slightly lower absolute values of the distance coefficients. All other results are completely robust to this inclusion.

10. This is in line with Buigut (2012) who shows that although the intra-EAC exports are mainly driven by Kenya and Uganda, the intra-EAC imports have largely increased for all countries.

11. Karugia et al. (2009) analyzed trade in the maize and beef sector in Kenya, Tanzania and Uganda based on regional survey data and detected a strong link between the low level of intra-EAC trade and these NTMs.

12. Historical data for the Tripartite countries can be obtained for Botswana in the year 2006, Egypt (1999–2001), Ethiopia (1995), Sudan (2001), Kenya (1993), Lesotho (2005), Madadascar (1995), Mauritius (1995), Mozambique (1994), Namibia (2006), Rwanda (1994), South Africa (1992–2006), Swaziland (2006), Tanzania (1993, 2001), Uganda (1993), Zambia (1993) and Zimbabwe (1997).

13. For instance, Minor and Tsigas (2008) find an increase of 26 percent for iron and steel, 25 percent for chemicals, 21–24 percent for light manufactures such as apparel, footwear, furniture and leather, and 21 percent for machinery and electric equipment.

14. We use Tau-3 here as missing values are replaced by the average of all significantly positive estimates from the corresponding GTAP category. In contrast, in Tau-1 missings are replaced with zero (potential bias) and in Tau-2 missings are replaced with Hummels' (2001, 2013) point estimates from US imports (inappropriate for our sample choice).

15. A similar approach is applied in Egger (2004), who examines the export patterns for developed countries and includes the difference in high-skilled to low-skilled labor ratios between trading partners.