Preferential trade agreements and manufactured goods exports: does it matter whom you PTA with?

Abstract

This article explores two questions. First, do preferential trade agreements (PTAs) affect manufactured goods exports of developing countries? Second, does it matter for developing countries whom they sign the PTAs with? We find that the answer to both questions is yes. Using bilateral manufactured goods exports data from 28 developing countries during 1978–2005; we find that South–South PTAs have a significantly positive effect on manufactured goods exports. In contrast, no such effect is detected in the case of South–North PTAs. We confirmed the robustness of these findings to estimation methodology, sample selection, time period, zero trade flows and multilateral trade resistance.

Keywords:

• manufactured goods exports
• preferential trade agreements
• south–south and south–north trade
• gravity equation
• industrial development
• developing countries

JEL Classification:

• F13
• F14
• F15
• O14
• O24
• L60

Acknowledgements

We thank Amitava K. Dutt, Kevin Gallagher, Ilene Grabel, Kevin Grier, Robin Grier, James Hartigan, Jeffrey B. Nugent, Arslan Razmi, Jaime Ros, and the session participants at the ASSA 2011 meeting in Denver, and the Eastern Economics Association 2012 meeting in Boston, and Department of Economics seminar participants at the University of Oklahoma and GLODEM seminar participants at Koç University for their comments and suggestions. For financial support, we thank the Office of the Vice President for Research at the University of Oklahoma, and the Dean of Faculty Office and the School of Critical Social Inquiry at Hampshire College. We thank Lord Andzie Quainoo for his research assistance. Authorship seniority is shared equally by the authors. All remaining errors and omissions are ours.

Notes

1. ¹ We did the search on 2 May 2012 using the PTA, regional trade agreements or multilateral trade agreements as keywords in the abstracts or titles.

2. ² Note that South–South bank lending and FDI flows have also increased significantly. The share of South–South FDI in global FDI flows, for example, increased from 16% in 1995 to 37% in 2003 (World Bank, 2006, 2008; also see Akin and Kose, 2008).

3. ³ Schiff et al. (2002) and Schiff and Wang (2008), for example, find that the impact of trade-related technology diffusion on Southern TFP is higher in South–North than South–South trade.

4. ⁴ Baldwin and Robert-Nicoud (2010) argue that these are perhaps best conceived as trade in tasks rather than trade in goods; however, they appear on official trade statistics as traditional inter-industry trade.

5. ⁵ For a discussion, see Bhagwati et al. (1998) and Panagariya (2000). Also note that empirical research on the net trade effects of PTAs is inconclusive. For example, Carrere (2004) finds significant intra-bloc trade creation but also evidence of trade diversion. In contrast, Egoume-Bossogo and Mendis (2002), Lee and Park (2005) and Mayda and Steinbertg (2009) report significant trade creation but no diversion effects.

6. ⁶ Note that this point was raised much earlier on by Myrdal (1956, p. 261), who argued that because of the colonial legacy, ‘governments and businesses in underdeveloped countries are conditioned and trained to negotiate and cooperate with their opposite partners in advanced countries but not with the governments and businesses in other underdeveloped countries’.

7. ⁷ The theoretical roots of structural gravity equation, including Linder’s (1961) preference similarity hypothesis, is also applicable to disaggregated goods (by sectors) and factor endowments (Anderson, 2011). Thus, besides the current article, several other studies have also used the gravity model to study disaggregated trade in manufactures (Rauch, 1999; Eaton and Kortum, 2004).

8. ⁸ Notice that we choose not to force the income elasticities to unity, given that all previous research as well as our empirical results indicate that they are significantly different from one. However, the alternative specification is also tested in the robustness section.

9. ⁹ The Hausman test confirms the choice of fixed-effects (with a p-value at 1% level) over random-effects model. Yet, the estimates obtained by country fixed-effects ‘might suffer from the so-called incidental parameters problem, due to insufficient degrees of freedom to consistently estimate the parameters of interest’ (Bastos and Silva, 2010, p. 106).

10. ¹⁰ In addition to the Bronze medal error, the common practice of using US CPI to deflate export values or the practice of using exporting country CPI to measure export price levels are theoretically incorrect and can cause serious bias in estimation, given the effect of changes in nontradable and exported commodity prices and their different weights in consumption baskets across countries.

11. ¹¹ As noted by Baldwin and Taglioni (2006), however, most gravity equations are not estimated on uni-directional trade but on average trade, the mis-calculation of which causes the ‘silver medal mistake’.

12. ¹² Like Baier and Bergstrand (2007), we excluded the global system of trade preferences (GSTP) from our PTA classification.

13. ¹³ Note that the PTA coefficient estimate in Silva and Tenreyro (2006) using the PPML estimator is 0.38.

14. ¹⁴ Except for CurCol, which appears with mixed coefficient sign and significance levels.

15. ¹⁵ We discuss this point more in the robustness section.

16. ¹⁶ For brevity, in this part, we report results using only the benchmark regression with the MR terms. However, other results were very similar to those reported. The estimated MR terms are also not reported for space considerations. All unreported results are available from authors upon request.

17. ¹⁷ The data source is WDI, IFS and country statistics. When not available, we used the bilateral real exchange rates with respect to the US dollar and relative producer prices.

18. ¹⁸ South Korea is classified as a high-income OECD country in 2005 by the WB, which is why we have South–South PTAs showing up for high income OECD group. This may also serve as a robustness check on the sensitivity of the results to Korea being classified as Emerging South. However, note that Korea was classified as ‘upper-middle income’ until 1994 and then again during 1998–2000 (World Bank, 2011). Besides, South Korea is a signer of the GSTP agreement.

19. ¹⁹ Because of data availability issues, production networks and sales by foreign affiliates of international firms are not separated from domestic firm sales in this article. As Anderson (2011) notes, while this is true for most gravity model estimates, it may not necessarily be appropriate if foreign firms face different trade costs than domestic firms.