Sectoral FDI, absorptive capacity and economic growth – empirical evidence from Egyptian governorates

ABSTRACT

Using a novel panel dataset of Egyptian governorates for the period 1992–2007, we investigate the effects of aggregate and sectoral foreign direct investment (FDI) on Egypt’s economic growth. We distinguish between FDI in the manufacturing, agriculture and service sector. The similarity of governorates in terms of institutional characteristics like culture, language and legal framework and the consistency of the data collection process enables an effective estimation of the effect of FDI on Egypt’s economic growth. Employing General Methods of Moments (GMM) panel estimations, we find that neither aggregate nor sectoral FDI has an unconditional effect on economic growth. We also reject human capital as a channel of absorptive capacity, but reveal an interesting effect of FDI in the service sector on economic growth in interaction with domestic private investment (DPI). Service FDI promotes economic growth only if the host governorate has a minimum threshold of DPI to absorb foreign knowledge and technology.

KEYWORDS:

• Foreign direct investment
• sectoral FDI
• absorptive capacity
• economic growth
• Egypt

JEL Classifications:

• F21
• F23
• F43
• O47
• O53

Acknowlegements

We thank Bernd Hayo, Moamen Gouda, Michael Kirk, the editor, two anonymous reviewers, participants of the brown-bag research seminar at Philipps-University Marburg, participants of the 2017 International Conference of the Korea Association of Middle East Society and participants of the 9th International Conference of the Korean Association of International Studies, for helpful comments and suggestions. We would also like to thank all the people that helped facilitate, directly or indirectly, the data collection in Egypt.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. A similar argument is made by Ford, Jonathan, and Elmslie (2008), who analyze the impact of FDI on economic growth at the US states’ level. Similarly, Zhang (2001) and Yao and Wei (2007) investigate FDI effect on economic growth in China using a panel of Chinese provinces.

2. http://www.middleeasteye.net/news/high-hopes-egypts-economy-conference-bombs-rock-cairo-1705596510 (Accessed 2 November 2015).

3. See Baldwin, Braconier, and Forslid (2005) for an endogenous growth model where MNCs directly affect the endogenous growth rate via technology spillovers. Productivity spillovers take place when the entry or presence of foreign firms generate productivity or efficiency benefits for domestic firms, whereas the foreign firm is not able to fully internalize these benefits (Blomström and Kokko 1998). It is usually assumed that a foreign firm that enters a certain domestic market has some superior knowhow that allows it to compete in a foreign country and/or compensate for the better knowledge of domestic markets by domestic firms (Graham and Krugman 1991; Blomström and Kokko 1998; Borensztein, De Gregorio, and Lee 1998; Ford et al. 2008).

4. See, for example, Rodriguez-Clare (1996) for a theoretical model of FDI spillovers via backward linkages through the intensive use of intermediate products by foreign firms. For further examples of spillover channels through backward and forward linkages, see Javorcik (2004).

5. Moreover, spillovers from foreign firms’ presence do not have to be positive. At least in the short run, the market entrance of new foreign companies may decrease the productivity of domestic companies in the same industry through market stealing (or competition) effect (Aitken and Harrison 1999; Marcin 2008).

6. In a theoretical three-period model, Markusen and Rutherford (2005) show that the speed and degree of positive spillovers from FDI positively depend on the absorptive capacity of the host country.

7. UNCTAD (2007) sheds light on potential linkages and knowledge spillover channels of tourism FDI with the local economy.

8. Producer services are, for example, banking, finance, consulting, insurance, real estate, engineering, accounting or legal services as defined by UNCTAD (2004). In addition, ICT services have become one of the most important business services. See Markusen, Rutherford, and Tarr (2005) for a theoretical model where producer services are modeled as intermediate inputs.

9. Fillat and Woerz (2011) further show differences across industries when using human capital as a further channel of absorptive capacity.

10. FDI data by the CBE are not reported at the governorate level.

11. In fact, separating the effect of greenfield FDI from mergers and acquisitions (M&A) while investigating their effect on economic growth is consistent with literature recommendations and findings by Wang and Wong (2009) and Harms and Méon (2014). However, data on M&A in Egypt are not available at the governorate level to test hypotheses on differences in growth effects by FDI mode.

12. The stock of non-petroleum greenfield FDI accounted for 23% of GDP in 2007.

13. According to GAFI, the category ‘other services’ includes consultancy, trade services, petroleum services, education and human resources, health as well as general services. We do not have data from GAFI on the breakdown of manufacturing FDI and agricultural FDI. However, according to the American Chamber of Commerce in Egypt (2008), manufacturing FDI stock in 2007 consists of FDI in chemicals (27%), building materials (21%), food and beverages (16%), pharmaceuticals (11%), engineering (9%) and others (15%). In our dataset, FDI in agriculture includes land reclamation and cultivation, livestock, poultry farming, fish and slaughter houses. According to FAO (2011), the majority of agricultural FDI is directed to land reclamation and cultivation projects.

14. For more details on the distribution of aggregate and sectoral FDI at the governorate level, see Hanafy (2015a).

15. The governorate dummy also captures the urbanization rate as well as port and airport availability at the governorate level, since available data show that these variables have hardly changed over time during our sample period (Hanafy 2015b).

16. In fact, some studies justify the use of FDI flows by the lack of data on FDI stock (e.g. Borensztein, De Gregorio, and Lee 1998).

17. Since the coefficient of a simple interaction of FDI and human capital would measure the effect when human capital equals zero, which is unlikely, we follow the valuable suggestion of one anonymous referee and mean-center the human capital variable in the interactive term. This provides an easier interpretation of results based on the marginal effect of FDI for different values of human capital.

18. The addition of an interactive term may lead to multicollinearity as the interactive term tends to be strongly correlated with the original variables that were used to construct it. In our case, the two investment variables lnFDIst and lnDPI are strongly correlated with a correlation coefficient of 0.78. Similarly, FDI stock and DPI are strongly correlated at the sectoral level. In order to alleviate this problem, we follow the approach by Burill (2007) and Azman-Saini, Baharumshah, and Law (2010) in orthogonalizing the interactive term using the following two-step procedure. First, we regressed the interactive term (FDIxDPI) on the FDI and DPI variables. Second, we used the residuals from the regression of the first step to represent the interaction term. We call the new interactive term (FDIxDPInew). We would like to thank two anonymous reviewers for helpful comments regarding the creation of the interactive terms.

19. Egypt currently consists of 27 governorates. However, we merged the governorates Qena and Luxor for the sake of data consistency, as most data sources do not report separate data for Luxor for our sample period. Luxor was only split from the governorate Qena to become a single governorate in 2010.

20. We opt for three-year periods due to the limited time span of our sample, similar to Aykut and Sayek (2007). Extending the time periods would not allow us to run a dynamic GMM model with the necessary statistical tests. The sub-periods for our dependent variable are constructed for the intervals 1992–1994, 1995–1997, 1998–2000, 2001–2003 and 2004–2007. Only the final time period consists of four years.

21. Prior to running our regressions, we studied the correlation matrix of all variables, which we do not include in this paper due to space constraints.

22. We would like to thank one anonymous reviewer who suggested the use of a GMM estimator.

23. We use four lags only as instruments to limit the number of instruments. However, our main findings are also robust to using three lags. We report the number of instruments under each estimation model of our results. We run the regressions using the statistical package gretl, which reports results similar to stata’s xtabond2 routine by Roodman (2009a) (see the gretl user guide).

24. Gretl reports the Sargan test which is similar to the Hansen test in stata’s xtabond2 according to the software manual and the lead software developer Allin Cottrel. We thank Allin Cottrell for the prompt reply to our inquiry.

25. This applies for all our reported models in this paper.

26. Note that this result remains unchanged when we run the regression without DPI or without the human capital variable.

27. For example, Hanafy (2015b) finds that labor force education is not a significant determinant of FDI location in Egypt.

28. The manufacturing FDI and DPI variables and their interaction term are also not jointly significant F(3,84) = 0.37.

29. That is: ∂y/∂ln(FDIservst)= -0.129+0.087*ln(DPIserv) = 0; ln(DPIserv) = 0.129/0.087; DPIserv = exp(0.129/0.087) = 4.02. See Ford, Jonathan, and Elmslie (2008) for a similar approach of a different FDI interactive term.

30. ∂y/∂ln(FDIservst)= -0.043+0.049*ln(DPI) = 0; ln(DPI) = 0.043/0.049; DPI = exp(0.043/0.049) = 2.39

31. In 10 of the 18 governorates, this DPI threshold is reached in all time periods of our analysis (Alexandria, Cairo, Giza, Matruh, North Sinai, Port Said, Red Sea, Sharkia, South Sinai and Suez). In 4 of 18 governorates the threshold is reached in most time periods (Aswan, Damietta, Ismalia and Menoufia), while 4 out of the 18 governorates meet this DPI threshold only in one, mostly the final, time period (Assuit, Kalyoubia, New Valley and Qena).

32. ∂y/∂ln(DPI)= -0.070+0.049*ln(servFDIst)=0; ln(servFDIst)=0.070/0.049; servFDIst=exp(0.070/0.049)=3.88.

33. The interactive term of FDI flows in services with aggregate DPI is positive and significant at the 5% level. The interactive term of FDI flows and DPI in services is positive but is insignificant at conventional significance levels (p-value = 0.15). However, combining both results highlights our finding on the cross-sectoral complementarity between FDI in services and DPI in the other sectors.

34. See Golub (2009) for an International Index of FDI restrictions in the services sector. Accordingly, the Middle East, East Asia and South Asia rank as the most restricted.

35. A consideration of special fiscal incentives for FDI in services would need to be further analyzed.

36. Similarly, Blonigen et al. (2007) use linear interpolation to fill missing data.

 

Additional information

Funding

Shimaa Hanafy acknowledges financial support from the Hankuk University of Foreign Studies.