Investigating geography and institutions as determinants of foreign direct investment in Africa using panel data

Abstract

This article uses a cross-country econometric approach to identify the determinants for foreign direct investment (FDI) in Africa. The contribution is 3-fold. Firstly, we recognize that the estimation techniques used elsewhere, such as ordinary least squares, may be flawed. We therefore use a dynamic one-step generalized method of moments (GMM) estimator due to Arellano and Bond (1991). The GMM-estimates identified a number of robust determinants of FDI, namely government consumption, inflation rate, investment, governance (political stability, accountability, regulatory burden, rule of law) and initial literacy. It is concluded that geography does not seem to have a direct influence on FDI flows to Africa. Neither market-seeking nor re-exporting motives of FDI seem to dominate, with different policy instruments being significant in the different specifications. This does not discount the importance of good policies, but probably indicates the importance of good policies made by good institutions. Institutions, in the form of political stability showed up as a significant determinant of FDI.

Acknowledgement

We are grateful to four anonymous referees for useful comments on earlier drafts. All errors and omissions remain our own.

Notes

¹ Foreign direct investment is when an investor based in one country acquires an asset in another country with the intent to manage that asset (Stoker, 2000, p. 117). It can take the form of equity capital (through mergers and acquisitions or Greenfield investments), reinvested earnings and borrowing or lending between MNEs and affiliates.

² Jenkins and Thomas (2002, p. ii) point out ‘the economic literature on private capital formation in developing countries is largely concerned with the issue of uncertainty and risk as disincentives to investment’.

³ It is assumed that there is no second-order autocorrelation in the differenced idiosyncratic error term.

⁴ The IV approach leads to consistent but not necessarily efficient estimates of the parameters because it does not make use of all the available moment conditions (Baltagi, 1995, p. 126).

⁵ The asymptotic standards errors from the two-step GMM estimator have been found to have a downward bias (Blundell and Bond, 1998).

⁶ Warner (2002) points out that recent research on economic growth and convergence has ‘framed the issue as a competition between geography and institutions’.

⁷ They found that there is no statistically significant relationship between growth and investment in Africa once an outlier country such as Botswana is excluded (Devarajan et al ., 2003, p. 547).

⁸ In all parts of the world, economic development in tropical zones lags far behind that in temperate zones. The underlying reason is a backlog in productivity growth. Differences in productivity growth and innovation between temperate and tropical zones reflect the interplay of a number of factors. First, many kinds of agricultural and construction technologies do not transfer well between ecological zones. Second, temperate zones have long had much higher rates of endogenous technological change than the tropics. Third, the tropics pose inherent difficulties in agriculture and public health. Fourth, the tropics are disadvantaged because they are far from the large mid-latitude markets.

⁹ The costs of international transport for landlocked developing countries are on average 50% higher than for coastal economies (Radelet and Sachs, 1998).

¹⁰ For a critical discussion of the concept of ‘institutions’ see e.g. Nelson and Sampat (2001).

¹¹‘Institutions are the rules of the game; organizations are the players’ (North, 1998, p. 2).

¹² An indicator of trust in the financial system is provided by the willingness of people to entrust their savings to financial institutions, proxied by the ratio of (M2-M1) to M2.

¹³ See www.worldbank.org/wbi/governance/wphtm#governance

¹⁴ Consequently ‘year’ was included as a regressor in the subsequent equations, although it was not found to be statistically significant.

¹⁵ A variable x_it in Equation 11 is said to be endogenous if E [x_itu_is ] ≠ 0 for s ≤ t, but E [x_itu_is ] = 0 for all s > t.