The impact of foreign direct investment on horizontal export diversification: empirical evidence

Abstract

Using data on stocks of Foreign Direct Investment (FDI) from 131 countries spanning the years 1984 to 2004 and the number of products exported by each country, we examine the effect of FDI on horizontal export diversification. To quantify the effects, we utilize parametric (quantile) and semi-parametric econometric methods. Results from both approaches indicate that, in general, an increase in the stock of FDI enhances the horizontal diversification of exports. The actual magnitude of the effect however, varies greatly across countries depending on the existing stock of FDI and stage of diversification, giving rise to an almost inverted U-shaped relationship. A further look at our results provides useful insights on the circumstances under which FDI may aid or inhibit the horizontal expansion of exports.

Keywords:

• export diversification
• foreign direct investment
• quantile regression
• nonparametric estimation

JEL Classification:

• F21
• F12
• F36
• F10

Acknowledgements

We would like to thank the anonymous referees of this journal for their valuable suggestions. All remaining errors are exclusively ours.

Notes

¹ This can be achieved among other things by increasing the share of a country's exports in the dynamic rather than stagnant products, and enhancing the overall competitiveness of a country in international markets.

² These include volatility and instability in foreign exchange earnings and investment planning, inflation, capital flight and undersupply of investments by risk-averse investors. The long term effects include a secular decline in the terms of trade that in turn exacerbates the short run effects (Lederman and Maloney, 2002). According to Collier and Dollar (2002), the later may also include political risks such as worsened governance, civil war and conflict.

³ Results from such methods often underestimate, overestimate, or even fail to distinguish the effect of the variable of interest on real nonzero changes in the dependent variable (Terrell et al., 1996; Cade et al., 1999).

⁴ Division of the dependent variable series into the deciles (the 0.10, 0.20, 0.30, 0.40, 0.50, 0.60, 0.70, 0.80 and 0.90) is carried out by estimating the conditional (i.e. subject to the set of our main variable of interest – the stock of FDI) distribution of the dependent variable series. While the unconditional categorization of the series is possible as is often done using the Ordinary Least Squares (OLS) estimation strategy; as noted in Heckman (1979), the resulting coefficient estimates may be biased due to the truncation of the series. Application of the SQREG technique to the full data set avoids the potential sample selection bias.

⁵ We use 1 year lag of the stock of FDI in each country to allow enough gestation period for the effects of FDI to be materialized.

⁶ We have not employed a pure nonparametric approach because of computational and dimensionality problems. Although in a different context, Mukherjee and Pozo (2011) also used a similar approach.

⁷ We also estimate a simple OLS to examine to the relationship. For brevity, however, we do not report the results here.

⁸ The deciles distribution of the dependent variable presented are commonly generated conditional on the explanatory variables included in each model. However, the decile distribution presented here are unconditional. Thus, it is possible to observe discrepancies between the descriptive statistics of the variables reported here (unconditional deciles) and the ones that would be generated either from either of the unconditional or the conditional distributions the dependent variable series.

⁹ In the semi-parametric regression model, we do not account for time and/or country-specific effects as it would require the use of panel data techniques for which, to the best of our knowledge, no formal procedure has been developed.

¹⁰ Given the semi-log form of the estimating equation, the elasticity of the variable of interest is computed as .

¹¹ Our observation of negative or no effects of FDI on horizontal diversification of exports in several countries should not, in any way, be taken to imply that increased FDI stock is detrimental to export diversification. It is possible that increased inflow of FDI may enable the host economy to concentrate (specialize) more on the production of a good(s) in which the nation has a comparative advantage; thus enhancing the vertical diversification or sophistication of exports.

¹² We estimate both the Poisson and negative binomial fixed effects models to account for the count data nature of the dependent variable series. In addition, we employ the Generalized Method of Moment (GMM) estimator proposed by Arellano and Bover (1995) and Blundell and Bond (1998) to evaluate the robustness of our findings. Again, the effects of FDI on horizontal export diversification consistently remain positive and significant. All robustness estimation results can be obtained from the first author upon request.

¹³ Note that the Baltagi (2005) approach rejects the null hypothesis of poolability of the data only at 10%.