Telecommunications and economic growth: an empirical analysis of sub-Saharan Africa

Abstract

We examine the effect of mobile cellular phones on economic growth in sub-Saharan Africa where a marked asymmetry is present between land line penetration and mobile telecommunications expansion. This study extends previous research along two important dimensions. First, we allow for the potential endogeneity between economic growth and telecommunications expansion by employing a special linear Generalized Method of Moments (GMM) estimator. Second, we explicitly model for varying degrees of substitutability between mobile cellular and land line telephony, so that greater expansion of mobile telecommunications can have a different impact whenever the level of land line penetration differs. We find that mobile cellular phone expansion is an important determinant of the rate of economic growth in sub-Saharan Africa. Moreover, we find that the contribution of mobile cellular phones to economic growth has been growing in importance in the region, and that the marginal impact of mobile telecommunication services is even greater wherever land line phones are rare. Given the low cost of mobile telecommunications technology relative to other broad infrastructure projects, especially land line infrastructure, we advocate that mobile telecommunication services be encouraged in the area.

Notes

¹ For discussions on direct and indirect benefits to economic growth of telecommunications sector, see Tisdell (1981), Leff (1984), Antonelli (1991), Cronin et al. (1993a) and Greenstein and Spiller (1995).

² Throughout the rest of the article, we opt for more commonly used terminologies, such as cellular phones and land-line phones, rather than the more formal but cumbersome ‘mobile telecommunications services’ and ‘land-line telecommunications services’.

³ Roodman (2006) provides with a full discussion on the derivation of the Arellano and Bover (1995)/Blundell and Bond (1998) difference GMM estimator. Our discussions on the modification of key assumptions about panel data approach draws heavily upon Roodman (2006).

⁴ As is standard in GMM estimation, the joint validity of the instruments is tested. The Hansen test of overindentifying restrictions is satisfactory for all regressions. Also, in all regressions, the Arellano–Bond test suggests that we do not reject the null hypothesis of no second-order autocorrelation in the first-differenced residuals at the 5% level.

⁵ Note that, in Regression (2), the estimated coefficient of LANDPHONES100 is statistically insignificant but larger than that of CELLPHONES100. However, its statistical insignificance should not lead to a conclusion that conventional land-line telephony is no longer relevant to the region's economic growth, especially from a theoretical point of view. A relatively large SD of the variable only suggests that the regression provides weak empirical evidence against the null hypothesis that the coefficient is not significantly different from zero. From 2000 to 2006, the mean value of LANDPHONES100 in the region varied little, from 2.62 to 3.5.