where n are countries for which Technologyi data is available, and 1-n are countries for which Technologyi data is not available, but a partial data series, called Networked Readiness Index (NRIi), is available. For comparison purposes the highest possible score for both Macro and Technologyi is 5.0.
To control for additional variations Version 3 adds regional dummies (Eastern, Western, Northern, and Southern Africa). Version 4 assumes that Technologyi can be decomposed into two components, viz., Cellfone and Internet. The “Internet” variable is the ratio of internet hosts to internet users, and “Cellfone” is per capita cell phones. Both measure the intensity of use of new technologies. 5 Version 5 adds regional dummies to Version 4. Versions 1–5 constitute the fundamental versions of the basic model in (2) or (3). Auxiliary Versions 6–9 are not essential estimations; rather, they are indirect checks on the robustness of the fundamental estimations. For example, Version 6 drops Macro, Version 7 drops Openness, Version 8 drops Internet, and Version 9 excludes Cellfone. The next subsection presents and discusses the results.
Results
Tables 2.1 and 2.2 report estimation results of fundamental and auxiliary versions, respectively. Excluding the constant term the second column of the first table shows, for example, that macroeconomic policy has the largest effect on per capita real GDP across these African countries. Per capita capital and openness to trade are also positively related to per capita real GDP. More than one third of a percentage point increase in y results from a one percentage increase in k (capital) and τ (trade).
Table 2.1 Fundamental Determinants of Per Capita GDP across African Countries, 2004–2005
Table 2.2 Auxiliary Determinants of Per Capita GDP across African Countries, 2004–2005
A large constant term suggests that some other determinants of real per capita GDP are missing from Version 1. The regression results in the next column of the table add a measure of technology using a technology index calculated by the World Economic Forum (2005). In this case while the coefficients of k i and Openness remain largely unchanged, the estimate of Macro more than doubles and the constant term falls. A major finding is a huge, negative, and statistically significant impact of technology on the per capita real GDP of these African countries. This implies that the low level of technology harms real income determination in these countries. In fact, the magnitude of this negative coefficient increases significantly when the regression includes regional dummies, suggesting that the sign is not spurious. Southern Africa has the largest regional dummy and Eastern Africa the smallest. With regional dummies included macroeconomic policy becomes even more important than in previous versions. Openness to trade remains positive for GDP, but it is no longer statistically significant.
Columns 4 and 5 of Table 2.1 consider explicit components of technology: the intensity of internet use (Internet) and mobile phone use per capita (Cellphone). 6 Parameters for both variables are large and statistically strong. Such impacts are unaltered by the inclusion of regional dummies, although in this case trade openness becomes insignificant. Even so, the impact of per capita capital at about 0.37 is robust across all five fundamental versions. The explanatory power ranges from 21 percent to 54 percent, not unreasonable for cross-section regressions and a relatively small sample.
The coefficients of Internet (3.3161) and Cellfone (5.0941) are large. This raises a question about whether or not these variables are overestimated. The results in Table 2.2 indirectly address that concern. For example, Version 6 to Version 9 retain k i and regional dummy variables as the key independent variables and drop one of the remaining variables from each regression. Version 6 drops Macro; Version 7 excludes Openness, Version 8 leaves out Internet and Version 9 goes without Cellfone. Consequently, there is a remarkable improvement in summary statistics in Table 2.2 compared to Table 2.1; both explanatory and predictive power of the regressions, for instance, increase. However, there is no major gain in the technical efficiency of individual parameters. In addition, the values of the log-likelihood functions decline. This seems to indicate that the fundamental versions in Table 2.1 are more informative than the auxiliary versions in Table 2.2.
Concluding Remarks
The objective of this essay is to quantify the impact of technology on per capita real GDP of 46 African countries in 2004/2005. The results are encouraging both for policy and further research. The first estimation begins with per capita capital, trade openness, and macroeconomic policy index as the main independent variables, assuming homogeneous technology across countries. These results show that 12 percent of variations in per capita real GDP are explained by those independent variables. A one percentage rise in capital and trade openness contributes more than a third of one percentage increase in GDP, and for macro-policy the effect is three-fourths of a percentage change.
However, the large constant term motivates the inclusion of a country-specific measure of technology. The negative impact of the technology variable means that technology is a major constraint of the performance of African countries. This conclusion is consistent with previous observations of a negative total factor productivity (TFP) and or “Africa dummy.” Since TFP is a catch-all “measure of our ignorance” (Abramovits 1956), subsequent estimations assess the effects on per capita GDP of the intensity of use of two newer technologies: Internet and Cellphone. Along with the macroeconomic environment, these two variables explain real GDP per capita across countries well. However, there are considerable regional variations.
A number of implications for research and policy emerge from the concluding remark. For instance, the results suggest a need for improved technology. Increasing the distribution and use of internet and cellphone technologies is one way of doing just that. These new technologies have a good chance of rapid diffusion because “social capability” and “technological congruence” already exist in these countries and the cost of diffusing newer technologies is lower than the cost of adopting and sustaining older technologies. As for further research a key implication of the results is a need to investigate the impacts of old technologies, increasing the sample size, and using alternative modeling and estimations techniques, and better data.
NOTES
1. Kenyan journalist B. Wainaina in “How to Write about Africa I and II” (2005; 2006/2007) argue that there is a standard way people expect about Africa, and unless anyone conforms to that standard, he or she will never be listened to.
2. 1. South Africa, 2. Kenya, 3. Tanzania, 4. Zimbabwe, 5. Mauritius, 6. Cote d’Ivoire, 7. Morocco, 8. Egypt, 9. Mozambique, 10. Namibia, 11. Uganda, 12. Botswana, 13. Zambia, 14. Rwanda, 15. Swaziland, 16. Nigeria, 17. Benin, 18. Democratic Republic of Congo, 19. Madagascar, 20. Senegal, 21. Cameroon, 22. Burkina Faso, 23. Guinea, 24. Tunisia, 25. Sierra Leone, 26. Mali, 27. Togo, 28. Libya, 29. Congo, Republic, 30. Ghana, 31. Malawi, 32. Angola, 33. Ethiopia, 34. Sudan, 35. Burundi, 36. Chad, 37. Somalia, 38. Central African Republic, 39. Lesotho, 40. Gabon, 41. Reunion, 42. Seychelles, 43. Mauritania, 44. Gambia, 45. Cape Verde, 46. Niger.
3. The Africa dummy and Africa TFP are not directly comparable because of different models and estimators. However, the negative signs of the coefficients have been revealingly consistent.
4.