This study is founded on economic development theory, in which economic diversification is viewed as driven by simultaneous changes in production, consumption and trade patterns (Schuh and Barghouti,1988; Barghouti et al., 1990; and Petit and Barghouti, 1992). Through the faster growth of sectors with high income elasticity of demand in addition to forces of unbalanced growth, economic diversification may be driven. Hence in the context of Sub-Saharan Africa and based on the arguments of this study, both ICT and financial development, working together drive the process of economic diversification.
Equation (1) is the general model specification for this study adapted from Andriannaivo and Kpodar (2012)
ECODIV=f(ICT, FINDEV, FININC, RLAW, GEXP, TROPEN, PSE, L, K, INF) (1)
Where, ECODIV= Economic diversification, ICT = information and Communications Technology, FINDEV = Financial development, FININC = Financial inclusion, RLAW = Rule of Law, GEXP= government expenditure, TROPEN = Trade Openness, PSE = Primary School Enrollment, L = Labour, K = Capital, INF =Inflation.
Economic diversification measure in this study is the Herfindahl-Hirschman index which is computed for this study. The index measures the concentration of the sectors in the total output of a country and while having a range between zero and one, values closer to one reflect lower diversification and vice versa. ICT indicators in this study are fixed broadband subscriptions (FIXEDBB), Fixed Telephone Subscriptions (FIXEDTSP), ICT good imports (ICTGIMPOT), Internet users (INTUSE), Mobile use (MOBUSE), and secure internet servers (SINTERNET) and rise in the indicators reflect higher levels of ICT development and vice versa. Financial development indicator used was private sector domestic credit to GDP ratio while financial inclusion indicator used was commercial bank branches per 100000 people. Hence equation (2) results.
HHI=f(FIXEDBB, FIXEDTSP, ICTGIMPOT, INTUSE, MOBUSE, SINTERNET, FINDEV, FININC, RLAW, GEXP, TROPEN, PSE, L, K, INF) (2)
Explicitly expressing equation (2) above and log-transforming large variables to ensure standardized regression estimates, equation (3) results.
Where, HHI = Herfindahl-Hirschman index, FIXEDBB = Fixed Broadband subscriptions (per 100 subscribers), FIXEDTSP = Fixed Telephone Subscriptions (per 100 subscribers), ICTGIMPOT = ICT good imports (% of total good imports), INTUSE = Internet users (per 100 subscribers), MOBUSE = Mobile subscriptions (per 100 subscribers), SINTERNET = secure internet servers, FINDEV = Financial development (in percentage), FININC = Financial inclusion (commercial bank branches per 100000 individuals), RLAW = Rule of Law (ranges between -2.5 and 2.5), GEXP= government expenditure (in Billions of US Dollars), TROPEN = Trade Openness (In percent), PSE = Primary School Enrollment (In Percentage), L = Labour (in Million), K = Capital (In Billions of US Dollars), INF =Inflation (in percentage). \(\epsilon\) = stochastic error term, i = country, t= time period. \({{\beta }}_{0}\) = constant, \({{\beta }}_{1}-{{\beta }}_{14}\) = marginal effects of independent variables. Log= Logarithm operand.
From equation (3) above, with the exception of rule of law which was sourced from the World Bank World Governance Indicators (WGI), all variables were sourced from the World Bank World Development Indicators (WDI). The model was estimated using Pooled Ordinary Least Squares regression, fixed effects estimation and two-step Generalized Method of Moments (GMM) estimation. The time frame covered by the data is between the year 2000 and 2019, while 37 of the 48 countries in sub-Sahara African (SSA) region form the sample. The selected countries are Botswana, Benin, Angola, Cabo Verde, Burkina Faso, Cameroon, Burundi, Comoros, Congo Republic, Cote d'Ivoire, Eswatini, Ethiopia, Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Lesotho, Madagascar, Malawi, Mali, Mauritania, Mauritius, Mozambique, Namibia, Niger, Nigeria, Rwanda, Senegal, South Africa, Sudan, Tanzania, Togo, Uganda, Zambia and Zimbabwe. The selection of these countries is primarily premised on the extent of data availability.