Terrorism and Regional Integration in Sub-Saharan Africa: The Case of the CFA Franc Zone

Additional evidence on the empirical significance of terrorism in Sub-Saharan Africa is provided by Elu and Price (2012) who find that terrorism possibly constrains living standards by diverting financial inflows – remittances – that would otherwise be used to finance the drivers of economic growth toward the financing of growth-reducing terrorist acts.

START data are publicly available at http://​www.​start.​umd.​edu/​gtd/​.

As of January 1, 1999 the currency arrangement among CFAZ member countries changed as the currency of these countries – the franc – was now pegged to the euro.

Penn World Table 6.​3 data are publicly available at http://​pwt.​econ.​upenn.​edu.

See Zorn (2001) for an overview of estimating GEE specifications when repeated observations data are possibly correlated. A GEE specification is a semiparametric approach where for k explanatory variables and j = 1,2 ⋯ C possibly correlated clusters of observations on the number of terrorist incidents in country i (N _i ), parameter estimates follow from solutions to a quasi-score function of the form:where G _i  = λ _i /β, and for scale parameter φ = 1, V _i  = [(A _i )^1/2 R _i (α) (A _i )^1/2]/φ. The A _i are T × T diagonal matrices with g (μ _it ) as the t ^th diagonal element, α is a vector of unknown parameters, and R _i (α) is the working correlation matrix. A Negative Binomial (NB) GEE specification follows from specifying λ _i  = h (θ _i β _i ), and for dispersion parameter r, N _i ~ NB (λ _i , λ _i [1 + r λ _i ]).

Our implementation of the Robins-Greenland-Hu test estimates the auxiliary OLS linear probability regression with robust standard errors. The difference between the tests in the static and dynamic specification differs in what is included in X _t – the dynamic specification includes the one period lag of terrorism, whereas the static specification excludes it.

This computation follows from the fact that given λ _it  = (β ^' θ + ε _it ), for a given change in the population-average level of remittances from I ^' to I ^' + δ, the change in N _it is given by [1 − (β _I ^' δ)] × \( \overline{\lambda} \).

This computation follows from scaling up the population average reduction in terrorism caused by CFAZ membership by the number – 36 – of CFAZ member countries in the sample.