Appendix 2 provides the summary statistics The correlation matrix is presented

Appendix 2 provides the summary statistics the

This preview shows page 7 - 9 out of 22 pages.

Appendix 2 provides the summary statistics. The correlation matrix is presented in Appendix 3. 2.2 Methodology Three empirical strategies are adopted to control for specific characteristics. First, Fixed Effects (FE) regressions are used to control for the unobserved heterogeneity. Then, the bite on endogeneity is increased with control for persistence in the dependent variable by employing the Generalised Method of Moments (GMM) which accounts both for simultaneity using instruments and further controls for the unobserved heterogeneity using time invariant omitted variables. Last, the Tobit model is employed to control for the limited range in the dependent variable. The panel FE model is presented as follows: t i i t i h h h t i t i t i W COCO CO IHD , , , 4 1 , 2 , 1 0 , , (1) where, t i IHD , is inclusive human development for country i at period t ; 0 is a constant; CO is a CO 2 emissions variable; COCO , is an interaction term representing the multiplication of two identical CO 2 emissions variables; W is the vector of control variables (education quality, private domestic credit, foreign aid and foreign direct investment); i is the country-specific effect and t i , the error term.
Image of page 7
8 Since we are employing an estimation technique that deals with interactive regressions, it is relevant to briefly discuss some pitfalls associated with interactive specifications. In accordance with Brambor et al . (2006), all constitutive variables should be involved in the specifications. Moreover, in order for the estimated interactive parameters to make economic sense, they should be interpreted as conditional or marginal effects. A plethora of reasons motivate the choice of an alternative system GMM estimation strategy, notably, it: (i) does not eliminate cross-country variations; (ii) controls for potential endogeniety in all regressors through instrumentation and accounts for the unobserved heterogeneity and (iii) mitigates potential small sample biases from the difference estimator (Asongu, 2013; Tchamyou et al ., 2018). Moreover, basic conditions for the use of the GMM strategy are also fulfilled, notably: (i) the condition for persistence is apparent because the correlation coefficient between the outcome variable and its first lag is higher than 0.800 which is the rule of thumb for establishing persistence in an outcome variable and (ii) the number of cross sections (or 44 countries) is higher than the number of periods in each cross section (or 13 years). In this study, we adopt the Roodman (2009a, 2009b) extension of Arellano and Bover (1995) which has been established to restrict over-identification and limit the proliferation of instruments (Love & Zicchino, 2006; Baltagi, 2008; Tchamyou, 2018). Hence, the corresponding specification is a two-step GMM with forward orthogonal deviations instead of differencing. We prefer the two-step to the one-step procedure because the latter is homoscedasticity-consistent while the former controls for heteroscedasticity.
Image of page 8
Image of page 9

You've reached the end of your free preview.

Want to read all 22 pages?

  • Summer '20
  • Dr joseph
  • Econometrics, Greenhouse gas, World energy resources and consumption, Millennium Development Goals

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture