This paper investigates the relationship between economic growth and energy consumption through non-linear causality tests. Eight developing countries from Europe and Central Asia spanning the period 1993 to 2008 are selected for the purpose of panel empirical analysis. Panel unit root and panel cointegration tests with and without considering cross section dependency (CD) problems are implemented. Next, linear panel cointegration tests are employed and, finally, a two-regime Dynamic Panel Smooth Transition Vector Error Correction (PSTRVEC) model is estimated for testing the presence of non-linear short- and long-run causality. To this end, a new estimator, called the Dynamic Non-linear Pooled Common Correlated Effect Estimator (DNPCCEE) is proposed. The empirical findings indicate that short and long-run causalities are regime-dependent.