The market volatility index (VIX) and the Baltic dry index (BDI) are evaluated as two important economic indicators, the former as being a gauge of investor's fear and risk and the latter as being a reflection of costs associated to shipment of dry cargo. However, the empirical analyses aiming at achieving future forecasts suffer drastically due to inherent threshold effects the leptokurtic distribution The purpose of the paper is to propose the application of threshold models that allow regime dependent dynamics in the conditional mean and variance processes simultaneously to overcome this difficulty. Accordingly, four different threshold GARCH specifications are evaluated: TAR-GARCH, TAR-TGARCH, TAR-TR-GARCH and TAR-TR-TGARCH which allow more complex threshold behavior as one moves from the former to the letter. The empirical findings show the following, i. all of the four models provide significant improvements in terms of out-of-sample forecasting, ii. the threshold effects are dominant in both series and the proposed threshold models are capable to overcome the ARCH effects, iii. without the simultaneous modelling of threshold effects in the mean and variance processes, additional within regime specifications are needed to account for the negative and positive innovations, iv. BDI and VIX are two indexes that should be modeled with caution and once controlled for the threshold effects, they possess significant potential to be taken as future leading economic indicators.