Given its socio-economical and environmental impacts, drought is one of the most important and challenging problems in applied climate prediction. Assessment of drought predictability allows the creation of more realistic and accurate scenarios that can assist in mitigating natural disasters. One area of particular importance related to drought is the potential for predictions to lead to better water resources management strategies. Agricultural and hydrologic drought prediction skill is derived both from knowledge of initial hydrologic conditions, and from climate forecasts, and therefore there is the potential in practice to exploit both sources of predictability. At present, however, the relative magnitudes of predictability from these two sources are not well known. We use the University of Washington west-wide seasonal hydrological forecast system to evaluate the ability to forecast agricultural (soil moisture) and hydrological (runoff) drought recovery over Mexico. Based on the implementation of two hydrologic prediction techniques, the Ensemble Streamflow Prediction (ESP) and the NCEP Climate Forecast System (CFS) ensemble forecasts, we evaluate the skill, and factors that control the skill of hydrologic forecast over different hydrological regimes in Mexico. The experiments consist of the generation of monthly forecasts of Standardized Runoff Index (SRI) and Soil Moisture (SM) percentiles during the 2007 drought event in Mexico. Drought prediction skill is evaluated through comparison of ESP and CFS forecast and observed runoff and soil moisture percentiles with the model simulations done using the observed forcings through standard deterministic ensemble mean scores. The comparisons highlight differences in spatial variations in drought prediction skill across Mexico, and also help identify regions where drought prediction skill is low. Furthermore, the results contribute to understanding of linkages between drought events in southern and northwestern Mexico.