National early warning systems of Sahelian countries are mainly based on biophysics models to predict agricultural production shortages and prevent food crisis. The objective of this paper is to show that cereal market prices also bring useful information on future food availability that could complete current early warning disposals. Indeed, at any point in time prices are informative not only on the state of present food availability but also on agents' expectations about future availability. The research is based on the exploitation of the statistical properties of price series. It aims at detecting movements in the prices trend signalling a coming crisis. We first identify markets playing a leading role in price formation at the national and regional level through the estimation of VAR models. The second step consists in identifying crisis periods and then the price shocks characterises during the period which precedes a crisis. This analysis leads to the identification of early warning indicators whose relevance is tested using panel data qualitative choice models. The data set encompasses 50 markets belonging to three countries: Mali, Burkina Faso and Niger, over the period 1990-2008. The results show that past price shocks on a few number of leading marketplaces can help preventing coming crises.