This article applies the Hinich portmanteau test to the daily return of the international Arabica Colombian coffee price the period 06/29/1990 to 07/01/2010. By splitting the data in windows and setting the null hypothesis on the time series, H : a stationary pure noise process with zero 0 tricorrelations and bicorrelations for each window. The main result of this paper is that there is significant evidence in 10 and 11 windows, respectively. This shows that there are statistical structures with nonlinear dependence. Thus it is difficult for economic agents to cope with forecast performance decisions in this market.