This article analyses the determinants of the discrete decision of a household on whether to participate in banana markets using the FIML bivariate probit method. The continuous decision on how much to sell or buy is analyzed by establishing the supply and demand functions while accounting for the selectivity bias.

Results indicate that buying and selling decisions are not statistically independent and the random disturbances in the buying and selling decisions are affected in opposite directions by random shocks. Transaction cost related factors such as geographical location of households, market information sources and travel time to the nearest urban centre do influence participation. Other factors such as labour availability, farming experience, gender of household head, off-farm income and the asset base of the household also affect the likelihood and intensity of participation.

Policies guiding central and local governments towards increased investment in rural infrastructure (i.e. feeder roads networks, trunk roads, telecommunication services and establishment of market places) can help reduce transaction costs and thereby improve participation of smallholder farmers in markets. Policies supporting group formation may lead to improving economies of scale and flow of information amongst farmers which may increase market participation.