The paper uses logistic regression model to determine factors affecting probability of income increase in agricultural holdings specialised in milk production. Analysis had a dynamic character and covered the period between 2009 and 2011. Independent variables included both quality and quantity features associated with farm activities. The analysis showed that, in each year, the most important variables in the logit model concerned the utilised agricultural area and the number of dairy cows, which had a positive impact on probability of income increase. All estimated models are characterised by high quality and thus can be used to correctly classify agricultural holdings.