The goal of this study is to develop an estimation procedure for the purchase cost of agricultural lands that are not necessarily on the market. To do so, we implement the first stage of the hedonic price method to estimate the price of agricultural lands that have been the object of real estate transactions. We work at the original scale of the parcel. The estimation results are then employed to predict the price for agricultural lands that have not been the object of transactions. This stage needs the construction of explanatory variables at the level of entire parcels. To reduce the acquisition cost of information, we focus on the French administrative department of Côte-d’Or (8,763 km2 and 983,116 cadastral parcels). We show that the predictor proposed by Meulenberg (1965) best suits our dataset.