The ABARE (now ABARES) survey of irrigation farms in the Murray–Darling Basin began in 2006–07 and provides a comprehensive farm-level panel dataset, which, to date, has seen limited econometric analysis (Ashton et al. 2009). At present, three complete years of irrigation survey data are available: 2006–07, 2007–08 and 2008–09. In each year, approximately 850 farms are sampled. As with the ABARES broadacre surveys, the irrigation survey is a rotating (unbalanced) panel dataset. This study makes use of the irrigation survey data to estimate production functions at both the farm and enterprise (crop/livestock activity) level. In addition to the traditional categories of input use (land, labour, capital and materials), the study incorporates measures of water use, tree and vine capital and local seasonal rainfall. The analysis incorporates fixed effects models to take advantage of the survey’s panel structure, as well as consideration of potentially endogenous inputs via instrumental variable methods. The study focuses on the short-run marginal revenue product of water implied by the estimated production functions. The results provide an encouraging demonstration of the kind of analysis that can be undertaken with the irrigation survey dataset. The estimated marginal product curves showed horticulture farms to have the steepest marginal product curve and broadacre farms to have the most elastic. There remain a number of promising areas for potential future research using the dataset, particularly if the survey were to continue for a longer and more representative sample of years.