Myeloproliferative neoplasms (MPNs) are an uncommon group of blood cancers that, if untreated, result in an increased risk of haemorrhagic event or thrombosis. Unlike other cancer types, diagnosis of MPNs requires a combination of microscopic, clinical and genetic evidence, which provide unique challenges given the typical notification processes of cancer registries. This, and the relatively recent advances in diagnosis and revision of the World Health Organization diagnostic criteria, may result in under-diagnosis or under-reporting of MPNs. We used population-based cancer registry data from the Australian Cancer Database and modelled the incidence and survival of MPNs between 2007 and 2016 using generalised linear models and Bayesian spatial Leroux models. Substantial evidence was found of spatial heterogeneity in the incidence of MPNs and significant differences in incidence and survival by state or territory. States with lower incidence tended to have poorer survival, suggesting that some less severe cases may not be diagnosed or notified to the registries in those states. Population rates of genetic testing and percentages of records diagnosed using bone marrow biopsies did not explain the differences in incidence by state and territory. It is important to determine the key drivers of these geographical patterns, including the need to standardise diagnosis and reporting of MPNs.