1 Hubbell's ‘Unified Neutral Theory of Biodiversity and Biogeography’ (UNTB) has generated much controversy about both the realism of its assumptions and how well it describes the species abundance dynamics in real communities. 2 We fit a discrete-time version of Hubbell's neutral model to long-term macro-moth (Lepidoptera) community data from the Rothamsted Insect Survey (RIS) light-traps network in the United Kingdom. 3 We relax the assumption of constant community size and use a hierarchical Bayesian approach to show that the model does not fit the data well as it would need parameter values that are impossible. 4 This is because the ecological communities fluctuate more than expected under neutrality. 5 The model, as presented here, can be extended to include environmental stochasticity, density-dependence, or changes in population sizes that are correlated between different species.