Forecasting accuracy of behavioural models for participation in the arts

► We estimate attendance at cultural activities using count data regressions and assess their forecasting performance by computing Brier scores. ► We use data from a representative sample of the adult population of the USA that exhibits lots of zeros in the dependent variable (SPPA2002). ► Brier scores bring evidence for the good forecasting properties of such models. ► Our models are especially useful to predict attendance for cultural activities with a larger variability. ► They prove to be better for non-attendants and moderate attendees, the broadest categories.