A simulation approach to modelling temporary populations

Elin Charles-Edward, University of Queensland

Conventional population estimates provided by statistical agencies for regions and localities refer to a single point in time but it is widely recognised that actual population numbers fluctuate over the course of a year. Direct measurement of temporary populations has proven prohibitively expensive, while indirect estimates based on symptomatic data have met with limited success. We propose an alternative approach based on a stochastic simulation model that couples seasonality of mobility with duration of stay and is operationalised using Monte Carlo methods. Drawing on data from the Australian National Visitor Survey, we identify families of distributions fitting seasonality and duration. Random samples are then drawn from these, and combined to produce estimates of ‘visitor nights’ for 20 selected municipalities. Results are promising with the synthetic estimates correlating closely with observed data from the Australian Small Area Accommodation Survey.