Does space matter? – a space-time model for night-time light intensity for South Africa

This study develops a space-time autoregressive moving average (STARMA) model for South Africa, modelling time histories of spatially located data. The central question is the relevance of space in the modelling process, i.e., does space matter? This is an interesting question since time series methodologies such as autoregressive integrated moving average (ARIMA) generally exclude space, only focusing on time. The study applies the STARMA methodology within the South African context focusing on the monthly median night-time light intensity of the 234 municipalities from 2013 to 2022. The results propose that it is indeed possible to include space in the modelling process and that the inclusion of space in the modelling process does matter. Based on the results obtained, it is evident that the STARMA model outperformed the ARIMA model. In this context, the findings demonstrate that spatial information contributes significantly to improving predictions, particularly for geospatially distributed data over extended periods. This enhancement underscores the importance of accounting for spatial dependencies, which are often overlooked in traditional econometric or ARIMA models. Space does seem to matter.