Spatial analysis of municipal water demand: a panel data approach

While the municipal water demand literature is well developed, one area that is understudied is the impact of spatial effects. After controlling for factors shown to impact demand, this study applies spatial econometric methods via a spatial weights matrix to a panel municipal water consumption data set. While diagnostics suggest the presence of spatial lag and spatial error, thus indicating the potential usefulness of spatial empirical methods, several important pitfalls must be acknowledged. First, the application of spatial weights in a panel setting is computationally intensive, especially when the number of time periods or observations is large, and perhaps necessitates aggregation. Second, because most users in a municipality are likely to be subject to similar utility action, climate, etc., a spatial lag signal may be spurious. Third, because premises served by the utility may enter or exit the data set through time, the requirement of balanced panels requires careful consideration. Fourth, if the option to use premises-level (or similar) data or aggregated data is available, it is typically advisable to use premises-level data despite the possible presence of spatial effects.