Spatial determinants of urban residential water demand in Fortaleza, Brazil

Abstract

This paper aims at estimating a residential water demand function for the city of Fortaleza, Brazil, considering the potential impact of spatial effects on water consumption. The empirical evidence is a micro dataset obtained by means of a household survey of water consumption in 2006. We estimated three econometric models, which had as explanatory variables the average price, the difference, income, number of residents and the number of rooms, under different spatial specifications: the spatial error model (SEM), the spatial autoregressive model (SAR), and finally, the spatial model autoregressive moving average (SARMA). The results show that spatial autocorrelation exists in two forms (error and lag), indicating that the SARMA specification is the "best"as shown by a series of tests. Such results are in contrast to that suggested by Chang et al. (2010), House-Peters et al. (2010), Franczyk and Chang (2008), Ramachandran and Johnston (2011), who favored the use of the SEM model. Our results point out to the necessity of considering spatial effects in the estimation of residential water demand, since the absence of spatial effects is a key misspecification error, underestimating the effect of important variables such as average price and number of residents, while overestimating the effect of other variables such as income and number of rooms.