The determinants of residential water demand: empirical evidence for a panel of Italian municipalities

Abstract

We present empirical evidence on the determinants of residential water demand for one Italian region, Emilia-Romagna, by using municipal panel data. The estimated water demand price elasticity is negative, showing values between −0.99 and −1.33, never significantly different from one, if we consider different specifications without and with additional socio-economic factors. Income results associated to a positive elasticity, though lower than one. The role of other socio-economic territory-specific determinants is less relevant, with the exception of altitude. The relative high value of price elasticity is deemed consistent with the higher level of regional water prices compared to the national average.

Notes

¹ Considering the price of the central block in the increasing-blocks tariff structure (three blocks are generally present).

² Official time series data on municipal income are not available. We estimated municipal income for the four years by using the Treasury 1998 municipal taxable income data, re-parameterized on the basis of yearly provincial income data provided by ISTAT (National Institute of Statistics).

³ It is worth noting that endogeneity will be a potential problem in following years, where the full transformation of water public utilities into private managed firms will correlate prices and costs stronger than in the past. For this reason, we believe that in the years 1998–2001 the water price could be considered as exogenous, showing an increasing trend following the implementation of the 1994 Bill, but still largely determined by political factors external to the model.

⁴ Specifications including the squared tariff were regressed but lead to less robust results.

⁵ Since the raw utility dummies do not vary over time, we have included a set of interaction terms between utility dummy variables and users (time variant).

⁶ To avoid collinearity problems and to check the marginal impact of each element on the ‘baseline model’, we first regress specifications with price, income, and one element at a time. We also control for water utilities and time period effects. The addition of water utilities interaction terms in specifications 3 to 9 reduce the significance level of the income coefficient, although in specifications 8, 3 and 4 the income effect remains significant. The addition of time period effects in specifications 3 to 9 leads to a significant income effect in model 9 only.

⁷ The specification including the interaction terms with the share of rural areas has not been considered because of the high correlation between the interaction term and population.

⁸ See, for instance, Arbues-Gracia et al. (2002), Martinez-Espineira (2002), Martinez-Espineira and Nauges (2004). Nevertheless, Dalhuisen et al. (2003) in their meta-analysis of water demand empirical studies, show price elasticity values in a range with a negative upper value much higher than one.