Sensitivity of Water Price Elasticity Estimates to Different Data Aggregation Levels

Abstract

The empirical literature on residential water demand employs various data aggregation methods, which depend on whether the aggregation is over consumption, sociodemographic variables, or both. In this study, we distinguish three dataset types-aggregated data, disaggregated data, and semi-aggregated data-to compare the consequences of using a large sample of semi-aggregated data vis-a-vis a small sample of fully disaggregated data on the water price elasticity estimates. We also analyze whether different aggregation levels in the sociodemographic variables affect the water price elasticity estimates when the number of observations is fixed. We employ a discrete-continuous choice model that considers that consumers face an increasing block price structure. Our results demonstrate that the water price elasticities depend upon the level of aggregation of the data used and the sample size. We also find that the water price elasticities are statistically different when comparing a large semi-aggregated sample with a small disaggregated sample.

The empirical literature on residential water demand employs various data aggregation methods, which depend on whether the aggregation is over consumption, sociodemographic variables, or both. In this study, we distinguish three dataset types-aggregated data, disaggregated data, and semi-aggregated data-to compare the consequences of using a large sample of semi-aggregated data vis-a-vis a small sample of fully disaggregated data on the water price elasticity estimates. We also analyze whether different aggregation levels in the sociodemographic variables affect the water price elasticity estimates when the number of observations is fixed. We employ a discrete-continuous choice model that considers that consumers face an increasing block price structure. Our results demonstrate that the water price elasticities depend upon the level of aggregation of the data used and the sample size. We also find that the water price elasticities are statistically different when comparing a large semi-aggregated sample with a small disaggregated sample.