Brock University1
Extending municipal water demand forecasting capacities by incorporating behavioural responses to pricing and other policy measures
Steven Renzetti, Diane Dupont and James Price, Department of Economics, Brock University, St Catharines, Ontario Canada
INTRODUCTION
An 2008 AWWA survey indicated that most American water utilities forecasted water demands by multiplying future population estimates by historical per capita water use. The limitations of this approach have been discussed widely in the research literature and demonstrated in real-world cases (Billings and Jones, 2008). Specifically, it fails to account for other demand drivers such as income, prices and household appliance holdings. The lack of data, resources and institutional capacity pose important challenges to many water utilities in both developed and developing economies that prevent them from adopting more sophisticated and accurate forecasting methods.
This paper demonstrates the application of a user-friendly, spreadsheet-based demand forecasting and simulation tool that provides water utilities with the capacity to forecast near and medium-term demands while simultaneously accounting for planned or expected changes in important demand drivers. Using price and income elasticities estimated from water utility data, the forecasting program has the capacity to provide several concrete benefits to water agencies: (1) develop the capacity to produce more accurate water demand forecasts, (2) a sophisticated planning tool that allows agencies to assess likely impacts of pricing and other policy measures, (3) the capacity to improve integration of capital investment planning and demand growth and to assess benefits of demand-side management (DSM) measures in terms of deferral of investments, (4) the capacity for water agencies to engage their stakeholders (city councils, members of the public) in discussions regarding the merits of alternative conservation measures.
METHOD
The project follows a three-stage methodology. The first step involved working with a number of partner Canadian municipal water agencies to collect the data needed to estimate price and income elasticities for residential water demands in their cities. These data included residential consumption (annual or monthly), prices, incomes, climate and household characteristic data and were derived from water agency user records, Statistics Canada census data and other sources. This data gathering process provided valuable lessons regarding data storage and extraction needs for demand relationship estimation. Residential water demand equations were then estimated separately for each water agency using several functional forms including linear, double-log and Stone Geary. Estimation methods included instrumental variables estimation to account for potential price endogeneity and error correction methods to allow for price responses over different time scales.
The second step involved the construction of a spreadsheet-based water demand forecasting program. The program simulates future residential demand growth based on, first, user-inputted rates of growth of important water demand drivers (such as prices, income, population and conservation policies) and, second, user-chosen elasticities. In both cases, the program user may select from pre-set values for demand driver rates of growth and demand elasticities or may input values reflecting the circumstances of her water agency's operating environment. Currently a Monte Carlo capability is being constructed for the model to allow users to carry out risk analysis regarding future scenarios.
The third stage of the project involves testing of the demand program and is currently underway. Data from municipal water agencies are being inputted and water demands projected under a variety of assumptions regarding rates of growth of demand drivers as well as assumed elasticity values. Alternative output formats (tabular, graphical, etc.) are being considered relative to different stakeholder needs.
RESULTS & DISCUSSION
The first stage of the project yielded estimated price and income elasticities. The specific values vary across municipalities and according to estimation method. For example, in the case of the Capital Region District in British Columbia, price elasticities estimated with a panel of household water consumption data range in value from -0.18 to -0.65 (at the sample mean) depending on whether a double log or Stone Geary specification is used.
The second stage of the project yielded a beta version of the demand forecasting model which was tested using the results of the first stage as well as user-inputted data from several municipalities. In one case of a rapidly growing municipality in southern Ontario, the model demonstrates the potential errors that can be introduced into forecasts by not accounting for price-related behavioural responses by households. The base-case scenario assumes water demands are driven only by income and population increases and projects a 35% increase in aggregate water demand by 2050. Projected inflation-adjusted annual water price increases of 1.8% are modeled to take place but not to impact demand growth in this scenario. Conversely, water demand growth is projected to be only 19% under the alternative assumption that households do respond to price changes with a price elasticity of demand of -0.22. Thus, in this example, failing to account for potential behavioural responses to price changes leads to water demands being significantly overestimated (with attendant possible implications for perceptions for needed infrastructure spending).
CONCLUSION
This project demonstrates the benefits of academic researchers partnering with municipal water agencies in order to transfer research findings and increase partners' operational capabilities in response to partner-identified needs. The specific goal was to improve water agencies ability to project water demand scenarios that not only incorporate commonly used drivers such as population and income but also account for behavioural changes likely to arise from price changes. The resulting spreadsheet-based model provides a user-friendly means to enhance agencies' capabilities for conducting demand projections, anticipating impacts of price changes and communicating these results to stakeholders.
1. Billings, R.B. and C.V. Jones. 2008. Forecasting Urban Water Demand. Denver, CO: American Water Works Association.
2. Renzetti, S., O.M. Brandes, D.P. Dupont, T. MacIntyre-Morris and K. Stinchcombe (2014) "Using demand elasticity as an alternative approach to modelling future community water demand under a conservation-oriented pricing system: An exploratory investigation" accepted for publication in Canadian Water Resources Journal.