Water conservation model

The team used data from a water conservation program, WaterFix, combined with survey data and integrated into a model which described the householders' choice to opt into the program.

What does the model describe?

This model is used to examine residential WaterFix adoption rates and the factors that have an impact on adoption. The model also shows how much water could be saved from WaterFix adoption, and how many tonnes of emissions-reduction could be achieved, over a period of 12 years. These savings are then translated into dollar benefits to households and the water company. We believe that it is a good tool for water planners to try and understand how different types of ways that planners could boost adoption rates for the WaterFix program.
Link to model

What is the WaterFix program?

The water company offers customers the WaterFix® program, where qualified plumbers check a home for leaks and other opportunities to save water. A licensed plumber may repair toilet, tap and pipe leaks, replace toilets, replace indoor and outdoor taps, replace shower heads, and install flow control device. There is a cost associated with upgrades of fixtures or repairs.
There were seven versions of the WaterFix program that were explored. All of these options are not necessarily cost efficient for the homeowner and the choice to opt-in therefore goes beyond cost optimisation. In the model we describe here, we are presenting four options for payments and they are:
  1. WaterFix at a price of $430 to be paid immediately.
  2. WaterFix paid over five years via the water bill which after accounting for water savings which would amount to $36 per year over five years.
  3. WaterFix with a 30% discount so this would amount to $300 paid immediately.
  4. WaterFix paid over five years with 30% discount: Paid via the water bill this would amount to $10 per year over five years.

Outputs

Adoption rates: This is the main output of interest, i.e. the proportion of the population that are choosing to participate in the program in a given year.
Population level adoption rate: Proportion of the entire population who have chosen to participate in the program.
Cumulative water savings: Total volume in MLs of water savings if these results were scaled up to the entirety of the Sydney metropolitan area.
% of dam levels: Conversion of cumulative water savings to give an indication of how much of a different this program would do to the overall dam levels over the time period.
Cost effectiveness - hhs: An estimate of $s spent by households divided by the volume of water saved. As a reference point, the current price of water in Sydney is $2.3 per kilolitre.
Cost effectiveness - SW: An estimate of $s spent by the water company divided by the volume of water saved. As a reference point, the current price of water in Sydney is $2.3 per kilolitre.
Cost effectiveness - ttl: An estimate of $s spent by the water company and households altogether divided by the volume of water saved. As a reference point, the current price of water in Sydney is $2.3 per kilolitre.
Ttl emissions reductions (tonnes): the total amount of reduction in emissions associated with a reduction in potable water supply, with an assumed energy intensity per kL and an associated tonnes of CO2-e emissions from 1kWh of energy.
Emissions reduction value ($Ms): conversion of the total emissions to a monetary value, based on $15 per tonne of emissions.
Average motivation: The average likelihood amongst the population of simulated agents of being motivated to participate in the program.
Promotional activity: The spending on promotional activity per year normalised to a value between 0 and 100.
Level of engagement: Modelled proportion of the population being actively engaged with water conservation as a priority issue.

Inputs

Drought scenarios: These describe how the drought occurs and impacts on dam levels across the scenarios, which is a key driver for community concern and behaviour. The following figure describes the changes over time for each of the scenarios.
How the program is paid for by the householder:
  1. WaterFix at a price of $430 to be paid immediately.
  2. WaterFix paid over five years via the water bill which after accounting for water savings which would amount to $36 per year over five years.
  3. WaterFix with a 30% discount so this would amount to $300 paid immediately.
  4. WaterFix paid over five years with 30% discount: Paid via the water bill this would amount to $10 per year over five years.
Making participation in the program easier: The users to switch on or off framing of the offering in terms of:
  • Making it easier, in terms of effort, to participate in the program.
  • Making it easier, in terms of time requirements, to participate in the program.
How participation in the program is sold to householders : The users to switch on or off framing of the offering in terms of:
  • Promotion of how participation in the program enables environmental benefits.
  • Promotion of how participation in the program helps address supply concerns.
  • Promotion of how participation in the program helps the householder receive better quality fittings.
Whether the water company should work with influencers to shape decisions : The users to switch on or off:
  • Influencing retailers to recommend participation in the program.
  • Influence online forums to recommend participation in the program.
  • Influence media to recommend participation in the program.
Increasing the level of "Investment": The user can choose which level of investment into promotional activities based on the input data which specifies Investment is a variable. There are three choices, Low level, Ramping up, and High level that describes for each year the level of investment as a number between 0 and 1, with 0 lowest level of historical investment and 1 being the highest level of annual investment during the millennium drought. There are described in the figure here.
Choice of marketing style: The user can choose which manner in which householders are approached, i.e. Whether via website (passive), via direct communication such as emails, phone calls or door knocking (targeted), or via media campaigns; or a combination of media and targeted.
Plumber sales: Whether or not to enlist plumbers to sell participation in the program whenever they are already onsite at someone’s home.
Summary of how the user can try to boost adoption rates? At an abstract level, there are four ways in which the adoption rates could be boosted:
  1. Increasing the level of community priority for water conservation, i.e. when community members prioritise water conservation as a key issue and therefore becomes knowledgeable about programs and behaviours. This happens partially due to things outside of the water company’s control, i.e. drought etc. It also happens partially due to investment by the water company. Investment levels are a key input here.
  2. Activate people into making a decision; people don’t always make decisions about these matters unless being prompted such as by media or promotional activity, or by being approached by sales agents. To activate people into making a choice, media interest or educational programs play an important role. Targeted communication, such as via letters, emails or a note in the water bill, have been found to be effective means of activating people to make a decision. To explore levers at this level, check what happens when you increase the marketing investment from low to high or the marketing approach from passive to targeted. Another interesting approach that can be explored is to enlist plumbers to sell participation in the program
  3. Making choices; once people are activated into making a decision those people choose to adopt or not adopt a behaviour with a certain probability. Here, behavioural drivers such as intrinsic motivation (such as perceived importance of doing something for the environment or for community benefit), as well as perceived importance of expenses in terms of time, effort and money are key factors. Important methods for making people more likely to choose to participate in WaterFix relate to changing the way that WaterFix is paid for, framing participation in terms of environmental benefits or other issues, or to reduce the amount of effort or time that is required for participating.
  4. Being subject to influence; when people are uncertain about what to do, they may seek further advice from friends or recommendations from what we call ‘influencing agents’. The influencing agents here are the likes of media, online forums and retailers. There is an option for working with influencing agents to promote participation in the WaterFix program.
What data is being used in the background? We used a combination of empirical survey data, expert judgment, calibration and sensitivity analysis to parameterise the model. Importantly we have data for:
  • Describing householders – based on survey data.
  • Describing how behavioural drivers influence motivations – based on multinomial regression models of stated intentions.
  • Individual level of data on how householders’ approach and trust different information sources when they seek recommendations – based on survey data.
  • The effectiveness of media campaigns on boosting the activation of decision making within the community.
  • To validate the data and define a couple of the model parameters, we used historical adoption data.

Describing the level of engagement with water conservation

Based on historical data provided by the water company, we have calibrated a model that describes the level of “engagement with water conservation” in the community, based on whatever data we have. Currently, the model is based on taking into account the 1) level of investment into promotion and media around water conservation by the water company, 2) the presence or not of drought, 3) the dam levels, and 4) an autoregressive component to represent the inertia in the dynamics.

Describing intention to participate

As an input into the Consumat model, we apply a multinomial regression approach to describe how motivation is influenced by behavioural drivers. This is described by the equations which can be found in the project report (Moglia et al., 2019). The variables that are statistically significant in explaining the data are the following:
  • Tenure category
  • Number of children
  • Time poverty
  • Employment
  • Prioritising being in nature
  • Prioritising having quality things
  • Prioritising protecting environmental values
  • Prioritising reducing greenhouse gas emissions
  • Prioritising the amount of effort involved
  • Being concerned about water supply security

Model validation

The ABM model was validated in terms of its capacity to replicate historical data levels of adoption of the particular water conservation program being modelled. The figure here shows the alignment between historical adoption rates and modelled adoption rates. We do note however that the modelled adoption rates are based on some level of calibration of a small number of parameters in the model.
Note: Adoption rates are the proportion of all the households in the population which chooses to opt into the program in a given year.