Preferences for Recycled Water

One control group and three information-treatment groups.

2026-08-28

2026-11-30

Policy preferences:
- Support for different household, municipal, and commercial applications
of recycled water.
- Support for introducing recycled water in the respondent's municipality
- Support for municipal investment in recycled-water infrastructure.
- Support for a petition promoting recycled water in Sweden.

Behavioral preferences:
- Intended use of recycled water for different household purposes.
- Willingness to pay for recycled water for different household and commercial purposes.

For household uses, respondents enter a willingness-to-pay amount between
0 and 170 SEK per cubic meter. A reference price of 85 SEK per cubic meter is
shown for ordinary tap water. For commercial uses, respondents are shown a
reference price for the same service using drinking water.

For outcomes measured on a Likert scale, we additionally construct binary
support indicators where applicable.

We also construct standardized indices for policy preferences and behavioral
preferences, as well as an overall index. Components are standardized to the Control group, averaged within each index, and the resulting index is restandardized to the Control group.

We examine policy support, willingness to use recycled water, and willingness to pay as the main outcomes.

To test H1--H3, we estimate the effects of Quality information and Goal information, as well as the interaction between the two information components.

We also test whether the effect of Quality information differs from the effect of Goal information, without specifying a directional hypothesis for this comparison. To test H3 specifically, we compare the Combined Quality and Goal condition with the Control, Quality-information, and Goal-information conditions.

The main analysis uses covariate-adjusted regressions controlling for the demographic characteristics collected in the survey. We also report unadjusted regressions. We use heteroskedasticity-robust standard errors and two-sided tests, and additionally report nonparametric tests.

In the follow-up survey, conducted 5--7 days after the main survey, we test the persistence of treatment effects using one measure of support for municipal use of recycled water and one measure of willingness to use recycled water for private outdoor purposes. We compare the estimated follow-up treatment effects with the corresponding effects in the initial survey. As a descriptive benchmark, we assess whether the follow-up coefficients are of a magnitude broadly consistent with previous experimental studies, where follow-up effects are often around 40--70\% of the initial treatment effects.

For the within-subject scenarios in the main survey, we compare willingness to
use and willingness to pay for recycled water across the following scenarios,
using data from the Control group:
- a water kiosk located 1 kilometer versus 10 kilometers away;
- collection from a water kiosk versus delivery by a water taxi; and
- normal conditions versus a warm and dry summer.

These comparisons are based on paired t-tests and corresponding standardized
within-subject effect sizes.

Secondary outcomes include:
- perceived quality and safety of recycled water;
- perceived environmental and societal benefits of recycled-water reuse;
- trust in recycled-water treatment;
- perceptions of water scarcity;
- normative views about whether water should generally be conserved;
- respondents' preferred prices for drinking water and recycled water.
The normative-view questions were added during the pilot.

As secondary analyses, we estimate heterogeneous treatment effects by age,
gender, education, household income, employment, municipality or region,
household composition, housing characteristics, home ownership, garden
ownership, number of cars, environmental attitudes, and previous experience
with water scarcity, where available.

The perception measures are used to examine potential channels through which
the information treatments affect preferences. Respondents' preferred prices
for drinking water and recycled water are analyzed as additional secondary
outcomes.

Adult respondents in Sweden participate in a randomized online survey experiment on preferences for recycled water. Participants are assigned to a Control group or to information treatments concerning the quality and safety of recycled water, the societal goals of recycled-water reuse, or both.

We examine how these information treatments affect perceptions of recycled water, policy support, willingness to use recycled water, and willingness to pay for it. The survey also includes within-subject scenarios that vary access conditions, service provision, and environmental conditions in order to study how these factors influence preferences.

Not available

Randomization will be conducted using the survey software. We target approximately 30% of
the final sample in Control and approximately 23.3% in each of the three treatment conditions.
Pilot observations count toward these final target shares.

respondent

No

Not applicable. Treatment is assigned at the individual level

We aim for a final analysis sample of approximately 3,000 completed respondents, including about 420 pilot observations. Depending on respondent availability through the survey provider in Sweden, the final sample may range from approximately 2,600 to 3,300 respondents. A sample of around 3,000 is close to the maximum feasible sample indicated by the survey provider for Sweden.

For a final sample of 3,000 respondents, the target allocation is approximately:
- 900 in the Control group;
- 700 in the Quality-information group;
- 700 in the Goal-information group; and
- 700 in the Combined Quality and Goal group.

A key consideration in choosing this allocation was to provide balanced
precision for the 2 x 2 treatment design, with a somewhat larger Control
group as the common comparison group, rather than to maximize power for the
follow-up survey.

A pilot survey with about 420 respondents has already been conducted. It included about 130 respondents each in the Control and Quality conditions and about 80 respondents each in the Goal and Combined Quality and Goal conditions. The pilot observations will be included in the final analysis. During the pilot, additional channel questions were added.

At a final sample of 3,000, with approximately 900 respondents in the Control group and 700 respondents in each treatment arm, comparisons of an individual treatment arm with Control provide approximately 80\% power at the 5\% significance level to detect a standardized effect of about 0.14 standard deviations. Comparisons between two treatment arms, with approximately 700 respondents in each arm, provide approximately 80\% power to detect a standardized difference of about 0.15 standard deviations. Assuming an 80\% follow-up response rate, as indicated by the pilot, the expected follow-up sample is about 2,400 respondents. In this follow-up sample, comparisons of an individual treatment arm with Control provide approximately 80\% power to detect a standardized effect of about 0.16 standard deviations.

Preferences For Recycled Water