Assessing Testing Effects and Retention in a Cluster-Randomized Workshop Intervention on Pollution Taxes

The intervention consists of a structured educational workshop on the principles, design, and economic rationale of pollution taxes. The workshop is intended to improve participants’ understanding of environmental taxation, including its objectives, behavioral impacts, and implementation challenges. Workshops were delivered across seven university sites, each using the same curriculum, instructional materials, and facilitator training to ensure consistency in delivery.

Under the cluster-randomized design, the seven workshop sites were assigned to one of two conditions:

Group A – Test Before Workshop:
Participants completed the survey instrument once, immediately before participating in the workshop. The same instrument was not administered again afterward.

Group B – Test After Workshop:
Participants attended the same workshop and completed the same survey instrument once, immediately after the session.

In addition, a third comparison group was recruited at similar campuses where no workshop was delivered:

Group C – No Workshop + Test:
Participants did not attend the workshop but completed the same survey instrument during the same time period as Group B. This group serves as an untreated control for estimating workshop effects.

The intervention allows for estimation of differences in attitudes toward pollution taxation across groups that completed the same test under three conditions: before exposure, after exposure, and with no exposure at all.

Participants were not informed of the existence of different survey-timing conditions or of the inclusion of a separate non-workshop comparison group (Group C). The study was presented uniformly as an educational project on pollution taxes with an accompanying survey measuring student attitudes.

In Group A, the survey instrument was administered as a standard pre-session activity before the workshop began. No indication was given that other participants might complete the survey at a different time.

In Group B, the same survey instrument was administered only after the workshop. Participants in this group were not informed that some workshops had completed the survey beforehand.

In Group C, students were surveyed in comparable campus settings without being informed that they constituted a non-workshop comparison group. The survey was presented simply as part of a study on environmental taxation attitudes.

Randomization of workshop sites into Groups A and B was conducted in advance using a stratified and attendance-balanced allocation procedure. Facilitators were blind to the randomization process and to the purpose of differing survey timings.

2025-06-13

2026-02-10

Primary Outcome: Willingness to Pay (WTP) for Pollution Taxes

The primary outcome is Willingness to Pay (WTP) for pollution taxes, measured using a 4-item behavioral intention scale based on Theory of Planned Behavior constructs. Each item is rated on a 7-point Likert scale, and a composite WTP score is computed as the mean of all items (after reverse-coding where applicable).

Because each participant completes the survey once, WTP is observed under three conditions:

Before the workshop (Group A)

After the workshop (Group B)

Without any workshop exposure (Group C)

These comparisons allow estimation of:

Workshop Effect:
Difference in WTP between Group B (workshop + test after) and Group C (no workshop + test).

Baseline vs. Post-Workshop Differences (Descriptive):
Comparison of Group A (test before) and Group B (test after).

Baseline vs. Non-Workshop Posttest Differences:
Comparison of Group A and Group C to assess whether untreated posttest attitudes differ from baseline levels.

Moderation Analyses

Two classes of moderators will be explored:

Pro-Environmental Behavior Index
A 12-item baseline measure capturing prior environmental behavior will be tested as a moderator of WTP differences across groups.

Demographic Moderators
Gender, religion, income, academic discipline, and other background variables will be examined for heterogeneous effects on WTP across conditions.

Willingness to Pay (WTP) for pollution taxes is measured using a 4-item behavioral intention scale adapted from the Theory of Planned Behavior. Items include intention to comply with pollution taxes, likelihood of paying required charges, and related behavioral commitments. Each item is scored on a 7-point Likert scale; the composite score is the mean of all items (reverse-coding when necessary).

Because each participant completes the survey once, the WTP measure is observed under three conditions:

Before workshop exposure (Group A)

After workshop exposure (Group B)

Without workshop exposure (Group C)

This structure allows for:

Causal estimation of the workshop effect through comparison of Group B vs. Group C

Descriptive baseline vs. post-workshop differences by comparing Group A vs. Group B

Baseline vs. untreated posttest differences by comparing Group A vs. Group C

Moderation analyses using demographic variables and a pro-environmental behavior index will examine heterogeneity in WTP across groups

Secondary outcomes include:

Theory of Planned Behavior (TPB) Constructs

Attitudes toward pollution taxes

Subjective norms

Perceived behavioral control

Moral norms
Each construct is measured using four Likert-scale items (1–7), averaged into composite scores.

Workshop Evaluation (Groups A and B only)
Items measuring perceived clarity, usefulness, interest, and likelihood of recommending the workshop. These are descriptive outcomes.

Heterogeneous Effects
The effect of the workshop across demographic characteristics (gender, income, religion, year of study, academic discipline).

Moderation by Pro-Environmental Behavior
A 12-item scale capturing prior environmentally friendly behaviors.

TPB constructs provide insight into psychological mechanisms underlying willingness to pay. Each construct is measured with four Likert-scale items, reverse-coded when appropriate. These constructs will be analyzed as:

Additional outcomes of interest

Moderators of the workshop effect

Potential explanatory variables in multivariate models

Workshop evaluation items provide feedback on the clarity and perceived usefulness of the session but are not used for formal hypothesis testing.

Demographic and behavioral moderators (e.g., prior environmental behavior) allow exploration of how workshop effects vary across subpopulations.

This study adopts a three-arm between-subjects experimental design, implemented across seven university workshop sites and an additional non-workshop comparison group. Participants complete the same survey instrument once, but at different points relative to the workshop intervention.

The three groups are:

Group A – Test Before Workshop
Participants completed the survey immediately before the workshop session.

Group B – Test After Workshop
Participants attended the same workshop and completed the survey immediately after the session.

Group C – No Workshop + Test
Participants did not attend any workshop but completed the same survey during the same time window.

This structure enables estimation of workshop effects and comparisons of baseline, post-treatment, and untreated attitudes toward pollution taxation.

Workshops were conducted at seven university campuses. Assignment to Groups A and B occurred at the cluster level, with entire workshop sites allocated to a single condition. Group C participants were recruited independently at comparable institutions where no workshop was held.

The design allows for:

Clean workshop effect estimation: Group B vs. Group C

Baseline vs. post-workshop comparison: Group A vs. Group B

Baseline vs. untreated posttest: Group A vs. Group C

All workshop sites used identical instructional materials and facilitator training to standardize the intervention.

Randomization was implemented at the workshop (cluster) level using a stratified and attendance-balanced allocation procedure:

Workshops were stratified into North (3) and South (4) regions.

Within each stratum, workshops were randomly assigned to either:

Test-Before Workshop (Group A) or

Test-After Workshop (Group B)

To prevent imbalance in sample sizes, a greedy balancing algorithm was applied using expected attendance. Workshops were sorted by anticipated size and iteratively assigned to the condition with the lower cumulative attendance within each region.

Group C was not randomized; it serves as an external untreated comparison sample recruited to establish a counterfactual.

The randomization unit is the workshop site (cluster).
All participants at a given workshop location were assigned to the same condition (A or B). Group C participants were surveyed outside the workshop setting.

Yes

Randomized clusters: 7 workshop sites

Group A: 3 clusters

Group B: 4 clusters

Non-randomized comparison sites: Group C (independent sample)

Planned total sample: ~1,200 students (≈500 for group A & B, 200 for group C)

Exact totals may vary depending on final attendance and recruitment.

Approximately 1,000 university students, based on expected attendance figures across the 7 workshop clusters.

Group A (Before the Workshop): 3 clusters, approximately 500 students

Group B (After the Workshop): 4 clusters, approximately 500 students

Assumptions: 7 clusters randomized to Groups A and B Group C treated as an independently sampled comparison group Intra-cluster correlation (ICC) for attitudinal outcomes: 0.01–0.05 α = 0.05, power = 0.80 Standard deviation of WTP ≈ 1.0 (standardized Likert composite) Under these assumptions, the design can detect a standardized mean difference (Cohen’s d) of: d ≈ 0.30–0.35 for comparisons between Groups A and B d ≈ 0.20–0.30 for comparisons involving Group C (due to larger effective sample)