Experimental Design Details
Sampling procedure: The study will be conducted in two states and the Federal Capital Territory: Enugu, Lagos and FCT Abuja, representing the Eastern, Southern and Northern parts of Nigeria. To ensure a representative sample, we will follow the sampling procedure of proportionate size to randomly sample 20 enumeration areas (EAs) (14 urban EAs and 6 rural EAs) from each of the three States. Enumeration Areas are the primary sampling units designated by the National Population Commission (NPC) and the National Bureau of Statistics (NBS). For each State, 6 rural EAs will be divided into two clusters of 3EAs each to form 1 treated group (TG) and 1 control groups (CG). The same principle will be applied to the urban EAs to form 1 TG and 1 CG. Twelve (12) households will be selected from each of the EAs to give 240 respondents for each State comprising 120 treated households and 120 control households. For the study, we will have a total of 720 respondents comprising two groups, treatment (N = 360) and control (N = 360). A well-structured interviewer-administered questionnaire will be developed and tested for data collection through a pilot survey. Data will be collected using computer-assisted personal interviewing (CAPI) coded into survey solutions.
Experimental procedure: The outcome variable for the study is the acceptability of carbon tax and a dichotomous response of 1 = Yes or 0 = otherwise will be used. The households’ carbon tax knowledge and acceptability at the baseline will be collected using a questionnaire. Participants will be asked to rank their knowledge about a carbon tax using a 4-point scale of 1 = Excellent, 2 = Good, 3 = Fair, and 4 = Poor. To determine how information provision on carbon taxation influences acceptability and to determine the preferred use of carbon tax revenues by households that will maximize their acceptability, the survey will follow a guide by PMR & CPLC (2018) on communicating carbon pricing, by providing a short text explaining how a carbon tax works and the revenue spending in comprehensible terms for different audiences; urban and rural. Only the treatment group will receive this information on carbon tax workings and revenue spending while the control group will not receive any information (treatment). They will be asked to allocate a percentage of the total carbon tax revenues (100%) to each of the four proposed revenue-use options (funding mitigation projects that reduce carbon emissions; funding community adaptation projects to extreme weather occurrences; revenue transfers to low-income households as compensation; and revenue transfers to all households) as if it was their decision. At the end-line survey, after 2 weeks of treatment, we will retest the acceptability of carbon tax by the treatment and control groups using the same questionnaire.
Analytical models: The effect will be estimated using a double-difference (DD) impact estimation procedure (Khandker et al., 2010). In DD, the treatment and comparison group (first difference) are compared before and after an intervention (second difference) rather than comparing at one point in time. That is, given a two-period setting where t = 0 before the intervention and t = 1 after intervention implementation, letting Y_t^A be the outcome (acceptability of carbon tax) for the treated and control units in time t, the average intervention impact will be estimated as follows: DD = E(Y_1^T-Y_0^T│T_1=1) - E(Y_1^C-Y_0^C│T_1=0). T1 = 1 denotes treatment (information provision and preferred revenue allocation) at t = 1, whereas T1 = 0 denotes untreated areas. The advantage of the double-difference estimator is that it nets out the effects of any additive factors (whether observable or unobservable) that have a fixed (time-invariant) impact on the outcome indicator (Ravallion, 2001). The steps for the DD include a baseline survey, information provision, and preferred revenue allocation intervention, an end-line survey, and then an estimation of the treatment effects (intent-to-treat (ITT) and average treatment effect on the treated (ATET)). The ATET is consistently estimated by differencing the mean outcome for the treatment and control groups over time to eliminate time-invariant unobserved characteristics and differencing the mean outcome of these groups to eliminate time-varying unobserved effects common to both groups. A probit model will be used to estimate the determinants of acceptability. Socio-demographic variables - trust in government, political orientation, climate concern, green membership, number of cars owned, monthly income, number of adults and children in the household, years of education, gender and age will be included as predictor variables in the model. We specify a probit model as follows: P(Y=1∣T)=Φ(β_0+β_1 T+β_2 X_2...+β_k X_k) where Y is the binary outcome variable, accept or do not accept, T is the treatment indicator variable (1 for treatment group, 0 for control group), X is the vector of predictor variables, Φ is the standard normal cumulative distribution function, and β are the coefficients to be estimated.