Consumer Preferences for Antibiotic-Treated Citrus: An Experiment Using GSO and BDM Mechanisms.

This study will address two interconnected challenges in consumer research: the methodological question of eliciting consumer valuations that are in line with market and the practical issue of consumer acceptance for antibiotic-treated citrus. As the U.S. citrus industry battles Huanglongbing (HLB), understanding consumer reactions to antibiotic treatments is crucial. We propose to test a novel experimental design comparing Game Structure Obvious (GSO) as compared to the Becker-DeGroot-Marschak (BDM) mechanisms and explore which one lead to preference elicitation that is more in line with the market and preference revelation. Additionally, we will examine how information about disease severity affects consumer support for antibiotic treatment in citrus production. This study extends recent theoretical work on preference elicitation mechanisms to field settings while providing actionable insights for agricultural policy and public health regarding antibiotic use in crop production.

2024-12-17

2024-12-23

The primary outcome is willingness to pay (WTP) for oranges from trees treated with antibiotics as compared to conventional oranges. We will additionally explore the impact of information provision on consumers WTP for

Our primary outcome variable will be willingness to pay measures that will be obtained directly from the BDM treatment. For the GSO treatment, WTP is the interval implied by the switch point [“Try to Buy”, “Not Buy”).

As secondary outcome, we will measure game form recognition using questions measuring understanding of each of the two mechanisms (BDM vs GSO).

The experiment employs a 2×2 between-subjects full factorial design with three factors: elicitation method (BDM vs GSO), reward type (hypothetical vs real). In addition, we will also present within-subject information provision (no information vs information about citrus greening) before the third round.

Not available

The randomization will be done by a computer and the software program (Qualtrics).

The randomization unit is at the individual level.

No

We will not use cluster.

295 subjects/treatment*4 treatment x 3 rounds x 2 products = 7080 observations

Our study will include a sample size of 295 participants per treatment group, resulting in a minimum total sample of 1180 participants across all four groups (295 participants * 4 groups).

Power analysis In this study, WTP for AB-treated citrus is our primary outcome of interest, a continuous variable measured in dollars amount. Given our 2*2 design, our power analysis is conceived to detect differences between two groups GSO vs BDM as well as information versus no information Hypothetical versus real. Therefore, we will need a t-test for independent samples. To determine the effect size needed, rather than relying on an isolated effect size estimate from a single study, we conducted a review of seven studies (Balogh et al., 2016; Brown et al., 2023; Kajale & Becker, 2014; Kendall & Chakraborty, 2022; Lombardi et al., 2019; Wongprawmas & Canavari, 2017) that are similar to our studies, which focused on consumer’s valuation. We identified six relevant studies, and assign weights based on the following characteristics: type of valuation (1 if WTP is used, 0 if not), elicitation method (1 if BDM and multiple price list is used in the study, 0 if not), and incentives (1 if incentivized, 0 if not). We then develop a composite weight based on the three latter and the sample size of the study. The effect size is estimated using the formula: ∆=((μ_0- μ_1 ))⁄σ represents the effect size. μ_0 and μ_1 represent the mean of two different treatments. We calculated a weighted average effect size based on the sample sizes and the degree of similarity to our study. This resulted in a weighted average Cohen's effect size of 0.25. To determine the optimal sample size needed, we followed different approaches divided in 2 steps for each preference elicitation mechanism. In step 1, we determined the sample size needed for the BDM mechanism. Since we have a between-subjects design, and we need means comparison, we followed Canavari et al. (2019) by using the following formula: n=(2(z_(1-α⁄2)+z_β )^2)/∆^2 Where z represents the statistics and using the conventional values for α (type I error) of 0.05 and β (type II error) of 0.20 since we consider a power of 80%. That led to 251 subjects/ group. Additionally, to account for potential attention check failures, we will increase the sample size by an average of 15% following Gupta et al. (2021) and Grodeck and Grossman (2024) who found inattentiveness to be 16%, and 14% respectively. Based on that, in stage 1, we will need 295 subjects/ group.