Biased Beliefs, Motivated Reasoning, and Consumer Valuation of Vegetable Labels in Vietnam

The study will be conducted over two waves spanning a total of two months. The first wave will include the experiment and survey questions. The choice experiment will endow participants with money and ask participants to: play a dictator game between themselves and a charity; to rank, rate, and state their willingness-to-pay (WTP) for vegetables prior to and after introduction of a food safety label (and accompanying definitions if appropriate; and to share their beliefs about the relative food safety of the vegetable options). Survey questions will assess: purchasing habits, food safety and food safety label knowledge, prior illnesses from vegetables, demographics, and exposure to chemical defoliants. The follow-up survey, conducted approximately one month later, will ask participants to recall what labels they saw during the experiment, the ranking they assigned to the vegetables after seeing the labels, and again, their willingness to pay. Participants will be properly incentivized to accurately recall their rankings but will also be given the option to state that they do not recall. This allows for individuals who engage in motivated reasoning to suppress their post-signal ranking at a personal financial cost.

2022-12-12

2023-02-28

This study seeks to answer the following questions:
1. To what extent do individuals who hold biased beliefs engage in Bayesian updating or motivated reasoning in the short- and long-term when provided with varying food safety labels as signals for product quality?
2. Is there a trade-off between the comprehensiveness and accessibility of these labels in markets where consumers hold biased prior beliefs?
3. If so, can the simultaneous provision of an exact label definition increase the effectiveness of the VietGAP label by reducing biases and increasing purchases of vegetables in retail markets?

We will use a difference-in-differences regression model to estimate treatment effects to assess the extent to which individuals update their beliefs in the short- and long-term given information that may conflict with their prior beliefs. The dependent variables for the regression specification are as defined in subsection \ref{Variables}. We may also explore a triple difference-in-differences framework and propensity score matching given the plausibly exogenous treatment of exposure to chemical defoliants.
The main dependent variables of interest are 1) ranking, rating, WTP, and perception of food safety for vegetables pre- and post-signal, 2) recall of initial ranking, rating, WTP, and perception of food safety for vegetables at follow-up, and 3) final ranking, rating, WTP and perception of food safety for vegetables at follow-up. The main right-hand side variables include: 1) assignment to treatment or control groups, 2) prior illness or disability stemming from consumption of contaminated vegetables 3) prior exposure or to chemical defoliants or having a friend/relative who was exposed to chemical defoliants 4) vegetable purchasing habits, food safety knowledge and beliefs, and 5) basic demographic controls.

This study will be conducted in Hue City, Vietnam. This city lies in Thua Thien Hue province, one of the most heavily sprayed areas during the Vietnam War, just south of the 17th parallel. We have partnered with the Center for Knowledge Co-creation and Development Research (CKC), a non-profit that is established in Hue which has experience administering and collecting survey data in the province. The head-of-household or primary shopper for the household will be asked to participate in the experiment.

The study targets both households directly affected and unaffected by Agent Orange. Households affected by Agent Orange will be included in the pool of households that will be randomly determined to participate in the study. The total sample size is expected to be 600 households split into two treatment and control groups. The treatment and control groups are not expected to differ from one another across any observable characteristics.

Households are assigned to two control groups and two treatment groups in this experiment. A vegetable label will be displayed to the control groups while the vegetable label and definition will be displayed to the treatment groups. The definition reduces uncertainty of the quality of the labeled good. The first control group is presented with one conventional vegetable and one pesticide-free vegetable (a 'pesticide-free' label that only addresses one dimension of food safety) without a definition of the label. While the first treatment group is presented with one conventional vegetable and one pesticide-free vegetable with the label and exact definition. Similarly, the second control group is presented with one conventional vegetable and one VietGAP vegetable without a definition and the second treatment group is provided with a conventional vegetable and a VietGAP vegetable with the label and exact definition. These labels and definitions are signals that provide the individual with additional information on the food safety quality of the good. The use of the two different labels and their definitions creates variation in the comprehensiveness and accessibility of the information presented.

randomization using random sampling in Stata

wards, households

Yes

23 wards

600 households across 23 wards

600 households