This market experiment is part of a larger study entitled "Food Quality and Child Diets" at our study site. That study includes a marketplace survey, during which the interviewer will conduct a discrete choice experiment to elicit respondents' marginal rate of substitution (MRS) between a hypothetical fortified cereal, a plain maize porridge, and complementary foods made of greens, beans and other nutrient-dense family foods. Before the choice experiment, the interviewer will read one of two randomly assigned descriptions of the fortified cereal, to test for significant differences in MRS caused by how the fortified food is described in its marketing messages or label claims. Following the choice experiment, the interviewer will also conduct a random-price auction to elicit respondents' willingness-to-pay (WTP) for actual infant foods available in a variety of locations around the study sites.
The discrete choice experiment elicits respondents' preference between combinations of foods that would be cooked and fed to the child at home during or between family meals. Options are represented by photos of local dishes. The choice is hypothetical but does provide a realistic description of food options likely to be available in the respondent's home. In this experiment, respondents state their choice between two options, where each is a combination of either: (1) a double or a single portion of plain maize porridge, plus (2) a double or a single portion of a nutrient-dense family food, with or without (3) a single portion of a fortified infant cereal which is darker in color than the maize porridge. We restrict choices to options that have either 3 or 4 portions in total, so that only four choices are needed to make a full factorial design. We then replicate those four choices with three different nutrient-dense family foods, for a total of 12 choices that appear in random sequence. Prior to this choice experiment, our intervention concerns how the photograph of fortified infant cereal is described. Treatments are randomly assigned and not known to either the interviewer or the respondent until they reach that stage in the interview script. At that point, in the control arm, fortified cereals are described as "made especially for infants, with more nutrients than ordinary maize." In the marketing treatment, fortified cereals are described as "made especially for infants, with many of the nutrients in greens, beans, fruits & vegetables". These statements correspond to the two main types of marketing and product packages currently observed for a variety of fortified infant cereals sold around the world.
The random-price auction follows Becker, Degroot, and Marschak (1964), eliciting WTP for a real product that could actually be sold to the respondent during the interview. The interviewer invites the respondent to state the maximum price at which she would purchase the product if it were for sale, and later draws a random price at which the sale occurs if that randomly-drawn price is equal to or below the respondent's stated WTP. This mechanism is known to elicit bids that are unbiased estimates of the respondent's actual market intentions. To elicit bids for multiple products without satiation, we solicit bids for plain cereal grain and all of the different fortified infant cereals available in local markets, and then randomly select one for which a price is randomly drawn to determine the sale. Such an auction serves essentially as a randomized double-blind experiment in which the randomized component is the price at which the respondent is able to purchase the product in question. In the setting of our trial, respondents will typically have come to the market without any cash beyond what is needed for that day's planned purchases. To ensure that respondents are able to make bids, they will be given an unconditional cash grant during the informed consent phase of the interview. This ensures that all respondents are able to participate in the auction, if they so choose, and decide whether or not they wish to make an offer to purchase each type of food. Any difference in behavior between WTP from a grant versus earnings will be absorbed by also soliciting WTP for plain maize, so that the endpoint of interest is the difference rather than the level of WTP.