Experimental Design
We visit 3000 poor households in Malawi across 150 villages, offering different preventive goods at randomly assigned prices. The study will be conducted in two rounds.
In round 1, the survey team visits 150 villages that have been previously selected and, in all villages, randomly identifies 20 respondents meeting the sample criteria described in the previous section. The respondent is then assigned to one of 8 treatment groups determined by the cross-randomization of two conditions:
1. Counter price: total price (in Malawian Kwacha) for which the household can buy the product that is offered, inclusive of the delivery fee;
2. Tag price (in Malawian Kwacha): actual price of the good, that is, the difference between the counter price and the delivery fee.
The table presented in the attached Pre-Analysis Plan summarizes the treatment conditions and the sample composition.
By randomizing what the delivery fee was, as part of that full price, households effectively pay different prices for the same good -- although with the same cash on hands by the end of that round. Analogously to Chetty (2009), this variation in how the price of the good is framed, affects reference prices for those products, without influencing other important parameters determining future demand.
Moreover, within each cell, the type of good offered to participants is randomly selected. Half of the respondents will be offered an experience good, i.e. 3 bars of soap, while the other half will receive a credence good, that is a 100ml bottle of Multivitamin Supplement.
Once the individual has been selected, she is asked to complete a short survey (see next session) and is offered the possibility of buying soap or vitamin supplements - depending on which good is randomly selected - delivered to their doors at either full or discounted price, according to the table shown above. At the end of the visit, independent of whether the participant had decided to purchase the product or not, the enumerator reveals more information about the composition of the price, i.e. how much of the counter price that was paid covered the price of the product (tag price), and how much was the delivery fee.
In the second round, which takes place approximately 2-3 weeks after the first visit, the survey team revisits the same households and, before administrating a short survey, it elicits willingness to pay for a larger set of preventive health goods, which we categorize into "experience" and "credence" goods.
1. Experience goods: soap, mosquito bednets, chiponde (supplementary feeding).
2. Credence goods: multivitamin supplement, chlorine, waterguard, and deworming tablets.
To make our measure of willingness to pay as reliable as possible, we elicit it in an incentive-compatible way use the Becker-DeGroot-Marschak (BDM) mechanism. Moreover, the following measures are taken to mitigate all possible measurement concerns:
1. After being explained how the game works, households are asked to state their maximum willingness to pay to receive a single dose of each product, taking into account that delivery is offered for free during this second visit.
2. Only one predetermined good, namely soap, is actually sold in the end, but households learn which one only at the end of the experiment.
3. At the beginning of the procedure, each respondent is shown a unit of each good to make sure that they don't perceive some goods as more likely to be sold than others.
4. Given that respondent are likely to be unfamiliar with this type of game, we run a trial round with peanuts, to make sure that the rules of the game are clear before proceeding, and that participants understand that it's in their best interest to answer truthfully.
5. We randomize the order in which the goods are presented.
6. Finally, a price is randomly drawn and, if it is lower than the household's stated willingness to pay, the good is sold at that price. Otherwise, the household is not allowed to buy the product.