Experimental Design
We adapt the real-donation dictator game introduced by Eckel and Grossman (1996) and subsequently applied to compare subsidy types (Eckel and Grossman, 2003; Davis and Millner, 2005; Davis, 2006; Eckel and Grossman, 2006a,b). In the standard version of the game, subjects decide how much of their endowment to hold and how much to pass to a charity. This choice corresponds to an expenditure donation. In our variant of the game, subjects decide how many units of the charitable good to fund at a given price, using their endowment. This choice corresponds to a quantity donation.
Our variant of the game requires a charitable good or service that is easily quantifiable. We approached a relief organization which makes extensive use of quantity donation calls in fundraising campaigns. Among their activities, we chose the treatment of malnourished children in a certain area of South Sudan as this service offered practical units and prices for our experiment. The children are treated in two ``bush clinics'' operated by the relief organization. Treating one child for one month using a special nutritional paste and high energy cookies requires a donation of $15. We divided this number into practical units of nutritional packages per child and day which implies a price (required donation) of $0.50 per package. In the donation task of the online experiment, subjects were introduced to the charity, the charitable good, and its marginal provision cost to the charity.
There are one baseline condition and six interventions. In the control condition, no subsidy was applied. Subjects were endowed with $2 and chose how many units of the charitable good to fund at a price of $0.50. The six treatment conditions were derived by applying the three subsidy types at two different levels each. We framed the rebate conditions as a 50\% (33\%) rebate, so that $0.25 ($0.17) per unit provided would be added to the final reward. Match conditions were framed as a match of each (every two) unit(s) the subject provides at no additional costs. Discount conditions were framed as a possibility of providing units for $0.25 ($0.33) apiece. For all subsidy types, subjects were informed that the subsidy is provided by ``a third party", subjects facing the discount subsidy learned that the reduced price results from a third party funding the remaining cost of $0.25 ($0.17) per package. The two subsidy levels imply effective prices of $0.33 and $0.25. Note however that for the 1:2 match, the effective price per package is not constant, since only every second unit provided by the subject induces an additional subsidy payment.
Treatment conditions were administered in both a between-subjects (BS) and a within-subjects (WS) design to two different subject samples. In the BS design, subjects were introduced to a specific subsidy condition or the control and had to choose the desired number of units from a drop-down menu. In the WS design, all seven conditions were listed in random order and subjects entered, for each condition, an integer number indicating their desired number of units. Subjects were informed that one of the conditions would be randomly selected through a lottery and implemented.
We recruited subjects from an online labor market, Amazon's Mechanical Turk (AMT), restricted to US residents. In the posted task, interested workers were informed that they would earn $2 for answering a 20-minutes academic survey on several topics. Donations were mentioned as one of the topics, but the real-donation dictator game was not particularly salient compared to other survey elements, so it is unlikely that subjects considered the donation task as the main subject of investigation. Interested workers followed a link which directed them to the survey containing the experiment.