Experimental Design Details
In both treatments, each subject has to choose between two different contracts. A contract consists of a wage w which the agent obtains when completing a specified number of real-effort tasks, N. To be precise, in our experiment, we choose an "encryption task'' as our real-effort task. In this task, agents are given a combination of letters which they have to convert into numbers by selecting the correct corresponding number from a table given to them (a single task takes about 40 seconds and previous work finds little hereogeneity across agents in ability and little learning over time).
In treatment A of our experiment, each subject chooses between contracts (10 tasks, 6 Euro) and (50 tasks, 9 Euro). Importantly, a subject herself does not have to solve all of the tasks specified in the contract because 5 tasks have already been pre-solved.
In treatment B of our experiment, each subject chooses between contracts (110 tasks, 6 Euro) and (150 tasks, 9 Euro). Importantly, a subject herself does not have to solve all of the tasks specified in the contract because 105 tasks have already been pre-solved.
Hence, in both treatments subjects have the choice between the very same two working contracts as the number of tasks to be solved is the same in the two treatments. Thus, rational choice, but also models of salience which assume that utilities attract attention would not predict a change in behavior. Crucially, however, the presentation of choices is different in the two treatments, due to the different framing of options. Hence, applying the concept of salience to numbers as seen on the decision screen (instead of to the corresponding utility values), implies that the difference in workload should be perceived as less salient in the decision situation in treatment B than in treatment A.
Thus, we predict that in treatment B more subjects should opt for the high-wage contract than in treatment A.
In addition, we make use of a novel memory task. Directly after choosing one of the two contracts, participants are required to remember and reproduce the content of the previous decision screen. That is, we pay them 0.1 EUR for each number they remember from the previous screen (the two payments and the two task requirements from the two contracts, and the number of pre-solved tasks).
This allows us to check whether participants remembered all aspects of their choice environment properly, or whether they paid more attention to some parts of the decision environment. In particular, we are able to link these choices to the contract choices of the agents. We will thus be in a position to examine whether participants were more likely to remember those aspects of the choice environment which theories of salience and limited attention consider to be more salient and attention-grabbing.