Experimental Design
This project analyzed a set of field experiments in an online labor market to answer two open questions about referrals: first do referrals contain information about worker productivity? Second, do referred workers work harder or more effectively because they are referred? Three experiments were conducted for this project: the peer influence experiment, the team experiment, and the selection experiment.
The peer influence experiment was designed primarily to determine whether referrals work harder as a result of being referred because they think their performance and persistence will affect either their referrer's position at the firm or their relationship with their referrer. In the peer influence experiment, referred and non-referred workers tested an airline flight website by answering questions about the site every other day for 12 days. Referrals in this experiment were randomized into two treatments. The non-monitoring treatment was designed to minimize peer influence. Referrals in this treatment were told their referrers would never know their performance and (after referring) referrers were told they would not be judged on the performance of their referrals. The monitoring treatment was designed to maximize peer influence. Each referrer in this treatment received an update on their referral's performance after each day of work. It was implied to each referrer that their referral's performance and willingness to continue working would affect whether the referrer was promoted.
The team experiment was designed to determine whether directly working with their referrers leads referrals to perform better. In the team experiment, the task was to work with an assigned partner to create a single, shared slogan for a public service announcement (PSA). Each of the two partners was given a different information sheet containing a distinct criterion for the slogan. Partners were asked to discuss the task and then to each submit the same slogan, which should have satisfied both criteria. Workers completed three such PSA tasks, each with a different partner. Each referral completed one task with their referrer and one task with another randomly-chosen referrer.
The selection experiment was designed explicitly to determine whether referrals contain information about worker quality. Four months after the peer influence experiment, the performance and persistence of referred and non-referred workers was measured in a job to which the referred workers had not been referred. Job offers were made to all referred and non-referred workers from the peer influence experiment. Similar to the peer influence experiment, workers who accepted the job offers were given a task that measured individual diligence over time. Workers were asked to visit the Twitter pages of three successful musicians and to answer a ten-question survey about those accounts every day for five consecutive days.