Incentives to Persevere

We design different incentive mechanisms aimed at promoting perseverant effort.

2016-07-25

2021-10-22

Full completion rate of the tasks; zero completion rate; completion rate after one failure; when do subjects start to work on the task in each week; the quality of the completed work .

Study 1
We addressed undergraduate students enrolled in the introductory microeconomics course. For each week of lecture during the semester, the lecturer created pre-lecture contents to facilitate learning at home. Each week, students were given an online quiz that tested their understanding of the contents prior to the lecture. There were 9 quizzes that contributed a total of 10% towards each student’s final grade in the unit. Students were randomly assigned one of the three treatments: piece rate, all-or-nothing, and self-select. At the beginning of the first tutorial session, tutors were given the scripts to read to students. The instructions provided details about the types of incentives.

Study 2
In Study 2, participants were invited to work on an online survey over four weeks (one in each week). Depending on treatment, the subject would either receive the payment based on the number of survey they complete (piece rate treatment) or would have to complete all surveys to receive the payment (exogenously imposed all-or-nothing treatment). We also allowed the subjects to choose between the piece rate or the all-or-nothing option (the self-select treatment).
Treatments in our subsequent studies (Study 3-6) that are similar to Study 1 and 2 except in some treatments we encourage the participants into choosing the ‘all-or-nothing’ reward mechanism in different ways. As in all the studies, we include both the baseline and the Self-select treatments, we are able to check whether we could replicate the effect of “self-select treatments” in different years and/or in different courses.

Study 1 (July to October 2016 – 343 participants)
Study 2 (October 2016 and May 2017 – 406 participants)
Study 3 (July to October 2017 – 336 participants)
The participants were undergraduate students, and the design was similar to Study 1 but we modified the treatments. There were three treatments – baseline (piece-rate) and self-select (no encouragement) and self-select (with encouragement). In the treatment with encouragement, we nudged the participants into choosing the ‘all-or-noting’ mechanism with a message.

Study 4 (February to June 2018 – 500 participants)
This study is similar to study 3, but we nudged the participants into choosing the ‘all-or-nothing’ mechanism by allowing them to miss 20% of the tasks. So there are three treatments - baseline (piece-rate) and self-select and self-select (with lenient all-or-nothing option). In addition, we also ran experiments to elicit the participants’ time and risk preferences and conducted surveys to examine their level of grit and self-control.

Study 5 (July – October 2018 -- 373 participants)
Similar to Study 4, but the tasks are online-surveys rather than quizzes. In addition, we also added the screener questions to identify the ‘shirkers’ who answered the survey questions without reading the questions carefully.

Study 6 (February to June 2019 – 756 participants) – Introductory Microeconomics &Introductory Macroeconomics
In this study, we examine how the previous experience in Self-select mechanism affect the performance when participants are assigned to the Self-select mechanism again. The study involve two phases. In Phase 1 (February to June 2019), students were assigned to either baseline (piece-rate) or self-select (with encouragement) (see study 3). In Phase 2 (July 2019), all the participants are assigned to “Self-select with encouragement” treatment. We would like to track the behaviour of the students who also participated in Phase 1. We are interested in the following research questions: would participants who chose the all-or-nothing mechanism but received zero reward in the first semester choose the same option again? Would participants who chose the all-or-nothing mechanism kept choosing all-or-nothing again? Would those who opted for piece-rate in Phase 1 decided to choose ‘all-or-nothing’ this time, especially when they did not have a full completion rate?

Study 6 will be repeated in 2020 to obtain more data, especially those who would self-select the all-or-nothing option in phase 1.
- Due to the pandemic the data collection was disrupted in 2020 but will resume in 2021.

Study 5 (March - June) and Study 6 (March -October) will be repeated in 2021 to obtain more data. In the additional trial for Study 5 (survey tasks), we will include an additional treatment with reminder - reminding those who have completed 80% of tasks that they are qualified for full payment and that their next task is optional.

Randomization done in office by a computer.
Study 2 - for data to be collected in 2021 - block randomization in qualtrics

Quizzes: students were randomly allocated to different tutorial groups, and the treatments were randomised by tutors
Surveys: randomization was done by gender and school

No

Approximately 3900 students

Approximately 3900 students

Study 1 (343): 103 baseline; 86 all-or-nothing; 154 self-select
Study 2 (406): 132 baseline; 131 all-or-nothing; 143 self-select
Study 3 (336): 103 baseline; 131 self-select (no encouragement); 102 self-select (with encouragement)
Study 4 (500): 114 baseline; 198 self-select (no bonus); 188 self-select (with bonus)
Study 5 (373): 114 baseline; 146 self-select (no encouragement); 113 self-select (with encouragement)
Study 6 (549): Phase 1 132 baseline; 417 self-select (encouragement)
Phase 2 (207): 207 self-select (encouragement)

2021 (planned)
Repeat Study 5 - 100 baseline; 150 self-select (no encouragement); 150 self-select (with encouragement)
Repeat Study 6
Phase 1 (Introductory micro class): 240 baseline; 160 self-select (encouragement)
Phase 2(Introductory macro class): 300 self-select (encouragement)