The research project aims to investigate one mechanism that could contribute to the phenomenon of overconfidence, namely the feedback individuals receive from others. We study this mechanism in the context of parents providing feedback to their child, which constitutes a key instance where this mechanism may be important. This lab-in-the-field experiment will allow parents to choose to send their child feedback in the form of an assessment of their future or past performance in different scenarios.
Within this context, we will address three main research questions. First, focusing on parents, we ask whether parents tailor the feedback they send to boost their child’s confidence. Second, we explore whether parents’ feedback decisions differ according to the (perceived) gender congruence of the task, comparing feedback behavior for a stereotypically male task (mathematics) with a stereotypically female task (Norwegian). Third, we shift attention to the child and ask whether the feedback decisions of parents influence the child’s beliefs and performance. Specifically, we ask whether parents’ feedback influences the gender gap in confidence, motivation, and performance of children (e.g., does parental feedback have a more positive influence on confidence in gender-congruent tasks). Our study will help to reveal the aims of parents in distorting the feedback they provide to their children.
Intervention
Our experimental design is described in detail in the attached PDF. This includes a detailed description of our interventions.
It will involve a lab-in-the-field experiment run online with parents and with children in Norwegian secondary schools (7th grade).
Experimental design
Our experimental design is described in detail in the attached PDF.
It will involve a lab-in-the-field experiment run online with parents and with children in Norwegian secondary schools (7th grade). We will vary the feedback that parents provide to their children, examining both how parents make their feedback decisions and the influence this has on their child's beliefs.
Design details
Randomization method
The randomization will be implemented by the computer program (oTree) with an equal probability (p=1/5) of being assigned to each of the five treatment conditions.
Randomization unit
Randomization is implemented at the level of parent-child pair (meaning the parent and their child).
Clusters
The children will take part in the experiment at their school. Randomization to the treatment groups will be done at the individual parent-child level, within schools.
We aim to collect 750 complete parent-child pairs.
Since parents and children complete the survey and experiment separately, this implies that we will collect data from some children (and possibly some parents) who are not part of a matched pair. In particular, parents need to complete their part of the survey in a short space of time in the morning before their child participates. Therefore, we expect that a fraction of the parents will not participate in time (perhaps 25%). We will still allow their child to participate in this case, but this data from children whose parents do not participate will not form part of our main dataset. These unmatched children will effectively face the T0 condition (no parental feedback), but they will constitute a selected sample. Therefore, we will continue to collect data until 750 parent-child pairs participate.
Since we expect this non-participation to take place before exposure to any treatment, we expect that it will be orthogonal to our treatment conditions.
Our anticipated effective sample size will therefore be 750 parent-child pairs, with ⅕ randomized into each of the 5 treatment conditions.
Observations
Each treatment arm in our effective sample will contain approximately the same number of participants (150), meaning that parents (along with their child) will have an equal probability (p=1/5) to being assigned of each of the five treatment arms. We will continue to recruit parent-child pairs until we reach a sample size of 750 complete parent-child pairs.
Arms
As noted above, we plan to continue to recruit parent-child pairs until we reach a sample of 750 complete parent-child pairs. They will be randomized to each of the five treatment conditions with equal probability. This will imply a sample of approximately 150 parent-child pairs in each of the 5 treatments.
Primary outcomes
PARENTS:
I) Belief about child's performance (Maths)
II) Belief about child's performance (Norwegian)
III) Feedback to child about the child’s performance (Maths)
IV) Feedback to child about the child’s performance (Norwegian)
CHILDREN:
i) Belief about own performance (Math): Confidence
ii) Belief about own performance (Norwegian): Confidence
iii) Competition choice