Parental Beliefs, Attention, and Action in Child Mental Health

The intervention consists of brief information messages delivered to parents during an online parent survey. Parents first report baseline beliefs about their child’s recent mental-health state. They are then randomly assigned to receive either control information or information related to adolescent mental health. Treatment messages are designed to make salient one or more of the following: the possibility that parents may underestimate children’s hidden emotional distress; the longer-run consequences of adolescent mental-health problems for later mental health, education, life satisfaction, and labor-market outcomes; and the importance of treating child mental health as an intrinsic parental objective alongside academic performance. The intervention is delivered within the survey, and outcomes are measured immediately after the information is provided.

The study includes two related information experiments.

Experiment 1 is embedded in a parent-child paired survey. Students complete paper surveys in class and parents complete linked online surveys. After parents report baseline beliefs about the child’s recent mental-health state, parents are randomly assigned to a treatment or control condition. Treated parents receive information based on two components: parent-child perception gaps in adolescent mental health and the consequences of adolescent mental health for adult outcomes. Control parents do not receive this information before post-treatment belief and intention measures. Post-treatment outcomes include belief updating, perceived seriousness/concern, and intended parental responses such as talking with the child, reducing academic pressure, and seeking help.

Experiment 2 is a parent-only mechanism experiment designed to separate information and preference channels. Parents are randomly assigned to one of five arms: (1) control information about general adolescent development; (2) perception-gap information emphasizing that parents can underestimate hidden distress in children; (3) consequence information emphasizing that adolescent distress affects later mental health, education, life satisfaction, and labor-market outcomes; (4) combined information including both perception-gap and consequence information; and (5) mental-health-priority information emphasizing mental health as an intrinsic parental objective alongside grades. Outcomes are measured immediately after treatment and include beliefs about child distress, beliefs about parents’ own underestimation, beliefs about long-run consequences, academic-versus-mental-health priorities, behavioral intentions, demand for additional mental-health guidance, and intended time allocation toward support-oriented activities.

2026-05-11

2026-09-11

The primary outcomes are:
1. Parent belief updating about the child’s recent mental-health state.
2. Post-treatment parental concern or perceived seriousness of the child’s mental-health situation.
3. Supportive parental behavioral intentions, including intentions to talk with the child, reduce academic pressure, seek help, and allocate time toward supporting the child’s emotional well-being.

Parent belief updating is defined as the difference between the parent’s post-treatment belief and baseline belief about the child’s recent mental-health state, with higher values indicating upward updating toward greater perceived distress.

Post-treatment concern is measured using a standardized index of parents’ perceived seriousness of the child’s mental-health situation and perceived need to rethink their response.

Supportive parental behavioral intentions are measured using a standardized index constructed from post-treatment intended responses. These include whether the parent intends to talk with the child about stress or emotions, reduce academic pressure or reassess pressure placed on the child, seek professional or other help if needed, demand additional mental-health information or guidance, and allocate more time to support-oriented activities intended to improve the child’s emotional well-being. Component outcomes will also be analyzed separately.

Secondary outcomes include:
1. Accuracy of parental beliefs about the child’s mental-health state.
2. Second-order belief accuracy: parents’ predictions of how the child would report their own mental-health state.
3. Beliefs about parents’ own likelihood of underestimating the child’s distress.
4. Beliefs about long-run consequences of adolescent mental health for adult outcomes.
5. Academic-versus-mental-health priorities.
6. Comprehension of the information treatment.
7. Heterogeneous treatment effects by baseline underawareness, belief type, academic-priority orientation, parenting style, and baseline beliefs about consequences.

Belief accuracy is measured by comparing parent-reported or parent-predicted child mental-health measures with the child’s own self-report when linked child data are available. Baseline error is defined as child report minus parent baseline belief; positive values indicate parental underestimation. Accuracy-related outcomes include error reduction, absolute post-treatment error, signed second-order error, and absolute second-order error.

Parents’ beliefs about their own underestimation are measured using survey items asking whether parents think their judgment may understate the child’s problems relative to the child’s own report. Consequence beliefs are measured using items on whether adolescent mental-health problems affect later mental health, education, life satisfaction, and labor-market outcomes. Academic-versus-mental-health priorities are measured using survey items capturing whether parents prioritize grades and exam success relative to mental health and emotional well-being. Comprehension is measured using post-treatment checks of whether parents understood the information. Heterogeneity analyses will examine whether treatment effects differ by baseline underawareness, truly underaware versus strategically inattentive belief type, academic-priority index, child-reported low-responsive parenting style, and baseline consequence beliefs.

The study uses randomized information experiments conducted with parents of middle- and high-school students. Parents complete an online survey in which they first report baseline beliefs about their child’s recent mental-health state. They are then randomly assigned to receive control information or one of several brief information treatments related to adolescent mental health. The treatments provide information about parent-child perception gaps, the long-run consequences of adolescent mental-health problems, and/or the importance of mental health relative to academic performance. Post-treatment outcomes are measured within the same survey and include belief updating, concern, priorities, and intended parental responses.

The project combines a parent-child paired survey with randomized information experiments.

In the paired survey, students complete paper surveys in class and parents complete linked online surveys. The sample includes 3,231 matched parent-child pairs from 8 middle schools and 8 high schools across 8 cities and 4 provinces in China. Parent and child surveys include parallel questions about the child’s emotional state, academic stress, parent-child communication, parental involvement, and parental responsiveness. Child mental health is measured using an 8-item CES-D scale. Parent beliefs are elicited before treatment.

Experiment 1 is embedded in the parent survey after baseline belief elicitation. Parents are randomly assigned to treatment or control. The treatment provides information about parent-child perception gaps and the consequences of adolescent mental health for adult outcomes. The main analysis compares treated and control parents on post-treatment belief updating, concern, and behavioral intentions. We will also examine heterogeneity by baseline underawareness, where baseline error is defined as child report minus parent baseline belief; by belief type, distinguishing parents who are truly underaware from those who are strategically inattentive using second-order belief questions; by academic-priority orientation; by parenting style; and by baseline consequence beliefs.

Experiment 2 is a parent-only mechanism experiment. Parents are randomly assigned to one of five arms: control, perception-gap information, consequence information, combined perception-gap plus consequence information, and mental-health-priority information. The design is intended to identify first-stage effects on intermediate beliefs and priorities and to assess whether these mechanisms mediate changes in supportive parental responses. Outcomes include perceived child distress, perceived likelihood that the parent underestimates the child, beliefs about long-run consequences, mental-health priority, behavioral intentions, demand for additional mental-health information, and intended support-oriented time allocation.

The primary estimating equation will regress each outcome on treatment indicators, controlling for baseline beliefs and pre-specified demographic covariates where applicable. For the paired-survey experiment, specifications may include controls for child gender, grade or school stage, parental education, income, and other pre-treatment covariates. Standard errors will be robust; if randomization or sampling induces clustering at a higher level, standard errors will be adjusted accordingly. Treatment effects will be reported as average treatment effects and, for pre-specified subgroups, as interaction effects.

Randomization is conducted by computer within the online survey platform after baseline belief elicitation. Assignment occurs automatically at the individual parent level. In the parent-only mechanism experiment, the computer assigns parents to one of five arms: control, perception-gap information, consequence information, combined information, or mental-health-priority information. In the paired-survey experiment, the computer assigns parents to treatment or control after baseline parent beliefs are recorded.

The unit of randomization is the individual parent respondent. Each participating parent is assigned independently to a treatment or control condition within the online survey.

No

Not applicable. The treatment is randomized at the individual parent level and is not clustered.

more than 3000 parent-child pairs

1500 parent-child pairs in control group, and the other 1500 in treatment group