Experimental Design
The design of the intervention was guided by our baseline survey data, which revealed that short-form video viewing time among children was highly skewed, with some students reporting extremely high levels---far above the median---and viewing time was negatively correlated with children's attentional capacity and academic performance. Based on these facts, we introduced two types of information interventions: (1) Information nudge. Parents received SMS texts that short-form video consumption may hinder children's human capital formation. (2) Peer comparison. Parents of children with above-median viewing time were randomly assigned either to peer comparison group, in which they were informed of the sample median consumption level of short-form videos, or to a no-peer comparison group. Then we cross-randomized students to three information nudge treatment arms and a peer comparison treatment arm. This cross-randomization allows us to estimate the effects of the information nudge treatments and the peer comparison treatment, as well as their interactions.
Information Nudge Treatment.---Each student was randomly assigned with equal probability to one of three information nudge treatment groups or to a control group. The amount of provided information about the harmful effects of short-form video consumption increased progressively across the three treatment groups. All information was delivered to parents twice a month via SMS messages.
(i) Attention Impairment: Parents received two specific pieces of information: (a) a reminder of their child's baseline daily short-form video consumption, and (b) a statement that short-form video viewing is associated with declines in children's attention. We assigned 1,386 students to this group.
(ii) Attention Impairment + Quantitative Evidence: This group received the same information as the Attention Impairment treatment, along with quantitative evidence from our baseline survey on the relationship between short-form video consumption and children's attentional capacity. Specifically, parents were additionally informed that students who watch short-form videos, on average, perform 4.9\% lower in attentional capacity, and that each additional ten minutes of viewing is associated with a 0.4\% decline in attention.
A total of 1,387 students were assigned to this group.
(iii) Attention + Cognitive Impairment: Since academic performance is a critical determinant of high school and college enrollment, as well as long-term labor market outcomes in China, parents may be more inclined to restrict their children's short-form video consumption once they are informed of its academic harms. We therefore designed a treatment providing information about the negative association between short-form video use and academic performance, which may induce stronger responses for parents who may place excessive weight on this most salient information. Specifically, this group received the same information as the Attention Impairment + Quantitative Evidence treatment, along with empirical evidence from our baseline survey showing that students who watch short-form videos score 2.4 points lower on average in Chinese and mathematics, and that each additional ten minutes of viewing is associated with a one-point decrease in average test scores. This treatment was assigned to 1,389 students.
(iv) Control: This group did not receive any information treatment regarding the effects of short-form video consumption. To address concerns that reminding parents about their children's baseline video consumption might influence parents' behavior, we provided parents in the control group only with their children's reported daily viewing time of short-form video, as obtained from the baseline survey. A total of 1,386 students were assigned to this control group.
Peer Comparison.---We also designed an alternative intervention to encourage parents to restrict their children's short-form video consumption. This design was motivated by existing literature demonstrating that when peers' behavior is observable, individuals may adjust their own behavior in response to social pressure or conformity to social norms. To implement this intervention, we cross-randomized the information nudge treatment with a peer comparison intervention targeted at students with relatively high viewing time. For the peer comparison intervention, we restricted the sample to children who reported watching short-form videos at baseline and assigned treatment to those with above-median viewing time. We make this restriction because factors that affect students who watch versus do not watch short-form videos, and factors that affect students who watch more versus less, may be systematically different. In other words, we focus on the intensive margin effects of the intervention on viewing time among students with relatively high viewing time.
Specifically, parents of children with above-median viewing time were randomly assigned with equal probability to either the peer comparison treatment or a control group.
Parents in the treatment group received SMS messages containing two pieces of information: (a) the median viewing time in our sample (15 minutes per day), and (2) their child's baseline daily viewing time. Parents in the control group received only their child's baseline viewing time.
Implementation
Our experiment was conducted over a five-month period, from February to June 2024. Starting in February 2024, parents in treatment groups received standardized SMS messages twice per month, resulting in a total of ten messages during the study period.
Each message followed a uniform structure, consisting of: (1) a personalized greeting including the child's name, (2) individualized information on the child's reported viewing behavior, (3) treatment-specific content, and (4) sender identification from the official Education Bureau.
The information on children's daily viewing time was drawn from the baseline survey. Within each treatment arm, the treatment-specific content was identical across all families.
Messages were automatically delivered on the 10th and 20th of each month at 6:00 AM.
All SMS messages were pre-programmed into an automated delivery system, which could track delivery confirmations. Any unsuccessful deliveries were automatically rescheduled within 24 hours to ensure consistent exposure.
To minimize cross-group contamination within schools, parents were not informed about the existence of alternative treatment conditions, and school personnel remained blinded to families' treatment assignments throughout the study.