Background & Summary

Adolescence is a stage of major physical, psychological, emotional and social development, representing a crucial period in human life. The experiences, skills and habits that are accumulated during this stage have a permanent impact on human life. Therefore, understanding the behavior of individuals throughout this period is essential to supporting their development and ensuring their success in adulthood. Indeed, there is great interest in underlying motivations of adolescent behaviors for the design of public policies1.

It is widely recognized that individual preferences and cognitive abilities are important determinants of real-life decision-making of adults in strategic and non-strategic situations2,3,4,5,6. To understand and predict adult behavior, it is essential to comprehend how their attitudes toward risk, social and time preferences, cognitive abilities, creativity, and other traits evolve, particularly in their younger years7,8,9,10,11,12,13,14,15,16,17,18.

The dataset presented here contributes to the literature on adolescence by eliciting-using the tools of experimental economics-rich information on economic preferences, cognitive abilities, strategic thinking behavior and other information from a large set of adolescents in Spain. We conducted lab-in-the-field experiments in 33 different educational centers, accounting for a total of 5,890 observations of Spanish students. The centers belong to 19 different locations. A total of 20 of them are public and the rest are semi-private. In addition to socio-demographic details and other variables related to the individual, the data includes several sets of variables: economic preferences, cognitive abilities, strategic thinking and other additional measures.

Our dataset can contribute to future research on adolescents in at least two ways. First, it allows researchers to study adolescent decision-making and understand developmental causes of anomalous behavior. Second, it provides information on economic preferences, cognitive skills and other individual information, enabling exploring the extent to which these variables are sensitive to the class and school environment.

This dataset has been previously employed in the following studies: (i) An analysis of the relevance of monetary incentives, experimental tools and protocols to collect data in schools19, (ii) a study of the impact of visual aids in experimental lottery tasks to reduce inconsistency among adolescents20, (iii) the development of time and risk preferences throughout the adolescence21, (iv) the dynamics of social preferences among girls and boys22 and (v) the use of coordination devices among adolescents23. These studies as well as information about the TeensLab can be found on our website (https://loyolabehlab.org/teenslab/).

Methods

Data acquisition

In conducting research with minors, adherence to legal frameworks and ethical guidelines is essential. Spanish law governing the protection of personal data of minors allows for data processing based on the consent of children over 14 years of age (Art. 724). Although our study included participants over 14, we obtained informed parental consent through a parental council that had approved the study, enabling the integration of the study into the school curriculum. This consent authorized participation and the anonymous sharing and use of data within the scientific community. In this way, parental consent is collected by the center itself at the beginning of the scholar-year where they present the activities planned, including this experiment. This strategy not only simplified the process but also facilitated scalability.

Participants were informed about the purpose of data processing, the confidentiality of their responses, and the legal framework governing their data. Teachers managed participant lists and assigned identification numbers to ensure confidentiality. All responses were recorded anonymously.

Informed consent was additionally obtained from all participants on the initial screen of the experiment. This mandatory screen provided essential information in compliance with data protection regulations. Table 1 presents a translation of this information, which includes the identity of the data controller and a description of the rights participants may exercise.

Table 1 Initial screen of the experiment.
Full size table

Our experiment was approved by the Ethical Committee of Universidad Loyola Andalucía (No. 20190318, 20200709 and 20230301) Furthermore, for 10-year-olds, it was also approved by the Bioethics Commission of the University of Barcelona (No. IRB00003099).

To mitigate issues related to non-standard samples and minimize missing data, we simplified response formats, predominantly using multiple-choice questions rather than open-ended ones. The design of the software required that participants could not skip questions. However, for potentially sensitive topics, they were allowed to choose the option “I would prefer not to answer”.

The participant pool was recruited through agreements with school headmasters, who agreed to integrate the experiment into their pedagogical curriculum and to carry it out as a classroom activity. Consequently, we achieved a high level of participation19. The experiments were conducted on-site at schools using an online platform named Social Analysis and Network Data (SAND; https://sand.kampal.com), enhancing data privacy control. This platform allows students to navigate and complete the experimental questionnaire, which is divided into several sections, on their devices (tablets, computers, or smartphones).

The questionnaire was administered entirely in Spanish. Due to the restrictive school policies on experiments involving real money, we used hypothetical rewards. However, it has been documented that the behavior of adolescents does not differ between incentivized and hypothetical payment schemes at least for risk and time preferences, suggesting the reliability of the findings25,26,27,28,29,30,31.

Measurements

Table 2 contains all the tasks included in the study. Apart from basic information regarding the school (province, city and public/semi-private) and the class (stage, grade, group, class size), our dataset includes individual-level measurements for the following three behavioral dimensions:

  • Economic preferences: Time discounting, involving choices between immediate and delayed rewards (patience)19,32; risk preferences, assessed through decisions involving probabilistic outcomes (prudence)19,20; social preferences, measured via resource allocation tasks (egalitarianism, altruism, spitefulness)33,34,35; and honesty, evaluated through opportunities to misreport outcomes36.

  • Cognitive abilities: Cognitive reflection, overriding intuitive responses19,37; financial abilities, solving simple financial calculations19; probability knowledge and accuracy, measured via decisions in probabilistic scenarios38 and creativity, generating multiple original ideas using a single object39.

  • Strategic thinking: Subjects choices and expectations in strategic environments (games)22.

Table 2 Experiment summary by dimensions and observations.
Full size table

We also collected information regarding the participant’s family background and outcomes in school:

  • Socio-demographics: Age, gender, self-reported income, migrant status and family composition (number of siblings and her ranking).

  • GPA: The self-reported number of A’s and B’s scored in Mathematics, English and Spanish Literature during the previous year.

  • Physical appearance: Self-reported height, weight and appearance by Stunkard figure scale40,41.

  • Mood: Three items from the Kidscreen questionnaire about their interactions at school, assessing whether they have fun with their friends or feel lonely42,43.

Finally, for certain sub-samples (see available observations in Table 2), we gathered additional auxiliary information:

  • Expectations: Information regarding subjects’ expectations about their future outcomes, such as their university degree, traveling around the world, living abroad and desired future job.

  • Self-assessed math abilities: Two types of questions: “How good are you at maths?” and “How much do you like maths?”44.

  • Time discounting II: Time preferences (patience) measured by the compound staircase version45.

  • Time perception: Questions about future actions at three levels46.

Data Records

The dataset can be found in Zenodo47 (https://zenodo.org/records/13720112) and is available in different formats (xls, cvs, dta). The screenshots of the complete experimental instructions are also available in the repository. We also provide STATA 1848 scripts for some basic summaries of the available variables.

Sample variables

The experiments were conducted over multiple sessions from 2021 to 2023. A total of 5,890 students started the experiment, but 609 did not finish the entire questionnaire.

In contrast to adults, it is well-known that children and adolescents often find it more difficult to maintain concentration over extended periods and to complete all tasks13. Some of them simply leave the survey at a certain point. We check the responses after each session and reassess the tasks which were not successful. As a result of various adjustments made during the experimental sessions, the survey tasks underwent some changes. Consequently, the number of observations for different variables in our dataset varies. Table 2 provides an overview of the available observations for each task.

The initial questionnaire screen (Table 1) provided essential information about the study, including an introduction to our team and funding sources.

Figure 1 displays the distribution of the final sample by age and gender. The sample is well-balanced in terms of gender; 49.68% are female students and 49.68% male. The remaining subjects (0.64%) are classified as unknown, either because they did not answer or they selected another category.

Fig. 1
figure 1

Distribution of the sample in terms of age and gender. Note: The histogram contains three gender categories: Male, female and other/I prefer not to say (PNS).

Full size image

As for educational stages, 8.62% of the sample belongs to primary education, 84.94% belongs to secondary education, 1.90% to sixth form and 4.53% to vocational training. Table 3 presents the distribution of the observations by educational stage. Additionally, it displays the response rate and summary statistics for the ages at each stage.

Table 3 Descriptive statistics by educational stage.
Full size table

Technical Validation

The study is a laboratory-in-the-field experiment. Data were collected in the school classrooms under the supervision of team members and research assistants.

The data recorded in the software were downloaded for cleaning using Stata48. Variables were coded and incomplete entries were not deleted. Only the age variable was imputed through the year of birth reported by the students and according to the course to which they belonged.

Our experiment includes standard tasks from the literature as well as tasks adapted by our research team from previous literature19. We have extensive prior experience in designing experiments for teenagers and collecting data in primary and secondary schools using lab-in-the-field techniques. Previous evidence suggests that there are no significant differences in outcomes when using hypothetical payoff tasks, such as eliciting risk preferences27,28,29,30,31. Prior to each task, students were provided with a brief description and they were informed of the economic implications of their decisions in hypothetical terms. This ensured that participants fully understood the nature of the tasks, while maintaining the validity of the experimental design and the scalability of the study. Some pilots of the tasks were carried out independently to configure the final design. The changes in the survey are detailed in the variables descriptor available in the repository.

One of the main problems in collecting data from non-standard samples is that some tasks are not understood and participants show inconsistent behavior across them. To address this, our design took into account the results of pilots that adapted the tasks to the adolescent context through the use of visual aids. As a result, the consistency rates are remarkably higher than those reported in the literature7,19,20.

Consistency is assessed by examining how the choices of participants align with their stated preferences in different situations, based on their personal decision-making patterns. Among the data reported for the economic preferences dimension, we find a high percentage of consistent responses in the tasks that require certain within-task consistency. We observe that 82.75% of the individuals who complete the time preference task exhibit consistent behavior. Similarly, 79.20% of individuals report consistent answers in the risk preferences task20. Table 4 includes a distribution of responses for both tasks across their 6 decisions, where a trend can be identified that may represent this high level of consistency. Such enhanced consistency indicates that the data collected from adolescents are reliable and coherent, providing a robust foundation for examining adolescent decision-making processes and developmental trends.

Table 4 Percentage of responses for each decision.
Full size table

Usage Notes

The Zenodo repository gives access to the available data together with a descriptive note on the variables and their coding. The variable descriptor includes a definition of the task, some general characteristics, and the specific name under which it is found in the database. We provide further information on the changes that the survey has made over time. In addition, the repository visualizes the experimental screens in the original Spanish language.

This article is licensed under a Creative Commons Attribution (CC BY) 4.0 International License.