Effect of extracurricular tutoring on adolescent students cognitive ability: A propensity score matching analysis

In recent years, there has been a substantial increase in the number of adolescent students attending extracurricular tutoring. However, extracurricular tutoring, being a distinct form of education, may have varying effects on cognitive capabilities compared to conventional education. Accordingly, the purpose of this paper is to conduct a thorough examination of the effects of extracurricular tutoring on cognitive abilities among adolescent students. The study utilized national data from the China Family Panel Study 2018 to explore the relationship between involvement in extracurricular tutoring among students and cognitive abilities. The study included 2567 valid respondents. A binary logistic regression model was built to determine the factors associated with tutoring involvement while controlling for 19 individual, family, and school factors. Furthermore, a propensity score-matched analysis was conducted in order to mitigate potential bias by using confounding variables identified in the previous step. The study results show that participation in extracurricular tutoring can significantly increase the level of cognitive ability of adolescent students, with adjustments made for age, gender, ethnicity, number of family members, net family income per capita, education and training expenditure in the past years, change of residence for enrollment, change of domicile address for enrollment, locality of the current school, class size, hold a position as a class cadre, average daily study time on weekdays, average study time per day during weekends. The findings imply that the government should provide tutorial subsidies to disadvantaged groups of adolescent students, allocate educational resources equitably, and invest more in education resources in less developed regions to foster fair and healthy development of education and improve the cognitive abilities of young students in the long-term.


Introduction
Participating in extracurricular tutoring has emerged as a global educational phenomenon, particularly in China. [1,2]This practice is shaped by various socio-cultural factors beyond the education system, including social competition and family expectations, cultural attitudes towards education, and traditional beliefs about knowledge and learning.In particular, in many East and South Asian countries, socio-cultural expectations and norms foster intense competition and high levels of stress among parents and students that drive the demand for extracurricular tutoring to enhance academic performance and skill acquisition.Besides, in some cultural contexts, better access to knowledge and educational opportunities is considered critical for personal success; as a result, extracurricular tutoring is viewed as a mechanism for improving learning outcomes and boosting competitive advantages.In certain Middle Eastern and Asian nations, extracurricular tutoring is viewed positively, with strong cultural values placing a high premium on education and learning.The multifaceted cultural impact of extracurricular tutoring reflects a combination of factors, including social, familial, and educational dimensions, and its effects on the long-term development of education and society remain uncertain.Despite the issuance of several government regulations in recent years to limit extracurricular tutoring, it continues to be a pervasive and popular phenomenon due to the growing competition for better life opportunities and the perception of the inadequacy of traditional school education to meet these challenges. [3]xtracurricular tutoring refers to diverse educational activities that take place outside regular school hours, including subjects such as language, mathematics, English, music, sports, and art lessons, among others. [4,5]Various types of extracurricular tutoring are emerging, ranging from personalized 1-on-1 sessions to group classes, online courses, homework clubs, and camp-style programs. [6,7]While some studies have shown that the average effect of private tutoring is not significant, it may have notable and positive effects on specific subjects for urban students with lower academic achievement or those in low performing schools. [8]Recently, the government has been emphasizing the need to appropriately reduce the academic burden on primary and secondary school students, and has stressed the importance of schools cultivating students interests and innovative spirit comprehensively.The government has also issued measures to enforce these requirements in schools.Schools have responded accordingly by implementing reforms, such as reducing the amount of in school class time and adding courses that help develop students interests.However, due to the competitive educational environment, extracurricular tutoring is still very popular in primary and secondary schools. [9]While research has emphasized the significance of cognitive development in educational outcomes, there is limited knowledge of the effects of extracurricular tutoring on this development, as it varies from traditional teaching methods.Therefore, it is essential to undertake more research on extracurricular tutoring to gain a better understanding of its impact on students cognitive development.
extracurricular tutoring plays a crucial role in enhancing students' cognitive capabilities.Cognitive ability encompasses a range of skills, including the ability to process, store, and extract information.[12] Furthermore, cognitive ability constitutes the foundation for completing activities [13] and can be divided into general and specific cognitive ability. [14]he former is generally referred to as intelligence, while the latter includes language, memory, calculation, reasoning, decision-making, and spatial abilities. [15]Therefore, cognitive ability is fundamental to learning, and it partly determines an individual's developmental prospects. [16]The impact of extracurricular tutoring on cognitive development is often manifested through the provision of additional learning opportunities, building a broad knowledge network, strengthening the analytical and problem-solving skills, improving memory, and cultivating good study habits.
Although extracurricular tutoring has been associated with positive outcomes on cognitive development, the impact of such programs on adolescent students is complex and influenced by several factors.There is limited research in this area, which can be attributed to various reasons.Firstly, there is a lack of a unified evaluation system and tools for research and comparison, as extracurricular tutoring comes in various forms. [17]Secondly, there is sample selection bias, as clients come from different family backgrounds and levels of education and economic status, making it challenging to compare different young people randomly.Finally, there is diversity in extracurricular tutoring, including the timing, form, and content of tutoring, which limits the ability to conduct specific breakdowns and analysis. [17]This complexity has resulted in the need for methods such as propensity score matching (PSM) to control for confounding variables that may influence the results of research studies.
In this study, to overcome these problems, we have used the cognitive ability test results from the China family panel studies (CFPS), [18,19] and PSM to investigate the impact of extracurricular tutoring on young people's cognitive abilities.The PSM method is a statistical technique that controls for jointly dependent covariates that affect decisions to procure extracurricular tutoring and cognitive development.In doing so, PSM reduces the selection bias caused by confounding variables across the 2 different groups of young people.[22][23][24][25][26] The selection of appropriate confounding variables is, however, crucial in the application of PSM.In our study, the confounding variables selected aligned with teenagers' cognitive ability, and their involvement in extracurricular tutoring was included in the study.
The literature suggests that engaging in extracurricular tutoring is influenced by individual, family, and school factors. [27]ndividually, students with higher grades, those living with their parents, and day students tend to be more likely to engage in extracurricular tutorials. [28,29]From a family perspective, factors that include parental education and family income play a crucial role in whether students attend extracurricular tutoring or not. [7,30]In terms of school characteristics, students who attend central city schools and public schools with higher rankings tend to participate in extracurricular English tutorial classes. [28]ble 1 Covariate measurement and attributes.Furthermore, the proportion of urban students who take extracurricular tuition is significantly higher than that of rural students, with a difference of 26.31%. [31]fter reviewing the literature, we noted that confounding variables can adversely affect the relationship between adolescent participation in tutoring and cognitive ability.These variables contribute to observed data selection bias, making it difficult to conclude whether participation in extracurricular tutoring affects adolescents' cognitive abilities.However, adopting the propensity score matching method can improve the methodological standards and help researchers make stronger causal inferences from observational data. [32]Although PSM method has its advantages, it is not the ultimate solution to all endogeneity problems.However, if applied appropriately, PSM can effectively reduce selection bias between the treatment and control groups.The PSM method involves matching adolescents not involved in extracurricular tutoring with the control group, balancing the treatment group.By calculating the average treatment effect on the treated (ATT) between the treatment and control groups for those treated, researchers can provide stronger evidence to validate the effect of extracurricular tutoring on adolescents cognitive abilities.In this study, we identified 19 relevant confounding variables and selected them synthetically as covariates for PSM based on relevant literature and variable availability.

Sampling
This study analyzed the data from the CFPS (https://opendata.pku.edu.cn/dataverse/CFPS), a social survey sponsored by the Institute of Social Science Survey of Peking University.Launched in 2010, the CFPS permanently tracked all family members and their future biological or adopted children [33][34][35] to obtain reliable data for academic research.The survey covered 25 provinces, municipalities, and autonomous regions, with a target sample size of 16,000 households.This study focused on young people aged between 9 and 18 years old, using the 2018 Individual, Family, and Household Economy database.The family information in the adult pool was first matched against the parent sample number in the individual pool, and then against the family sample number of the individual's family to obtain the family economic information.
In order to achieve the research objectives of this paper, sample cases were selected based on data from the sample of young people and parents in the CFPS survey using the following criteria: Exclusion of cases with missing values for selected variables; Exclusion of individual cases with learning disabilities, intellectual disabilities or other cognitive impairments; Exclusion of cases where answers did not match reality, such as those with sleep hours exceeding 12 or <4 hours, those with an average of more than 6 hours of extracurricular tutoring per day from Monday to Friday, and cases with an average of more than 12 hours of extracurricular tutoring per day on weekends.The final valid sample size obtained was 2567.The CFPS was reviewed and approved by the Biomedical Ethics Committee of Peking University, and the data collection was carried out with the approval of the ethical review.The approval number is IRBO000105214010.

Cognitive function
The cognitive measures used in this study were based on a separately designed cognitive module in the CFPS questionnaire.The module used a set of word test and a set of math test to examine the cognitive level of the respondents.The word test assessed the linguistic level of the respondents by flipping the tab to the appropriate position and asking respondents to pronounce the word or phrase in that position.The math test included 4 different sets of mathematical questions, with 1 set randomly selected for respondents to answer.This measure of cognitive level provided a more comprehensive picture of the respondent's cognitive ability.
Cognitive ability was quantified by the respondents' word test scores (ranging from 0-34) and math test scores (ranging from 0-24). [19,36]A higher score indicated a better level of cognitive development.

Extracurricular tutoring
extracurricular tutoring is a paid educational service provided by private individuals or training institutions to schoolchildren outside of regular school hours with the aim of improving academic performance or specific skills, such as art, sports, or academic competition.extracurricular tutorials in this study included both curriculum tutorials aimed at improving academic performance and comprehensive quality development aimed at developing students' specific skills in a particular area.The 2018 CFPS database included a question regarding participation in tutoring or tutoring classes.Respondents who answered "yes" were considered to have participated in extracurricular tutoring.Of the 2567 eligible participants, 564 had participated in extracurricular tutoring.

Covariates
For this study, 19 questions from the CFPS database were selected as covariates that might affect the cognitive development and decision-making of adolescent students in relation to extracurricular tutoring.The questions covered 3 main aspects: including: adolescents characteristics, such as gender, age, and ethnicity; family characteristics like the number of family members and family financial situation; and school characteristics such as whether the adolescent attends a public school or not and currently enrolled in which stage of education.Table 1 displays the 19 confounding covariates used in this study and their attributes.

Statistical analysis
Data analysis was performed using Stata 15.1 SE.Normally distributed continuous variables were expressed as means ± standard deviations, and categorical variables were expressed as frequencies or percentages.Two independent samples t tests were used for continuous variables, and the Chi-square test was used for categorical variables. [37]Missing continuous variables were imputed using means or medians. [38]o address selective bias in the observed data, PSM was used to establish cohorts with similar baseline characteristics.PSM is a statistical technique that measures the effects of an intervention without the need to randomize the subjects of the intervention.Augmenting the foundational information, the paper elaborates on the PSM method by explaining the necessary steps to conduct the analysis.The first step is to calculate a propensity score, and the second step is to match against the propensity score to estimate the impact of the ATT. [39,40]Thus, statistical analyses were conducted using Psmatch2 in Stata 15.1 SE to assess the relationship between participation in extracurricular tutoring and adolescent cognitive abilities.To start, we estimated a binary logistic regression model using participation in extracurricular tutoring as the dependent variable and 19 individual, school, and family factors as independent variables.Next, we conducted propensity score analysis to control potential bias using confounding covariates identified in the previous step.For this analysis, PSM used concepts and terminology from experiments, considering the nontreatment unit as the control group and the treatment unit as the intervention group.In our study, the "control group" consisted of adolescents who did not participate in extracurricular tutoring, and the "intervention group" consisted of those who did participate.
Moreover, to compare cognitive abilities between adolescents who did and did not participate in extracurricular tutoring, paired t tests were conducted for the propensity score matching group.We used 3 common matching methods, kernel matching, radius matching, and nearest neighbor matching, between the steps of matching and estimating ATT, to ensure the reliability of the results. [24,41]raditional ATT estimation produces erroneous standard errors due to the high correlation between matched samples and the error between propensity values and actual values.To combat this issue, empirical standard errors were obtained by using the bootstrap method.For the treatment group, this study used the bootstrap method to estimate ATT and empirical standard errors. [42] Results

Descriptive statistics
This study presents an overview of the participants' basic characteristics and provides a descriptive analysis of 2 distinct Participating in these activities showed a higher level of cognitive ability among adolescent students.However, it is not straightforward to infer a causal relationship between extracurricular activities and cognitive ability.Nonetheless, there were significant differences between the 2 student groups on 13 covariates.According to the study, students who were younger, had fewer family members, had higher per capita household income and higher education and training expenditure in the last year, had larger class sizes, had more  average weekly study time, were female, were Han Chinese, had changed their residential and domicile addresses for further studies, attended urban schools, and were class cadres were more involved in extracurricular tutoring and had higher cognitive ability.This finding suggests that a causal relationship cannot simply be inferred between adolescent students' participation in extracurricular tutoring and their level of cognitive ability.

Propensity score
First, logistic regression was performed with participation in extracurricular tutoring as the dependent variable and 19 covariates as independent variables.All variables showed statistical significance except for: age, change of residence for enrollment, change of domicile address for enrollment, currently enrolled in school, enroll in the nearest school, have you ever requested leave from school, and have you ever  skipped class.These results suggest that the model is strongly predictive of whether adolescents participate in extracurricular tutoring, as shown in Table 3.The regression results were then used to build a predictive model to calculate the propensity score of all youth participating in extracurricular tutoring in the sample.The higher the propensity score, the greater the likelihood that adolescents will engage in extracurricular tutoring.

Matching and balanced test
We used calculated propensity scores to match 564 students who participated in extracurricular tutoring with 2003 who did not.To ensure the robustness of the results, we employed 3 common matching methods, including kernel matching, radius matching, and nearest neighbor matching.The kernel matching successfully matched 553 pairs of samples, the  radius matched 538 pairs of samples, and the nearest neighbor matched 553 pairs of samples.The overall matching success rate was close to 100%, indicating that the 3 methods had a good overall matching effect.The standard deviation of the 2 samples after PSM should be <5%, and the independent samples t-test for each covariate between the control and treatment groups should no longer be significant.In addition, there were no significant differences in the 19 covariates between the control and treatment groups after matching (Table 4 and Figs.1-6), indicating that these covariates did not provide any valid predictive information for participation in extracurricular tutoring after matching.Therefore, it passes the overall balance test.

Sensitivity analysis
Rosenbaum proposed that the purpose of sensitivity analysis was to explore the extent to which hidden selection bias could alter the outcome of the treatment effect obtained, [43,44] and he therefore suggested that sensitivity analysis must be performed.The higher the sensitivity value, the lower the sensitivity.In general, if the sensitivity value is greater than or equal to 2, the sensitivity of the study is low and the results are reliable.In this study a sensitivity analysis was carried out after matching using the Rbounds module in Stata and a sensitivity value of 2.7 was calculated.For this reason, we can assume that the results are reliable.

Final results
[47] We estimated the ATT and standard errors using the results of the 3 matching methods.After PSM, adolescent students who participated in extracurricular tutoring (treatment group) had distinctly higher levels of cognitive ability than those who did not.In all 3 matching methods, there were apparent improvements in students' word test levels by 1.221, 1.159, and 1.171, and math test levels by 0.767, 0.790, and 0.659 (Table 5).

Discussion
This study aims to use Propensity Score Matching (PSM) analysis to investigate the evidence of the impact of extracurricular  tutoring on the cognition of adolescents.First, this study examines the impact of adolescents participation in extracurricular tutoring on their cognition by using data from the 2018 China Family Panel Studies (CFPS).Participation of adolescents in extracurricular tutoring exhibited higher cognitive abilities.These findings confirm the significant influence of extracurricular tutoring on the cognitive development of adolescents.This study provides more robust evidence to support the findings of many related studies on this topic.Tan et al [48] reported that private tutoring can positively impact academic performance and reduce the workload of mainstream teachers, but can also widen the gap between students and increase pressure to engage in tutoring services.Conversely, Bray et al [49] opposed supplementing school work with private tutoring due to it depriving students of their originality and initiative.Furthermore, previous studies have found negative correlations between the time spent participating in extracurricular activities, such as tutoring, interest classes, watching TV, surfing the Internet, playing games, and adolescent students cognitive ability. [50]he above findings may be due to overestimating the effect of covariates.However, this study achieved an ideal control group by exact matching, meeting the basic requirements for causal inference between the treatment and control groups.As a result, this study provides robust evidence that extracurricular tutoring can improve the cognitive performance of adolescent students.
This study utilizes a novel approach to further understand and validate the impact of extracurricular tutoring on cognitive development from a methodological standpoint.Observational studies based on survey data lack randomization, resulting in selection bias and insufficient evidence for causal inference.To address this issue, this study employed PSM as a convenient and intuitive method, which is increasingly popular in different social science fields.An appropriate set of confounding covariates can be used for exact matching.This study identified 19 covariates from the literature that influence adolescent students cognitive abilities and decisions to participate in extracurricular tutoring.Additionally, no significant differences were found between the matched control and treatment groups, demonstrating successful matching and minimized selection bias.Despite China's high level of compulsory education coverage, true educational equity has yet to be achieved, and extracurricular tutoring exacerbates this inequity.The government can implement measures to address this, such as improving school hardware facilities, strengthening teacher training, and enriching teaching methods, providing students with equitable access to quality and homogeneous educational resources during school hours, promoting adolescent students cognitive development.Based on the above study's results, this paper proposes 2 recommendations.First, the government should provide tutorial assistance to disadvantaged young students, promoting educational equity and healthy development.Second, the government should allocate education resources in a rational and effective manner, particularly by increasing investment in less developed areas, to achieve maximum educational equity.

Limitations
There were several limitations to this study that should be considered in interpreting the findings.First, while 19 covariates were used to control for potential confounding factors, it is possible that other important variables may play a role in adolescent students cognitive development and their participation in extracurricular tutoring.On the 1 hand, the use of secondary survey data may have limited the selection of variables included in the analysis.On the other hand, the diverse and complex factors that influence youth participation in extracurricular tutoring make it challenging to study them comprehensively.However, the 19 covariates selected were deemed satisfactory in covering most aspects affecting adolescent students participation and cognitive ability.Second, in addition to the PSM method, other statistical techniques can be used to address selection bias, such as regression adjustment and instrumental variables.Therefore, combining multiple methods for cross-validation will definitely improve the reliability and persuasiveness of the conclusions.Finally, it is important to note that the study utilized the most recent dataset available, which was the 2018 database, in order to provide current and relevant findings.However, this limitation also means that the study's conclusions may not be generalizable to other time periods.

Conclusion
Our research results indicate that, after using PSM matching to control confounding variables, the cognitive ability level of teenagers who participated in extracurricular tutoring was significantly higher than those who did not participate in extracurricular tutoring.These findings suggest that the government should provide tutoring support to disadvantaged adolescent students as part of their efforts to promote healthy and equitable education development.
To further achieve equity in education, educational resources should be allocated rationally and effectively, with a particular focus on increasing investment in under-developed areas.Such measures will facilitate educational equity and ultimately better promote cognitive development among adolescent students.

Figure 1 .
Figure 1.Common range of values for the kernel matching propensity score.

Figure 2 .
Figure 2. Plot of standardized deviations for each variable of kernel matching.

Figure 3 .
Figure 3. Common range of values for the radius matching propensity score.

Figure 4 .
Figure 4. Plot of standardized deviations for each variable of radius matching.

Table 2
Descriptive characteristics of the total sample and the pre-matching participated groups versus not Participated groups.
Continuous variables were analyzed by t test, and categorical variables were analyzed by chi-square test.SD = standard deviation.

Table 3
Logistic regression estimates of adolescent students participating in extracurricular tutoring (N = 2567).
* P < .05. † P < .01. ‡ P < .001.www.md-journal.comsamplegroups.Meanwhile, we conducted a comparison of the differences between the 2 sample groups, as depicted in Table2.After dealing with missing values and outliers, a total of 2567 participants were eligible for this study.The participants average age was 13.534 ± 2.460 years, and there were more male (1363) than female (1204) participants.Majority (87.03%) of the respondents were Han Chinese, and the average number of household members was 4.98 ± 1.76, with a net family income per capita of RMB 17.917 ± 25.081 thousand, and an education and training expense of RMB 8.761 ± 11.301 thousand in the past year.A total of 55.01% of the respondents were middle school students, and 77.95% were currently studying in rural areas.The average class size was 47.97 ± 15.68, and 32.92% of the respondents held a class cadre position.The average daily study time on weekends and weekdays was 3.80 ± 2.98 hours and 8.75 ± 2.88 hours, respectively, among the participants.Out of 2567 participants, a total of 564 took part in extracurricular tutoring.

Table 4
The t test of covariates between treatment and control group after matching.

Table 5
Average treatment effect on the treated in three matching methods