The Effect of Preschool Education on Non- cognitive Skills of Middle School Students: Em- pirical Study Based on CEPS Xin Gong, Zhenyi Li Central China Normal University, Wuhan, China Abstract. Preschool education has been being seen as a crucial factor of individual cognitive skills, but it is yet clear regarding its effect on chil- dren’s non-cognitive skills. Apart from cognitive skills, the “capabili- ties” in the human capital model should also include non-cognitive skills. Using the baseline data of the “China Education Panel Survey”, this study explored the influence of preschool education experience on the non-cognitive skills of middle School students by using the method of propensity score matching. The results showed that children’s chances of receiving preschool education has been affected by factors like the health status of the children before schooling, family economic status, and parents’ education level. More importantly, preschool education had a statistically significant positive impact on the students’ ability of curiosity, acceptance of new things and self-discipline, and showed a negative impact on negative emotionality. Best Evid Chin Edu 2019; 3(1):287-303. Doi: 10.15354/bece.19.ar1070 Keywords: Preschool Education; Middle School Students; Non-cognitive Skills; Propensity Score Matching Correspondence to: Xin Gong, College of Education, Central China Normal University, Wuhan, China, E-mail: gongxin@mail.ccnu.edu.cn. About the Author Zhenyi Li, College of Education, Central China Normal University, Wuhan, China, E-mail: lizhenyi0206@163.com. Conflict of Interests: None. Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 288 Research Background and Problem RESCHOOL education, is of great significance to children’s physical and men- tal health, habit formation and intellectual development, and has a foundational role in individual’s lifelong development. In recent years, more and more atten- tion has been paid to preschool education in China, and the gross enrollment rate of preschool education has increased from 35.9% in 2001 to 77.4% in 2016. The rapid progress of preschool education, on the one hand, benefits from the rapid growth of the social economy; on the other hand, the increasing realization of the economic and non- economic value of preschool education for children’s overall growth. James J. Heckman, a Nobel Laureate in economy, has found that early childhood development interven- tions before formal elementary school are the highest-reward form of human capital investment, which is worth more than later investments in school education and on-the- job training (Heckman, 2006), and the economic benefits of universal preschool educa- tion is far greater than the traditional economic plan (Qiu & Liu, 2011). In terms of the specific impact of preschool education on children’s development, it mainly focuses on the following two aspects: cognitive ability including in language, mathematics, and academic achievement; and non-cognitive ability such as sociability, self-confidence, motivation, etc. Under the traditional human capital model, the word “ability” is always equat- ed with cognitive ability, while the non-cognitive ability is usually ignored. However, in empirical studies, the cognitive ability from the controlled individuals only cannot ex- plain the whole variety of individual outcomes. Therefore, Bowles proposed that “abil- ity” in the human capital model should not be limited to “cognitive ability”, but should also include “non-cognitive ability” (Bowles & Osborne, 2001). In recent decades, sup- porters of the new human capital theory headed by Heckman divided the “ability” in human capital into two parts: cognitive and non-cognitive abilities, emphasizing the important role of non-cognitive ability that is independent from cognitive ability (Heckman, 2007). In studies over the past decade on non-cognitive ability, most of them have confirmed the important role of non-cognitive ability in individual’s academic perfor- mance and income (Eren & Ozbeklik, 2013; Heckman & Rubinstein, 2001; Heineck & Anger, 2010; Scick & Steckel, 2015). Li et al. (2017) found that non-cognitive ability has a significant positive impact on students’ academic performance. Le et al. (2017) used the data of the “Chinese Family Panel Survey” in 2012 and 2014, and according to “The Big Five-Factor Model of Personalities”, showed that the non-cognitive ability has a substantial effect on the income of laborers, and its contribution was even higher than that of the core variable concerned by traditional human capital- education years. Huang and Xie (2017) found an explanatory effect of the non-cognitive ability on labor income difference that is independent from cognitive ability. Compared with cognitive ability, the non-cognitive ability has a greater impact on workers’ income (Yu et al., 2017). The non-cognitive ability has also been found to improve individual’s happiness and was P Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 289 more conducive to a healthy and long life (Friedman et al., 2010; Kubzansky et al., 2009). In addition to the “instrumental” value of non-cognitive ability such as on an individual’s academic performance, income, and health, researchers start to realize its own value of the non-cognitive ability. For example, from an emotional perspective, if an individual rarely shows negative emotionality, then its immediate happiness will be relatively higher indicating the intrinsic value of the non-cognitive ability. In recent years, attention has been paid to the key competencies of students in China, and formu- lated the overall framework of “key competencies of Chinese students’ development”, emphasizing on the cultivation of comprehensive quality of students’ both cognitive and non-cognitive abilities. With the attention paid to the non-cognitive ability, more discussion was raised on the cultivation of non-cognitive ability and its potential influence. Studies showed that the formation of non-cognitive ability needs several periods within a long develop- ing period (Almlund et al., 2011), and the important period for the non-cognitive ability acquisition is the early childhood (Coneus & Laucht, 2014; Heckman, 2006). Moreover, abilities formed in different stages were closely related each other, and the development of one skill would help to acquire other skills during the next period (Borghans et al., 2008). Preschool stage is a sensitive period essential to cultivate children’s initiative, good personality, sense of responsibility, self-esteem, and self-control (Zhu, 2003; Shen, 1997), and to develop qualities such as curiosity, thirst for knowledge, imagination, and sociability as well (Pang et al., 2003). In particular, various game activities during the preschool period had an important impact on the cultivation of the non-intellectual abilities (Shen, 1997). Duncan et al. (2007) found in their empirical study that stable emotional characteristics cultivated in the early stage contribute to the development of children’s exploration ability later. A large number of empirical studies have confirmed that preschool education has a positive impact on the development of children’s cogni- tive and non-cognitive abilities (social skills, attention, etc.) (Nores & Barnett, 2010). The earlier the intervention would, the more beneficial would follow to the formation of both the cognitive and non-cognitive abilities. Besides, families and preschoolers might be more motivated to focus on the cognitive and non-cognitive development as it is too early to consider the competitive College Entrance Test during the preschool period. Based on the analysis and consideration above regarding the education reality in China, we proposed that the acquisition of preschool education may possess an im- portant role in the development of children’s non-cognitive abilities. However, it is yet known about the clear relationship between preschool education and the non-cognitive ability. In view of this, this study was designed to use the baseline data of China Educa- tion Panel Survey (CEPS) to observe the influence of preschool education on the devel- opment of the non-cognitive abilities in middle school students’ via Propensity Score Matching Method and, at the same time, to investigate the factors that affect the pre- school education opportunities. As the prominent role of non-cognitive ability is in- creasingly recognized, it is particularly pivotal to explore the role of preschool educa- tion in developing the non-cognitive ability. First, it is important for public education policy makers in establishing capital investment, implementation and improvement of Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 290 the preschool education plan, popularization of preschool education, and promotion of human capital accumulation; second, it is crucial to provide with empirical evidence to the role of preschool education in the development of students’ non-cognitive ability. Literature Review Accumulating data have showed the contributing effects of preschool education on both cognitive and non-cognitive abilities. At the same time, some pieces of evidence have confirmed the significant role of preschool education in the improvement of children’s non-cognitive ability. Reynolds (1995) investigated the impact of the Child-Parent Cen- ter Educational Program on children’s cognitive and social development, and found that students in grades one to six who had received one or two years of preschool education had 0.34 standard difference higher in cognitive and social development than those who had not received the preschool education. Adaptability is another important characteris- tic for those who receive preschool education. Sammons (2010) found that preschool education has a long-term impact on children’s “hyperactivity”, “self-discipline” and “anti-social” behaviors. Furthermore, Sammons et al. (2014) used data from the UK and found that high-quality preschools (v.s. low-quality ones) had a significant impact on the non-cognitive ability (such as self-discipline) of 16-year-olds. Andersson et al. (1992) obtained similar results in a study carried out in Sweden that students who par- ticipated in the public childcare system were more independent and confident. Gormely et al. (2011) used the fixed effects regressions with propensity score matching examined the effect of the early childhood education on children’s social-emotional outcomes in Tulsa, Oklahoma, and found that children participated in early childhood education programs showed lower timidity and higher concentration, and this effect was more significant on male children who came from the low-income families. Heckman (2006) indicated that the reason why the Perry Preschool Program in the United States could effectively improve the future success of participants was majorly due to its role in im- proving the non-cognitive ability. Chetty et al. (2010) found that high-quality education in K grade in the United States, had a significant impact on students’ non-cognitive abilities such as autonomy and effort in the grades of 4 and 8, and could reduce students’ destructive behaviors. However, in China, few studies did the same observation. Several studies fo- cused on the impact of preschool education on children’s cognitive development, but most of them were small sample and did not examine the long-term impact. Peng et al. (2011) investigated 182 children from three rural elementary schools in Sichuan Prov- ince and found that children with kindergarten learning were more likely to be praised by their peers and less likely to be downgraded than the comparisons. Rao et al. (2012) observed 370 rural children in Guizhou Province and showed that children who had attended kindergarten or preschool classes were significantly better than those who had not in reading and mathematics. Liu et al. (2013) studied 1,360 children in Beijing and Shanxi Province and found that one-year preschool education had a significant effect on the mathematical learning and language development, reaching 0.61 and 0.33 standard deviations, respectively. Li et al. (2016) investigated the influence of preschool educa- Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 291 tion on children’s language, mathematics, and social cognition ability and found a sig- nificant relation between them. Chen and Liu (2017) used a sample of 5,177 students from the Program of International Student Assessment (PISA) Shanghai 2012 data and Propensity Score Matching and OLS estimation methods, they found that preschool education had a significant positive impact on the mathematics, reading and scientific literacy for 15-year-old students, and the academic achievements in those who received more than one-year preschool education were higher than those who received only one- year preschool education. Gong et al. (2016) used assessed the long-term impact of pre- school education on the development of rural children with the age over 10 on the basis of the “Chinese Family Panel Survey” (CFPS) data through fixed-effects model, OLS estimation method, and Propensity Score Matching, and found that preschool education had a substantial impact on the social and emotional skills, which were displayed as preschool attendees in a rural area tended to have 1.32–1.53 more friends and an 11%- 17% higher chance to be a school leader than the non-attendees. The above-mentioned studies focused on the influence of preschool education on students’ cognitive abilities such as mathematics, language, and reading, but rarely explored its effect on the non-cognitive ability. However, the impact of preschool edu- cation on children’s development is a multi-dimensional, multi-level and multi-faceted complex system, and researchers and decision-makers should pay more attention to its impact on children’s non-cognitive development. This study evaluated the effect of pre- school education from a new perspective and empirically tested the influence of pre- school education on the development of the non-cognitive ability of middle school stu- dents with a nationwide sample. Data This study used data from the China Education Panel Study (CEPS) baseline survey for the 2013-2014 academic year. The survey was designed and implemented by the Na- tional Survey Research Center at Renmin University of China. The middle school (7th graders) and high school (9th graders) students were the objects, and multi-stage proba- bility proportional to size (PPS) sampling method was used. After four sampling stages, the county (district), school, class, student/parent/teacher/main subject teacher/school leadership and other four sampling units were extracted. In the baseline survey, a total of 28 counties, 4 schools from each county, and 4 classes from each school were select- ed, and a total of 19,487 seventh- and ninth-graders from 112 schools and 438 classes were sampled including 18,034 sample booklets with complete information on the pre- school experience. Variables Outcome Variables The outcome variable of this study was the non-cognitive ability. At present, there is no unified standard for the definition and measurement of non-cognitive ability. It is gen- erally believed that the non-cognitive ability is not a representative of a single ability, Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 292 but a combination of multiple abilities. In western, the concept of “non-cognitive skills” was defined as those that are different from calculation, reading or reciting and can be measured by personality traits. For example, curiosity, attention, grit, self-control, so- cial-emotional ability, self-discipline, motivation, self-esteem, sociability, time prefer- ence and so on (Borghans et al, 2008; Duckworth & Yeager, 2015; Gutman & Schoon, 2013). In China, Yan (1988) divided “non-intelligence factors” into three levels: (1) in the broad sense, they refer to all psychological factors other than intelligence factors; (2) in the narrow sense, they mainly include motivation, interest, emotion, intention, char- acter and other psychological factors; (3) specific non-intellectual factors, including achievement motivation, thirst for knowledge, and enthusiasm for learning; self-esteem, self-confidence, and competitiveness; sense of responsibility, obligation and honor; Self-control, tenacity and independence. In the past two decades, the “big five personal- ities” measurement method developed by personality psychology has gradually been recognized and applied to measure the non-cognitive abilities that include extraversion, openness, emotional stability, agreeableness, and conscientiousness. Since no uniform measurement index exists, empirical researchers often reconstruct the proxy index of the non-cognitive ability according to specific questionnaire index and the characteris- tics of the “big five personalities”. Openness According to the four related questions in the questionnaire, they can express opinions clearly, respond quickly, learn new knowledge swiftly, and be curiosity about new things (1 = completely disagree to 4 = completely agree). Through factor analysis, this study found that these four items naturally stayed together to form one factor value (α = 0.72). Combined with the “big five grid” model, we define this factor value as “open- ness”, and set it up as a measure of non-cognitive ability to keep consistent with the method used by Li et al. (2017). Negative Emotionality The CEPS questionnaire asked students if they felt dispirited, depressed, unhappy, list- less, or sad in the past seven days, using a five-point scoring system (1 = never to 5 = always) to measure the frequency of these feelings. Factor analysis showed that these five measures had a high similarity and could be naturally aggregated into one factor value (α = 0.85). Combined with the negative emotionality in the temperament type of children, we define this factor value as “negative emotionality”. The higher the factor score is, the stronger the negative emotionality would be. Self-Discipline According to the CEPS questionnaire, whether students often skip classes or late, a fac- tor value (α = 0.69) was obtained by factor analysis. Referring to the research idea of Cawley et al. (2001) that took truancy, lateness and other behaviors as proxy variables of “self-discipline”, our study reversed scores of the above-mentioned factor value (1 = Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 293 completely agree to 4 = completely disagree) and defined it as “self-discipline”. The higher the factor score is, the stronger the self-discipline would be. Sociality We took the number of students’ self-reported good friends as a social indicator. In pre- vious studies, the number of friends has been used as a proxy indicator of social com- munication skills. For example, Gong et al. (2016) used “the number of good friends they have” and “whether they are class cadres” as proxy indicators of non-cognitive social skills. Some western studies also took the number of good friends as the proxy variable of social communication (Kingery et al., 2011; Pettit et al., 2011). In addition, studies have shown that people who have more friends as childhood had better social and emotional well-being in adolescence and adulthood. Kingery et al. (2011) found that students who had more friends showed reduced loneliness and increased self- esteem. Wentzel et al. (2004) found a positive correlation between the number of friends and various adaptive indicators. Above-mentioned studies indicated that the larger the number of friends, the stronger the transitional adaptability from one stage of development to the next. In addition, our study added a reflection of the quality of friends that was composed of friends’ academic performance, effort, educational expec- tations and discipline. 1 Independent Variables The key independent variable of our study is preschool education experience. The CEPS questionnaire asks a student whether he or she has attended preschool or a one year preschool class before grade 1 after the age of 3, and we construct a binary variable (yes = 1, no = 0) based on the student’s answer whether he or she has received pre- school education (79.8%). 2 Control Variables On the basis of the literature reviewed above, this study controlled for the other factors that may affect both preschool education acquisition and the non-cognitive ability of middle school students, including:  At the personal level, gender (female = 1, male = 0) and grade are used (nine grade = 1, grade 7 = 0), nationality (Han = 1, minority = 0), household registra- tion (agricultural as the reference), one-child (one-child = 1, non-only child = 0), birth weight (low birth weight or birth weight less than 2.5 kg = 1, normal birth weight = 0), illness before elementary school (severe illness = 1, not se- vere = 0), and other variables.  At the family level, two variables are included: the years of education of the parent; the subjective evaluation of the family economic status before the chil- dren enters elementary school, including “family economic status is medium” and “family economic status is rich” two binary variables, taking “family eco- nomic status is difficult” as the reference. 3 Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 294  At the regional level. A region of the average level of teaching education to a certain extent reflects the local social and economic development. To observe the influence of regional social, economic and cultural level on preschool edu- cation and the development of the cognitive ability, we selected students in the county level with average education years as a regional social, economic and cultural development level proxy variable. 4 Method First, we used descriptive statistical methods to analyze whether there were differences between students with or without preschool education in terms of individual characteris- tics and family background, so as to understand whether the basic characteristics of both groups matched. On this basis, the logit method was used to investigate the factors that affect the preschool attending. Second, in order to accurately evaluate the influence of preschool education on children’s non-cognitive ability, we adopted the OLS estimation and the Propensity Score Matching (PSM) methods. Conventional OLS estimation roughly assumes that the observed value of the treatment group is the influence after receiving preschool edu- cation, which will lead to biases of the estimation results. It is not a natural process if receive preschool education or not or how long preschool education lasts, but generally is influenced by factors such as students’ individual characteristics and family socioec- onomic background. So it is impossible to observe the outcomes of the same individual in the two states of both preschool and non-preschool education at the same time. Selec- tive bias is a problem we need to take care of in order to accurately estimate the effect of preschool education on the non-cognitive ability. Compared with OLS, PSM can match the background of both groups more effectively and reduce the selective bias. Therefore, our study was based on the estimation of PSM. PSM was first proposed by Rosenbaum and Rubin, and has been increasingly recognized and widely used in recent years. The basic feature of PSM is its good match of the control group with the treatment group to balance the data of the probability (propensity score). It degrades the multidimensional standard of both groups into one dimension to greatly improve the matching degree and guarantee the comparability of both groups, and to get a better estimation of the intervention effect of the treatment group. The net effect of preschool education on the non-cognitive ability of middle school students was recorded as Average Treatment Effect on the Treatment group (ATT) in the PSM model, and the model was as follows, ATT = E {E [Y1i - Y0i | Di = 1, p (Xi)]} where Y1i and Y0i represent the scores of the non-cognitive ability of individual student in preschool and non-preschool education, respectively. Di is the processing variable, which represents the dummy variable of whether students receive preschool education. If student i receives preschool education, Di = 1; otherwise, Di = 0. P(Xi) is the propen- sity score value, which represents the conditional probability of student i receiving pre- Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 295 school education under the control of sample characteristic covariate X. PSM is gener- ally divided into four steps: (1) select appropriate covariates. Based on the previous research experience, we selected a group of variables that have an influence on both preschool education participation and children’s non-cognitive abilities (see Control Variables above); (2), calculate propensity scores. The propensity score was calculated by logit regression; (3), match samples according to propensity scores. In this study, the nearest neighbor matching method is used, and the radius matching and kernel matching methods were used to test the robustness of the estimated results; (4) by comparing the estimation results of various matching methods, the net effect of preschool education on the non-cognitive ability of middle school students is obtained. Results Descriptive Picture Table 1 shows significant differences between preschool attendees and non-attendees in terms of individual and family background. Specifically, among the samples receiving preschool education, non-agricultural households account for 43.5% of the total sample, which was significantly higher than the proportion of non-rural households in the group without preschool education (30.2%). In addition, compared with non-preschool at- tendees, the proportions of individuals with low birth weight and serious illness before elementary school were much lower in students who received preschool education, and the proportion of only children was higher. From the point of family economic back- ground, the parents of preschool attendees have more years of education, with a differ- ence of 1.27 years than the non-attendees, and the proportion of moderate and rich family in the preschool attendees was significantly higher than the comparisons. Finally, there was a significant difference in the average length of education (~0.68 yrs). In sum, obvious difference exists in individual characteristics and family background between preschool attendees and non-attendees. Factors Affecting Individual’s Acceptance of Preschool Education On the basis of above-mentioned descriptive analysis, we found significant differences between the two groups in terms of individual characteristics and family background. In order to observe the factors that influence individual acceptance of preschool education, we used logit regression analysis method to analyze the specific factors affecting indi- vidual acceptance of preschool education, and the estimation results are shown in Table 2. Since most independent variables in the model are discrete variables, the table pre- sents both odds ratio and the original non-standardized coefficient of the model. Table 2 shows that factors such as gender, ethnic group, grade, individual health status, family economic status and years of education of parents all had signifi- cant influences on individual’s acceptance of preschool education. Specifically, only- child families were more likely (22.9%) to receive preschool education than non-only children. Seventh-graders were 12%more likely than ninth-graders to receive preschool education, which may be related to the development of preschool education. Students in Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 296 Table 1. Sample Characteristics. Variables Overall Sample Preschool Attendees Preschool Non-attendees t-test Preschool Attendance 0.798 1 0 - Female 0.485 0.490 0.470 0.020* Han Ethnicity 0.910 0.924 0.857 0.067*** Grade 9 0.473 0.464 0.505 -0.041*** Non-Agricultural Household 0.409 0.435 0.302 0.133*** Low Birth Weight 0.322 0.295 0.433 -0.138*** Had Major Illness before Enter- ing Elementary School 0.091 0.083 0.126 -0.043*** The Only Child 0.434 0.468 0.300 0.168*** Parent’s Education 10.767 11.082 9.793 1.273*** Family Economy Medium 0.731 0.749 0.652 0.097*** Family Economic Affluence 0.056 0.060 0.041 0.019*** Average Number of Years of Education in the County 9.485 9.481 8.938 0.543*** Note: *p<0.05, **p<0.01, ***p<0.001. The value in the table is the mean value of each varia- ble. Since most of the result variables are standardized factor score values, the mean value of the result variables is not shown here. poor physical condition before elementary school were significantly less likely (25.7%) to receive preschool education than those in good physical condition. In comparison, students came from families with economic difficulties had a 43.3% of preschool edu- cation acceptance, but those from family with good economic status had a 63.1% of the preschool education acceptance, this is mainly because of poor families are unable to afford the high cost of preschool education. In addition, the level of local education lev- el had a statistically significant impact on students’ access to preschool education. Therefore, the analysis confirmed that students in a better position were more likely to receive preschool education. Moreover, the family’s economic status is the most signif- icant factor influencing students to receive preschool education. Influence of Preschool Education on Non-Cognitive Ability of Middle School Stu- dents We first used the OLS regression method to estimate whether preschool education had a significant impact on students’ non-cognitive ability after controlling other confounding factors. Then, the PSM method was used for further analysis. The results are shown in Table 3. The PSM results showed that a stable and significant relationship exists be- tween preschool education and students’ non-cognitive abilities such as thinking open- ness, negative emotionality, and self-discipline and friend quality. After controlling the Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 297 Table 2. Influencing Factors of Preschool Education: Logit Re- gression Results. Variables Preschool Attendance Coef Odds Ratio Female 0.082* 1.086 (0.039) Grade 9 -0.128*** 0.880 (0.039) Han Ethnicity 0.195** 1.217 (0.065) Non-Agricultural Household 0.033 1.034 (0.469) Low Birth Weight -0.291*** 0.743 (0.041) Had Major Illness before Entering Elementary School -0.131* 0.872 (0.062) The Only Child 0.206*** 1.229 (0.046) Parent’s Education 0.064*** 1.070 (0.012) Family Economy Medium 0.355*** 1.433 (0.043) Family Economic Affluence 0.487*** 1.631 (0.112) Average Number of Years of Education in the County 0.198*** 1.219 (0.017) N 18,034 Note: *p<0.05, **p<0.01, ***p<0.001. personal and family factors, the thinking openness in students who had received pre- school education were 0.04-0.05 standard deviations higher than those who had not re- ceived preschool education, that is, students who have received preschool education were better than those who had not received preschool education in terms of curiosity and ability to accept new things. Regarding the negative emotionality, students who have received preschool education showed less negative emotions such as depression, sadness and unhappiness indicating that in terms of emotional regulation, students who had received preschool education were stronger than those who had not received pre- school education. In addition, on self-discipline, students who had received preschool education rarely ran away from school and late when compared with those who had not received preschool education. Finally, in terms of the quality of friends, students who had received preschool education showed better performance. Among the friends they mentioned, most were students who learnt hard, achieved outstanding results and per- Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 298 Table 3. Influence of Preschool Education on Non-cognitive Abil- ity of Students. Outcome Variable OLS PSM Nearest Neighborhood Matching Radius Matching Kernel Matching Openness 0.036** 0.042** 0.036** 0.045*** (0.015) (0.019) (0.017) (0.017) Negative Emotion -0.071*** -0.071*** -0.073*** -0.071*** (0.017) (0.022) (0.020) (0.019) Self-Discipline 0.028* 0.011 0.032* 0.034* (0.014) (0.019) (0.017) (0.017) Number of Good Friends 0.547 + 0.278 0.149 0.198 (0.279) (0.344) (0.307) (0.301) Quality of Friends 0.151*** 0.176*** 0.197*** 0.206*** (0.037) (0.052) (0.047) (0.046) Note: + p<0.1, *p<0.05, **p<0.01, ***p<0.001. Other control variables are included in both OLS and PSM estimates. In PSM estimation, the nearest neighbor matching (K=4), radius matching (caliper= 0.01), kernel matching and other matching methods were adopted. The standardized mean differences of almost all variables after matching were less than 10%, and the standardized mean differences of most variables were significantly reduced. ATT was reported in the table. formed well. In addition to preschool education, parents’ education level and the status of family economy were also important factors affecting these non-cognitive abilities. 5 Chinese scholars have analyzed the important role of preschool education in cultivating children’s curiosity from the theoretical point, but no corresponding empiri- cal study was conducted. Our study demonstrated that preschool education has a signif- icant positive effect on the development of students’ curiosity that provides an empiri- cal evidence for the theoretical analysis mentioned above. In addition, this study found that preschool education has a significant positive impact on the self-discipline of mid- dle school students, which is consistent with the conclusion of Sammons (2010). Dif- ferent from Gong et al. (2016), they found that preschool experience had no significant effect on the number of students’ friends, which may be related to the differences in the samples used in the two studies, Gong et al took rural students as the research objects, whereas this study covered students from both urban and rural areas, and this variable was given priority, and therefore may exist a certain difference. Although the conclu- sions were inconsistent in terms of the number of friends, our study found that the stu- dents who received preschool education were more likely to have good academic per- formance, learn hard and perform well with their best friends, which supplemented the study of Gong et al. These two studies verified the significant influence of preschool Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 299 education on the development of the non-cognitive ability of elementary and middle school students from different aspects. Matching Effect Analysis The propensity score matching method was used to satisfy two basic assumptions: (1) equalization of covariate distribution in the post-match treatment and control groups; (2) the propensity scores of both treatment and control groups with common support areas. Our study focused on the balance of variables that also affect the preschool enrollment and the non-cognitive abilities of children. The balanced variables include parents’ edu- cational attainment, family economic status before elementary school, household regis- tration, and only-child family. The balance test found that these variables matched well and standardized mean differences were less than 5%. Meanwhile, the standard mean difference of the other control variables was less than 10%. According to the distribu- tion map of propensity scores between the treatment and the control groups, there were overlapping areas of propensity scores between the preschool attendees and non- attendees indicating that the matching reduced the initial difference of characteristic variables between both groups. Conclusions and Suggestions Given the prominent role of the non-cognitive ability in personal growth and develop- ment, it is increasingly recognized that “ability” in the human capital model includes not only cognitive ability, but also the non-cognitive ability. It has become the latest trend of world education practice to try new teaching methods to develop children’s non-cognitive abilities in social emotion, curiosity, and thirst for knowledge, self- discipline, and initiative. In this context, it is particularly important to explore the role of preschool education in the development of children’s non-cognitive ability. Based on the national sample of the 2013-2014 baseline data of the CEPS, this study empirically investigated the effect of preschool education on the non-cognitive ability of middle school students in China using the propensity score matching method, and attempted to answer the following questions: What factors affect the access to preschool education? Does preschool education affect the non-cognitive ability of the middle school students? The main research conclusions are: First, there is inequality in access to preschool education. Our study found that students’ physical condition before elementary school, family economic condition, par- ents’ years of education and local economic level all possess role in affecting students’ access to preschool education. In comparison with students from poor families, students from middle and wealthy families are 43.3% and 63.1% more likely to receive pre- school education. Second, preschool education has a significant impact on the non-cognitive abil- ity of middle school students. We found that students who have received preschool edu- cation had a strong curiosity and ability to accept new things, less negative emotions, better self-discipline, and less truancy and lateness. In addition, in terms of the quality of friends, students with preschool education had more good friends who are more like- Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 300 ly to get good academic and other school performance. Therefore, preschool education can improve the non-cognitive ability of human capital and play a positive role in pro- moting individual’s future development. Based on these findings, we strongly suggest that the government should con- tinue to pay attention to and invest in the preschool education. Meanwhile, because children from poor families with poor health are significantly less likely to receive pre- school education, so the government’s financial expenditure on preschool education should incline to vulnerable children to guarantee the right of the disadvantaged chil- dren to receive preschool education. We need to point out the limitations that our study has. First, the two methods we used in this study are non-experimental ones. Although the PSM method can reduce the selective bias to some extent, it only controls the influence of observable variables. If there is an unobservable endogenous variable, so “hidden bias” cannot be avoided. Second, we observed a number of non-cognitive indicators such as self-discipline, nega- tive emotionality, and thinking openness, but most of these indicators are self-reported measures. So they were more subjective. With the development of measurement tech- nology, more objective indicators can be investigated in the future. Finally, future stud- ies can observe the impact of preschool education with different types and quality on the development of children’s non-cognitive ability. Note: 1. The CEPS questionnaire asks the following questions: “do some of the good friends mentioned above have the following situations: good academic performance, hard study, and want to go to college”, then the answer is “no such” (1) “one or two” (2) and “many such” (3) Another related question is: “good friend mentioned above have the following situation: skipping classes, truant, kip, in violation of the rules being crit- icized, disposition and fight back and learned”, because the behavior such as skipping classes and grades excellent performance, such as is on the contrary, to facilitate re- search, do the reverse score here, answer “no such assignment 3,” answer “one or two” assignment 2, answer “many of these” assignment 1. On this basis, the index of “friend quality” (a = 0.80) was synthesized by principal component analysis. 2. The preschool education here is mainly for children aged 3-6, but not limited to this. In the CEPS survey, no matter how old a child is enrolled in kindergarten or pre-school, or how many preschools he or she has received, the figure is better than the three-year gross enrollment rate reported by the national statistics. 3. During the survey, the hospital asked parents and students for subjective evaluation of the economic status of the family before elementary school. We give priority to the par- ents’ questionnaire data, and then use the students’ questionnaire data when the par- ents do not answer, and finally generate derived variables that reflect the family eco- nomic status of students before they enter elementary school. Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 301 4. Based on the data of the sixth national census, the CEPS project works out the average years of schooling in the samples’ districts and counties. 5. Limited by space, other control variables from the results of omission. References Almlund, M., Duckworth, A. L., Heckman, J. J., & Kautz, T. D. (2011). Personality Psychol- ogy and Economics. In Hanushek, E.,Machin, S.,&Woessmann L(Ed.), Hand- book of Economics of Education(pp.117- 150).North-Holland Press. Andersson, B.E. (1992). Effects of Day-Care on Cognitive and Socioemotional Competence of Thirteen-Year-Old Swedish Schoolchil- dren. Child Develop, 63(1): 20-36. Borghans, L., Duckworth, A. L., Heckman, J. J., & Baster, W. (2008). The Economics and Psychology of Personality Traits. J Human Resource, 43(4): 972-1059. Bowles, S., & Osborne, G. M. (2001). The de- terminants of earnings: a behavioral ap- proach. Journal of Economic Litera- ture, 39(4): 1137-1176. Cawley, J., Heckman, J. J., & Vytlacil, E. (2001). Three Observations on Wages and Measured Cognitive Ability. Lab Econ, 8(4): 419-442. Chen C J & Liu Q . (2017) .A Quasi-experiment Study of Impact of Early Childhood Educa- tion on Students’ Academic Achievements When They Were 15 Years Old Based on the Shanghai Data from PISA. Stud Early Child Edu, (1): 3 -12.(in Chinese) Chetty, R., Friedman, J. N., Hilger, N.,Saez, E., Schanzenbach,D.W.,& Yagan,D. (2011). How Does Your Kindergarten Classroom Affect Your Earnings? Evidence from Pro- ject STAR. Q J Econ, 126(4): 1593. Coneus, K., & Laucht, M. (2014). The Effect of Early Noncognitive Skills on Social Out- comes in Adolescence. Edu Econ, 22(2): 112-140. Duckworth, A. L., & Yeager, D. S. (2015). Measurement Matters: Assessing Personal Qualities Other Than Cognitive Ability for Educational Purposes. Edu Res, 44(4): 237- 251. Duncan, G. J., Dowsett, C. J., Claessens, A., Magnuson, K., Huston,A. C., &Klebanov, p.,et al.(2007). School Readiness and Later Achievement. Deve Psy, 43(6): 1428-1446. Eren, O., & Ozbeklik, S. (2013). The Effect of Noncognitive Ability on the Earnings of Young Men: A Distributional Analysis with Measurement Error Correction. Lab Econ, 24(6): 293-304. Friedman, H. S., Kern, M. L., & Reynolds, C. A. (2010). Personality and Health, Subjective Well-being, and Longevity. J Personality, 78(1): 179-216. Gong, X., Xu, D., & Han, W. J. (2016). The Effects of Preschool Attendance on Adoles- cent Outcomes in Rural China. Early Child Res Q, 37: 140-152. Gormley, W. T., & Adelstein, S. (2011). Social- Emotional Effects of Early Childhood Edu- cation Programs in Tulsa. Child Develop, 82(6): 2095-2109. Gutman L M, Schoon I. The Impact of Non- cognitive Skills on Outcomes for Young People: Literature Re- view[EB/OL].https://v1.education endow- ment foundation. org. Uk/uploads/pdf/Non- cognitive _skills_literature_review_2.pdf Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 302 Heckman, J. J. (2006). Skill Formation and the Economics of Investing in Disadvantaged Children. Science, 312(5782): 1900-1902. Heckman, J. J. (2007). The Economics, Tech- nology, and Neuroscience of Human Capa- bility Formation. Pro National Academy Sci 104( 33):13250-13255 Heckman, J. J., & Rubinstein, Y. (2001). The Importance of Noncognitive Skills: Lessons from the GED Testing Program. Am Econ Rev, 91(2): 145-149. Heineck, G., & Anger, S. (2010). The Returns to Cognitive Abilities and Personality Traits in Germany. Lab Econ, 17(3): 535-546. Kingery, J. N., Erdley, C. A., & Marshall, K. C. (2011). Peer Acceptance and Friendship as Predictors of Early Adolescents’ Adjustment Across the Middle School Transition. Mer- rill-Palmer Quarterly, 57(3): 215-243. Kubzansky, L. D., Martin, L. T., & Buka, S. L. (2009). Early Manifestations of Personality and Adult Health: A Life Course Perspective. Health Psy, 28(3): 364-372. Le, J. J., & Hu B. W. (2017). The Effects of Non-cognitive Abilities on Wages: Evidence from China Family Panel Studies. Chin J Pop Sci ,(4): 66-76.(in Chinese) Li, K., Yi, P., Hu, B.,Burchinal, M., Marco, A. D.,&Fan, X.,Qin,J. L. (2016). Early Child- hood Education Quality and Child Outcomes in China: Evidence from Zhejiang Province. Early Child Res Q 36(Supplement C): 427- 438. Li, L., Zhao W. L., & Bian W. J. (2017). Family Background and the Non-cognitive Skills. Res Edu Develop, (1): 45 -52.(in Chinese) Liu, Y., Zhao J. H., & Zhang L. (2013). Value Added Analysis of Teaching Effectiveness of One-Year Preschool Education. J Edu Stud ,(3): 56 -66. Nores, M., & Barnett, W. S. (2010). Benefits of Early Childhood Interventions Across the World: (Under) Investing in the Very Young. Econ Edu Rev, 29(2): 271-282. Pang, L. J., Hu J., & Hong ,X. M. (2003). Value of Preschool Education. Studies Early Child Edu ,(1): 7 -10.(in Chinese) Peng, J. Y., Gao Q., & Xiao J. (2011). The Out- come of Kindergarten Education in the Western Rural Area. Studies Early Child Edu, (7): 9-13.(in Chinese) Qiu, Z. H., &Liu Y. (2011). The Research on the Economic Development Benefits of Pre- school Education of the Foreign Countries. Interl Com Edu,(6): 1-4. (in Chinese) Rao, N., Sun, J., Zhou, J., & Zhang, L. (2012). Early Achievement in Rural China: The Role of Preschool Experience. Early Child Res Q, 27(1): 66-76. Reynolds, A. J., Temple, J. A., Ou, S. R., Arteaga, I. A.,&White,B .A. B. (2011). School-Based Early Childhood Education and Age-28 Well-Being: Effects by Timing, Dosage, and Subgroups. Science, 333(6040): 360-364. Sammons, P. (2010). Do the Benefits of Pre- school Last? Investigating Pupil Outcomes to the End for Key Stage 2 (aged 11).In Syl- va K, Melhuish E, Sammons P. Siraj- Blatchford I, and Taggart B. (Eds.). Early Childhood Matters: Evidence from the Ef- fective Pre-school and Primary Education project(pp.114-148).Routledge, New York. Sammons, P., Sylva, K., Melhuish, E.,Siraj, I. (2014). Influences on Students’ Social- behavioural Development at age 16 Effec- tive Pre-School, Primary & Secondary Edu- cation Project. (EPPSE).[EB/OL].https://www.researchgate. net/publication/266372995_Influences_on_s tudents%27_social- behaviour- al_development_at_age_16_Effective_Pre- School_Primary_Secondary_Education_Proj ect_EPPSE Schick, A., & Steckel, R. H. (2015). Height, Human Capital, and Earnings: The Contribu- tions of Cognitive and Non-cognitive Ability. J Human Capital, 9(1): 94-115. Shen, D., L. (1997). Theory and Practice of Non-intellectual Factors. Edu Sci Press. (in Chinese) Tam, T., & Wu, H. F. (2013). Noncognitive Traits as Fundamental Causes of Health Ine- quality. Chin Soc Rev, 46(2): 32-62. Wentzel, K., Barry, C., & A. Caldwell, K. (2004). Friendships in Middle School: Influ- ences on Motivation and School Adjust- ment .Edu Psy,96(2):195-203. Yan, G., C. (1988). A Few Questions about Non-Intellectual Factors. J Shanghai Normal Gong & Li. Preschool and Non-cognitive Skills of Middle School Students Vol.3, No. 1, 2019 303 Univ(Philosophy & Social Science Edi- tion),(4): 134-139. Yu, F., Wang, C., Shen, J.,Shi, Y.,& Li, T. (2017). Effect of Cognitive Abilities and Non-cognitive Abilities on Labor Wages: Empirical Evidence from The Chinese Em- ployer-Employee Survey. China Econ J, 10(1): 76-89. Zhu, Z., X.(2003). Child Psychology. People’s Edu Press. (in Chinese). Received: 22 July 2019 Revised: 10 August 2019 Accepted: 15 September 2019 The Chinese version of this article has been published in Edu Econom 2018; 34(4):37-45. The English version has been authorized for being publication in BECE by the author(s) and the Chinese journal. 龚欣, 李贞义. 学前教育经历对初中生非认知能力的影响: 基于 CEPS的实证研究. 教育与 经济, 2018; 34(4): 37-45. Article-Title-BECE_V3N1_30Sep2019-ZYL Article-Maintext_BECE_V3N1_30Sep2019-ZYL