You are on page 1of 15

China Economic Review 62 (2020) 101477

Contents lists available at ScienceDirect

China Economic Review


journal homepage: www.elsevier.com/locate/chieco

Parenting style and the development of noncognitive ability in


T
children
Lanfang Denga, Tingting Tongb,

a
School of Economics & Management, South China Normal University, 378 Waihuanxi Road, Guangzhou 510006, China
b
International Business College and Institute of Supply Chain Analytics, Dongbei University of Finance and Economics, 217 Jianshan Street, Dalian
116025, China

ARTICLE INFO ABSTRACT

Keywords: We investigate the impact of parenting style on children's noncognitive ability development using
Parenting style data from China. Based on an extensive set of questions of caregiver-child interaction, we dis-
Noncognitive ability tinguish parenting style (i.e., respectful and disciplinary) from traditional material and time
Parental investment parental investment. We find that respectful parenting style significantly influences the formation
Endogeneity
of children's noncognitive ability. Our results suggest that with detailed measures of parenting
style and parental investment, children taken care by parents and grandparents do not exhibit
JEL codes:
significant differences in noncognitive ability. A further investigation of unobserved hetero-
D13
J13 geneity based on different techniques reveals a causal and long-lasting relationship from par-
J24 enting style to noncognitive ability.

1. Introduction

There is growing evidence that noncognitive ability such as perseverance, motivation, self-esteem is critical for individuals' future
outcomes, including wage (Eren & Ozbeklik, 2013; Kuhn & Weinberger, 2005; Lindqvist & Vestman, 2011), occupation level
(Grönqvist & Lindqvist, 2016), and school achievement (Heckman & Rubinstein, 2001). Dela。ys in noncognitive ability development
have been shown to negatively affect individuals' future outcomes, such as wage and productivity in the labor force (Heckman,
Stixrud, & Urzua, 2006). In addition, previous studies (e.g., Cunha, Heckman, Lochner, & Masterov, 2006; Cunha, Heckman, &
Schennach, 2010) have demonstrated that compared to cognitive ability, noncognitive ability has a higher level of malleability,
which can be improved in later stages of childhood. Therefore, the importance and malleability of noncognitive ability has prompted
researchers to investigate the determinants of children's noncognitive ability, especially the role of family, which has been considered
to have a powerful impact on children's skill formation (Cunha & Heckman, 2008).
The influences of family on the children's noncognitive development are very complex (Cunha & Heckman, 2008). In the eco-
nomics literature, there have been growing studies estimating cognitive/noncognitive skill production function, see for example,
Todd and Wolpin (2003, 2007), and Cunha and Heckman (2007, 2008, 2009). Most studies focus on the role of a specific aspect of
family characteristics, such as household income (Blau, 1999; Loken, Mogstad, & Wiswall, 2012), mother's time investment (Bernal &
Keane, 2011; James-Burdumy, 2005), and birth order of the child (Black, Grönqvist, & Öckert, 2018). The role of parenting style and
the main caregiver in shaping children's noncognitive ability development has been largely overlooked.
Even though a large psychological literature has linked parenting style to outcomes in childhood and adolescence, economists


Corresponding author.
E-mail addresses: 20190298@m.scnu.edu.cn (L. Deng), tongtingting@dufe.edu.cn (T. Tong).

https://doi.org/10.1016/j.chieco.2020.101477
Received 31 March 2019; Received in revised form 7 May 2020; Accepted 7 May 2020
Available online 11 May 2020
1043-951X/ © 2020 Elsevier Inc. All rights reserved.
L. Deng and T. Tong China Economic Review 62 (2020) 101477

have only recently begun to explicitly consider parenting style as inputs in the traditional models of human capital development
(Cobb-Clark, Salamanca, & Zhu, 2019). For instance, some studies have modeled parenting style in a game-theory structure, where
both parents and children are players, and parents optimally choose a parenting strategy to maximize children's good performance
(Burton, Phipps, & C. L., 2002; Cosconati, 2009). Some other studies empirically evaluated the influence of parenting style and found
that respectful, positive, and warm parenting styles are beneficial for the development of noncognitive ability (e.g., Cobb-Clark et al.,
2019; Dooley & Jennifer, 2007; Fiorini & Keane, 2014).
In China, more and more parents pay attention to childhood education, and the strict “tiger mother” parenting practice becomes
popular in Chinese families. Chinese parents' concept of “careful and intensive care” for their children has motivated them to con-
tinuously invest time and money in children's education. For instance, in 2019, nearly 40% of families spend 20% to 30% of annual
income to improve their children's knowledge and skills, and many parents gave up their own interests to be able to accompany their
children (51JOB, 2019). Although parents agree on the importance of companionship and investment, it is not clear what kind of
parenting style and investment is more effective for children's accumulation of human capital in China, especially for the non-
cognitive ability, which is considered as soft skills and are more difficult to measure (Fan, Wei, & Zhang, 2017).
In addition, grandparents' participation in children's education is prevalent in China, which plays an important role in child
development (see, for example, Bengtson, 2001, Jæger, 2012). In China, approximately 40% of the children are mainly taken care by
their grandparents. On the one hand, grandparents contribute to the development of children's human capital by serving as role
models and providing care, discipline, and supervision (Chen, Lin, & Mair, 2011; Zeng & Xie, 2014). On the other hand, the over-
protection of grandparents and the absence of parental care may adversely influence noncognitive ability development (Chang et al.,
2019). Therefore, it remains unclear about the role of grandparents in skill formation, and such analysis becomes critical considering
China's “open two-child policy” implemented in 2015.
To fill this research gap, this study investigates the role of parenting style on children's noncognitive ability and compares the role
of parents and grandparents in childcare based on data from China Family Panel Studies (CFPS). Based on an extensive set of
questions of caregiver-child interaction, we distinguish parenting style (i.e., respectful and disciplinary) from traditional material and
time parental investment. We find that both respectful and disciplinary parenting style significantly influence children's noncognitive
ability. Our results suggest that with detailed measures of parenting style and parental investment, grandparenting does not have a
significant impact on children's noncognitive ability. A further investigation of unobserved heterogeneity based on panel data
techniques and Oster's (2017) coefficient stability method reveals a causal and long-lasting relationship between parenting style and
noncognitive ability.
This study makes several contributions to the existing literature. This study contributes to the existing literature in a number of
ways. First, current economic literature about parenting style on noncognitive ability is still limited, and none of them is conducted in
the context of Asian countries, where parents tend to be more disciplinary, and the role of the grandparents is more prominent
compared to the West. Thus, our analysis contributes to both economics and psychological literature by conducting a comprehensive
analysis of the relationship between parenting style and noncognitive ability. Second, we specifically compare the role of parents and
grandparents in shaping children's noncognitive skills, which provides new evidence about intergenerational parenting. Third, we
investigate the dynamic nature of children's noncognitive ability development and examine how the current noncognitive ability is
affected by previous parenting style and noncognitive ability. To our knowledge, only a few studies (Cunha & Heckman, 2008; Fiorini
& Keane, 2014; Khanam & Nghiem, 2016) have specifically examined such dynamic nature. Finally, we analyze and address the
endogeneity issue of parenting styles by applying the Lewbel IV estimation method (Baum & Lewbel, 2019; Lewbel, 2012) and the
coefficient stability method (Oster, 2017). Therefore, by utilizing appropriate econometric techniques, we attempt to identify a causal
impact from parenting style to noncognitive ability.
The rest of this paper is organized as follows. Section 2 reviews and summarizes the literature. Section 3 describes the data and
measures of noncognitive ability and parenting style. Section 4 presents our empirical results, in which we examine the role of
parenting style and main caregiver. We further investigate the potential unobserved heterogeneity issue in Section 5. Section 6
concludes.

2. Literature review

Our study is related to two strands of literature. The first strand relates to the impact of parenting style on human capital
development, and the second strand evaluates whether the development of human capital differs with respect to different caregivers
(i.e., parents or grandparents). We summarize the literature in below.

2.1. Parenting style and human capital development

There are numerous studies about parenting style in developmental psychology (Aunola, Stattin, & Nurmi, 2000). According to
Baumrind (1971, 1989) and Maccoby and Martin (1983), the main differences in parenting style lie in two dimensions, respon-
siveness and demandingness. Responsiveness refers to the extent parents are supportive and sensitive toward their children, while
demandingness refers to the extent parents supervise, enact disciplinary efforts, and hold maturity demands for their children. Based
on the combination of these two dimensions, parenting styles can be classified into four categories: authoritative, authoritarian,
permissive, and disengaged (Baumrind, 1991). Authoritative parenting is both demanding and responsive, which is associated with
high parental involvement, trust, and encouragement (Ginsburg & Bronstein, 1993; Paulson, 1994; Pulkkinen, 1982); Authoritarian
parenting is demanding but not response, with strict control but less communication and encouragement (Maccoby & Martin, 1983);

2
L. Deng and T. Tong China Economic Review 62 (2020) 101477

Permissive parenting is responsive but not demanding, with parents having high levels of warmth and a child-centered attitude but
less parental discipline (Baumrind, 1989); Disengaged parenting is neither demanding nor responsive, with parents showing little
interest in children's activities, opinions, and emotions (Maccoby & Martin, 1983).
There is a large literature in developmental psychology that investigate the link between parenting style and a range of outcomes
such as school performance (Dornbusch, Ritter, Leiderman, Roberts, & Fraleigh, 1987; Spera, 2005), personality (Weiss & Schwarz,
1996), social-communicative competence (Hart, Newell, & Olsen, 2003), achievement strategies (Aunola et al., 2000), and various
risky behaviors (Chan & Anita, 2011). In general, the results show that authoritative parenting is usually associated with children's
positive outcomes while the rest three parenting styles have been linked to passivity and underachievement (Baumrind, 1991).
Despite the extensive analysis about parenting style in psychological literature, we can find only a few papers in economics
analyzing the influence of parenting style on child development in developed countries (Burton et al., 2002; Cobb-Clark et al., 2019;
Cosconati, 2009; Dooley & Jennifer, 2007; Fiorini & Keane, 2014; Khanam & Nghiem, 2016). Some studies measure parenting style by
a single index. For instance, Cosconati (2009) developed and estimated a model to understand the impact of parenting styles on the
development of human capital in children. More specifically, parenting style is measured by the strictness of the limits parents set for
their children about time allocation. The results indicate that the efficiency of parenting style depends on how much a child values
human capital. Burton et al. (2002) estimated the impact of parenting style on child disorder based on a simultaneous model. They
measure parenting style by a single index about how the parents respond when the child misbehaves, and their results indicate that
worse parenting behavior increases conduct disorder.
Other studies measure parenting style from various aspects. For instance, Cobb-Clark et al. (2019) examined the impact of
parenting style (i.e. respectful and monitoring) on the development of human capital based on a household production function
approach, and results showed that parenting style, especially respectful parenting, significantly influences youth's outcomes such as
noncognitive ability and risky behavior. Fiorini and Keane (2014) studied how the allocation of children's time and mother's par-
enting style (i.e. mother warmth, effective mother discipline) affect cognitive and noncognitive development. They found that mother
warmth and discipline are associated with better performance. In addition, Dooley and Jennifer (2007) also found that positive
parenting is related to better behavioral and emotional well-being, while hostile parenting is related to worse well-being. In general,
despite with different measures of parenting style, existing economics studies find that parenting style has significant impacts on
human capital development in developed countries.

2.2. Main caregiver and human capital development

Considering the significant role of grandparents in the intergenerational mobility, it becomes “necessary to look beyond the
nuclear family” (Bengtson, 2001). An increasing number of studies start to examine the influence of grandparents on grandchildren,
but results are inconclusive. For instance, Zeng and Xie (2014) estimated the impact of grandparents on grandchildren's education
attainment in rural China and found that the educational level of co-resident grandparents directly affects grandchildren's educa-
tional outcome. Chan and Boliver (2013) found a significant association between grandparents' and grandchildren’ class positions in
England. Grandchildren is likely to be in a more prestigious occupational class level if their grandparents are themselves in that class.
In contrast, Jæger (2012) found that the socio-economic characteristics of grandparents have few direct effects on grandchildren's
educational success.
The role of grandparents in shaping children's human capital can also be reflected by the studies about left-behind children, who
are mainly taken care by grandparents. Some studies found that parental migration has a negative impact on children's academic
performance and mental outcomes (Chang et al., 2019; Meng & Yamauchi, 2017; Meyerhoefer & Chen, 2011; Zhang, Behrman, Fan,
Wei, & Zhang, 2014; Zhao, Yu, Wang, & Glauben, 2014). According to Chang et al. (2019), lack of parental care is the main factor
impacting mental health of left-behind children. On the other hand, some studies do not find any significant impact of parental
migration on the performance of left-behind children, indicating that grandparents can also take good care of them (Graham &
Jordan, 2011; Lu, 2012). In our view, previous studies of left-behind children have not systematically compared the role of parents
and grandparents in affecting children's human capital development, especially in the noncognitive ability.
In summary, current economic literature about parenting style on noncognitive ability is very limited, and none of them is
conducted in the context of Asian countries, where parents tend to be more disciplinary compared to the West. In addition, previous
studies have not adequately compared the role of parents and grandparents in influencing children's noncognitive ability. More
research is needed in this area, especially with the substantial involvement of grandparents in child care in Asian countries.

3. Theoretical framework

Our theoretical framework is based on the cognitive/noncognitive skill production function extensively studied by Todd and
Wolpin (2003, 2007), and Cunha and Heckman (2007, 2008, 2009). According to the skill production function, child development is
a cumulative process depending on the innate ability and the history of family and school inputs, in which family plays an essential
role in shaping children's ability. Following Todd and Wolpin (2003, 2007) and Cunha and Heckman (2007, 2009), we model skill
formation of a child with a production function containing three main components: endowment, school input, and family input,
which can be written as the following equation,
Noncog = f (Endow, School, Family ), (1)
where Noncog is a measure of noncognitive ability, Endow is the endowment of the child, School denotes school inputs, and Family

3
L. Deng and T. Tong China Economic Review 62 (2020) 101477

Table 1
Parenting style and noncognitive ability measures in CFPS data.
CFPS Parenting style questions Noncognitive ability questions

2010 None Emotional stability for children aged 10 to 15, Locus of control and self-esteem for children aged 13 and
15
2012 Available for children aged 11, 13, and 15 Emotional stability for children aged 10 to 15
2014 Available for children aged 10 to 15 Self-esteem, Locus of control, emotional stability for children aged 10 to 15
2016 None Emotional stability for children aged 10 to 15

denotes family inputs. According to Cunha et al. (2010), family inputs incorporate parental investment and parental environments of
childhood. Thus, noncognitive production function can be written as:
Noncog = f (Endow, School, Family (Investment , Environment )). (2)

Parental investment (i.e., Investment) includes monetary resources that are invested to children, such as nutrition, school supplies,
and health care (Todd & Wolpin, 2003, 2007). For instance, affluent family may send children to expensive extracurricular classes to
improve their noncognitive ability, such as communication skills and leadership skills. In addition to the material-intensive in-
vestment, parental investment also includes the time spent with children. Many studies have shown that mother's time spent with
children plays an important role in developing children's ability (Bernal & Keane, 2011; James-Burdumy, 2005).
In addition to parental investment, Cunha and Heckman (2007) show that a good parental environment (i.e., Environment)
enhances the effectiveness of all investments in child development, and noncognitive skills are an important product of a successful
parental environment. Some previous studies used the Home Observation for Measurement of the Environment (HOME) as a measure
of parental environment, with a higher score indicating a more desirable parental environment (Caldwell & Bradley, 1979). In the
current study, we adopt a more direct and comprehensive measure of parental environment, which is parenting style. We construct
different measures of parenting style based on caregivers' interactions with children. For instance, some parents maybe more dis-
ciplinary while some maybe more laissez-faire, which could have different impact on children's noncognitive ability.
An important element of parental environment typically overlooked by previous studies is the caregiver. Intergenerational par-
enting is prevalent in many Asian countries, especially in China. Grandparents may have different parenting styles compared with
parents. In addition, the impact of the same parenting style on children's noncognitive ability may differ depending on different
caregivers. Therefore, we further define parental environment as a function of parenting style (Style) and the caregiver (Care),
Environment = f (Style, Care ), (3)

where Style is a vector of parenting style and Care is the main caregiver (parents vs. grandparents).
Therefore, we obtain the following empirical model of noncognitive skill production:
Noncog = 0 + 1 Endow + 2 School + 3 Investment + 4 Style + 5 Care + 6 Control + u, (4)

where Endow represents endowments. We use educational attainment of parents as proxies for children's endowment (Bernal, 2008;
Cunha & Heckman, 2007, 2009). School represents school inputs, such as the quality and location of the school, Investment includes
the family's material investment and time investment in children, Style represents parenting styles of the main caregiver, such as
respective and disciplinary, Care represents the main caregiver, Control includes control variables such as family income and parents'
age, and u is the error term.

4. Data

Our data come from the China Family Panel Studies (CFPS) conducted by the Institute of Social Science Survey (ISSS) at Peking
University. CFPS aims to reflect the social, economic, demographic, educational, and health changes in China by tracking and
collecting data at the individual, family, and community levels. CFPS has officially released four waves of survey data in 2010, 2012,
2014, and 2016. Detailed information for each wave of data is summarized in Table 1.1 According to Table 1, CFPS in 2010 and 2016
do not have questions about parenting styles, and CFPS-2012 contains only one noncognitive ability measure. In addition, we cannot
construct a panel data set because there is not enough overlap between CFPS-2012 and CFPS-2014 samples. Thus, our main analysis
is based on the third wave of the survey (i.e., CFPS-2014). Compared to other waves, CFPS-2014 contains not only numerous
questions about individuals' noncognitive ability, but also a rich set of questions about the interactions between child and the main
caregiver (usually parents or grandparents), which provides better measures of parenting style and noncognitive ability.
The original sample in CFPS-2014 covers 8617 children and 37,147 adults in 13,946 families. However, the noncognitive ability
questions were only asked to children between age 10 to 15, and some of these children's parents did not respond to parenting style
questions, which leaves us a sample size of 1412. We compare the descriptive statistics of some basic characteristics between our final
sample and the original whole sample, and find that the statistics are similar, which indicates that our results may not be biased by
sample selection issues. For instance, children taken care by parents account for 57% in the original sample and 61% in our sample.

1
The detail information is available on the official website: http://opendata.pku.edu.cn/dataset.xhtml?persistentId=doi:10.18170/DVN/45LCSO

4
L. Deng and T. Tong China Economic Review 62 (2020) 101477

Male account for 53% in the original sample and 55% in our sample. Children going to key schools account for 23% in the original
sample and 24% in our sample. Other basic statistics such as parental age and educational level are also quite similar.

4.1. Parenting style and parental investment measures

Based on response to questions about caregiver-child interactions, we construct 18 interaction measures, which can be classified
into two types, attention-intensive interactions and investment-intensive interactions. Attention-intensive interactions include
questions about how caregiver communicates or disciplines child, such as “caregiver encourages you to do things (asked to child)”
and “Do you often require child to finish homework (asked to caregiver)”, while investment-intensive interactions include questions
about the goods- and time-investment that caregiver made to child, such as “caregiver reads stories to you (asked to child)” and
“money spent on child's education for the past 12 months (asked to caregiver)”. Following the study by Cobb-Clark et al. (2019), we
attempt to distinguish parenting style from traditional parental investment measures. Specifically, parenting style variables are
derived from attention-intensive interactions, and parental investment variables are constructed from investment-intensive inter-
actions.2
Principal component analysis (PCA) is employed to derive parenting style and parental investment variables among 18 caregiver-
child interaction measures. PCA has been frequently used to construct parenting styles from multiple items in economics studies (for
instance, Cobb-Clark et al., 2019, Fiorini & Keane, 2014). Table 2 displays the results of two sets of principal component analysis. The
first set of PCA is based on 9 attention-intensive interactions (column 1 and column 2). We report the rotated factor loadings of two
components with eigenvalues larger than one, which account for 41% of the variation.3 Factor loadings larger than 0.3 in absolute
value are displayed in bold. Table 2 shows that the first component loads highly on questions about how caregiver communicates
with child and whether caregiver respects child, and we interpret this factor as an index of “respectful” parenting. The second
component loads highly on questions about how caregiver disciplines their children, and we label the index “disciplinary” parenting.
The second set of PCA is based on 9 investment-intensive interactions (column 3 and column 4). Similar with the first set, we display
the rotate factor loading of two components with the criteria of eigenvalue larger than one. Based on factor loadings, we interpret the
two resulting indices as “material-intensive” and “time-intensive” parental investments.4

4.2. Noncognitive ability measures

In the current study, we construct three measures of noncognitive ability, including Rosenberg self-esteem scale, locus of control,
and emotion stability. Rosenberg self-esteem scale measures self-worth by measuring both positive and negative feelings about the
self (Rosenberg, 1965). Locus of control measures the degree to which an individual perceives success or failure as being dependent
on one's own action (internal control), as opposed to external forces beyond his control (external control) (Rotter, 1966). Emotion
stability measures an individual's emotional status, which is adapted from the study by Dooley and Jennifer (2007). These measures
are widely used in economic studies. For instance, Heckman et al. (2006) employed Rosenberg self-esteem scale and locus of control
to measure noncognitive ability.
The questions used to construct noncognitive ability measures are presented in Appendix A. Rosenberg self-esteem scale is
constructed based on 10 questions, and the answer values for each question range from 1 (strongly agree) to 5 (strongly disagree).
The self-esteem scale is the sum of scores of these 10 questions, ranging from 10 to 50, and a higher scale indicates a higher degree of
approval toward oneself. Locus of control is constructed based on 9 questions in a similarly way, with a higher score indicating a
higher internal locus of control. In addition, emotional stability is derived from 6 questions, with a higher score indicating a better
emotional status. Table 3 presents the descriptive statistics of noncognitive ability measures. Rosenberg self-esteem scale ranges from
19 to 47, with an average of 36.12. Locus of control ranges from 23 to 49, with an average of 36.06. Emotion stability ranges from 12
to 30, with an average of 26.82.

4.3. Caregiver type

CFPS includes two questions about caregiver, including “who is the main caregiver during the day” and “who is the main
caregiver during the night”. Thus, there are four kinds of care scenarios, which are 1) parents during the day and grandparents during
the night; 2) grandparents during the day and parents during the night; 3) grandparents during both day and night; 4) parents during
both day and night. We define caregiver to be one if the child is taken care by parents during both day and night (i.e., scenario 4), and
zero if grandparents are involved in childcare during either day or night (i.e., scenarios 1–3). Table 3 shows that 61% of the children
are cared by parents, while the rest 39% are cared by grandparents or both.

2
Parental investment refers to the investment made by the main caregiver.
3
We employ varimax rotation matrix to facilitate interpretation. The results of other rotation matrix remain robust, and the results are upon
requests.
4
The identification of parenting style and parental investment follows the study by Cobb-Clark et al. (2019) and Khanam and Nghiem (2016).

5
L. Deng and T. Tong China Economic Review 62 (2020) 101477

Table 2
Principal analysis of parental investment and parenting style.
Parenting style Parental investments

Respectful Disciplinary Material-intensive Time-intensive

Eigenvalues 2.63 1.90 2.29 1.46


Variation captured 0.24 0.17 0.21 0.13
Attention-intensive interactions:
Child: When you are wrong, caregiver asks for the reasons and discusses with you 0.42 −0.01
Child: Caregiver encourages you to do things 0.45 0.00
Child: Caregiver talks to you friendly 0.37 0.06
Child: Caregiver encourages you to think independently 0.42 −0.02
Child: Caregiver explains the reason to you if they want you to do something 0.37 −0.03
Child: Caregiver likes to talk to you 0.40 0.02
Caregiver: Do you often require child to finish homework 0.02 0.46
Caregiver: Do you often stop child from watching TV 0.03 0.53
Caregiver: Do you often restrain the types of TV show that child watches −0.01 0.44
Investment-intensive interactions
Child: Caregiver asks you about your school life −0.03 0.48
Child: Caregiver reads stories to you −0.14 0.52
Child: Caregiver plays with you 0.03 0.53
Child: Caregiver attends meetings in your school 0.03 0.42
Caregiver: Whether family saves money for child's education 0.24 0.08
Caregiver: Whether there is a computer at home 0.29 −0.04
Caregiver: Money spent on child's education for the past 12 months 0.55 −0.02
Caregiver: Money spent on child's extracurricular class for the past 12 months 0.59 0.17
Caregiver: Number of extracurriculars classes child participated for the past 12 months 0.44 −0.09

Note: Columns 1–2 report the rotated factor loadings of PCA based on a set of attention-intensive caregiver-child interactions, and columns 3–4
report the loadings based on investment intensive interactions. Factor loadings of the first two components are presented with eigenvalues larger
than one. The loadings are varimax rotated to facilitate interpretation, and loadings greater than 0.3 in absolute value are reported in bold.

4.4. Control variables

CFPS contains school quality measures. First, we include school type (i.e., key school) as an additional control variable. Key
schools are high quality schools designated by the Ministry of Education of China, with higher graduation rates, high-quality tea-
chers, and better school environment, teaching facilities, and learning atmosphere, which could impact the development of non-
cognitive ability. We set Key school = 1 if the child studies in a key school. Otherwise, Key school = 0. In addition, we control for
school locations, including provincial capital, city, county, and village, as proxies for school quality. It is commonly recognized in
China that schools in provincial capital and cities have higher quality compared with schools in counties and villages. Descriptive
statistics of school characteristics are presented in Table 3.
Descriptive statistics for individual characteristics such as age, gender, education level, and hukou status, and family background
information such as parents' age, education, and family income are also presented in Table 3.

5. Empirical results

Table 4 presents the results of the influence of parenting style and parental investment on children's noncognitive ability using the
weighted least square estimation method. To be presentative to the whole population, we use the weight provided by the survey,
which is CFPS-2014 cross-section weights: national samples.5 To facilitate understanding, measures of parenting style, parental
investment, and noncognitive ability are standardized to have mean zero and standard deviation one in the following analysis. We
find that respectful parenting has significant influence on all three noncognitive ability measures. A one standard deviation increase
in respectful parenting index increases Rosenberg self-esteem scale, locus of control, and emotion stability by 0.138, 0.205, and 0.227
standard deviation, respectively. The positive impact of respectful parenting on noncognitive ability development is consistent with
previous literature (Baumrind, 1991; Maccoby & Martin, 1983). For instance, Cobb-Clark et al. (2019) found that a one standard
deviation in respectful parenting is associated with a 0.314 standard deviation increase in locus of control using Australia data.
The disciplinary parenting style, on the other hand, is negatively associated with locus of control, but is unrelated to self-esteem
scale and emotion stability. Specifically, a one standard increase in disciplinary parenting decreases locus of control by 0.094
standard deviation. It indicates that parents setting strict requirements and limits toward children, such as requiring child to finish
homework and restraining the types of TV show child can watch, reduce children's belief that they can control over their lives
through self-determination or self-motivation (i.e. internal local of control). Consistent with our results, we also find evidence in
literature showing that individuals from controlling families are more external in locus of control (Baumrind, 1991; De Man, LeDuc, &

5
The same weight is applied to the rest of the analysis except for Table 8. In Table 8, we adopt the CFPS-2016 weight because the dependent
variable in Table 8 is obtained from CFPS-2016.

6
L. Deng and T. Tong China Economic Review 62 (2020) 101477

Table 3
Descriptive statistics.
Obs. Mean SD Mix Max

Noncognitive ability
Rosenberg self-esteem scale 976 36.12 3.75 19 47
Locus of control 1401 36.06 4.01 23 49
Emotion stability 1408 26.82 3.71 12 30
Parenting style
Respectful 1412 0 1.62 −5.76 3.4
Disciplinary 1412 0 1.38 −3.15 5.12
Parental investment
Material-intensive 1379 0 1.32 −3.19 3.96
Time-intensive 1379 0 1.19 −2.06 9.87
Caregiver
Caregiver (parents = 1) 1312 0.61 0.50 0 1
Child characteristics
Age 1412 11.3 1.21 10 15
Gender (Male = 1) 1412 0.55 0.5 0 1
Middle school = 1 1412 0.17 0.37 0 1
Hukou (City = 1) 1412 0.21 0.41 0 1
Family background
Mother's age 1412 37.45 5.07 21 60
Mother's education level
Illiterate 1412 0.26 0.44 0 1
Middle school 1412 0.53 0.5 0 1
High school or above 1412 0.21 0.41 0 1
Father's age 1412 39.24 5.07 26 62
Father's education level
Illiterate 1412 0.15 0.36 0 1
Middle school 1412 0.59 0.49 0 1
High school or above 1412 0.26 0.44 0 1
Annual family income/person 1356 9547.33 9810.36 100 170,000
School quality
Key school = 1 1402 0.24 0.42 0 1
School location
Provincial level 1402 0.05 0.21 0 1
City level 1402 0.13 0.33 0 1
County level 1402 0.19 0.39 0 1
Village level 1402 0.63 0.39 0 1

Labreche-Gauthier, 1992; Trusty & Lampe, 1997).


In addition, we find both material- and time-intensive parental investment are beneficial for the development of self-esteem. A
one standard deviation increase of material- or time-intensive investment increases self-esteem scale by 0.113 and 0.072 standard
deviations, respectively. We also find that the influence of time-intensive investment on locus of control is positive and significant,
while material-intensive investment has a negative impact on children's emotion stability. The importance of parental investment on
child outcomes is largely consistent with previous literature (Fiorini & Keane, 2014; Khanam & Nghiem, 2016; Spera, 2005). By
comparing the impacts of parenting style and parental investment, we find parenting style, especially respectful parenting has a
strong and positive association with noncognitive ability development, indicating the importance incorporating parenting style into
the analysis of human capital development.
Furthermore, consistent with previous literature, we find parental education has significant impact on the development of
children's noncognitive ability (Khanam & Nghiem, 2016). Moreover, family income is another significant factor affecting all three
noncognitive ability measures.
Table 5 presents the results controlling for the main caregiver. The influence of parenting style and parental investment remain
highly consistent with previous estimation in Table 4. However, after controlling for parenting style and parental investment, we do
not observe any significant impact from caregiver type. Compared to those taken care by grandparents or both, children taken care by
parents do not show better noncognitive ability. The insignificance of caregiver types seems different from previous studies, which
indicates that children taken care by grandparents (i.e., parents out-migrated) have worse performance and mental health (e.g.,
Chang et al., 2019; Zhang et al., 2014). One possible reason is that previous studies do not control for detailed measures of parenting
style and parental investment.6
To explore whether the impacts are heterogeneous across genders, we split the sample for girls and boys. The impact of parenting
style and parental investment may differ across genders in China due to the skews intra-household allocation in favor of sons (Brown

6
We try to incorporate interaction terms between parenting style, parental investment, and caregiver. However, most of interactions terms are not
significant, and parenting style measures become insignificant after controlling for interactions terms due to high multicollinearity. Due to space
limit, results with interaction terms are available upon requests.

7
L. Deng and T. Tong China Economic Review 62 (2020) 101477

Table 4
Impact of parenting style on noncognitive ability.
Rosenberg self-esteem Locus of control Emotion stability

Parenting style
Respectful 0.138⁎⁎⁎ 0.205⁎⁎⁎ 0.227⁎⁎⁎
(0.049) (0.044) (0.047)
Disciplinary −0.013 −0.094⁎⁎⁎ −0.004
(0.045) (0.035) (0.034)
Parental investment
Material-intensive 0.113⁎⁎ 0.028 −0.084⁎
(0.050) (0.039) (0.050)
Time-intensive 0.072⁎ 0.077⁎⁎ −0.014
(0.042) (0.038) (0.040)
Age −0.007 0.021 0.003
(0.063) (0.037) (0.038)
Gender −0.103 −0.025 −0.045
(0.080) (0.070) (0.075)
Middle school = 1 0.089 0.161 −0.058
(0.183) (0.113) (0.108)
Hukou (City = 1) 0.108 −0.092 −0.071
(0.125) (0.112) (0.122)
Mother's age 0.016 −0.009 −0.020⁎
(0.012) (0.012) (0.011)
Middle school_Mother 0.167 0.116 0.305⁎⁎⁎
(0.102) (0.085) (0.109)
High school or above_Mother 0.099 −0.095 0.400⁎⁎⁎
(0.147) (0.121) (0.128)
Father's age −0.012 0.01 0.024⁎
(0.012) (0.011) (0.013)
Middle school_Father 0.196 0.174⁎ 0.06
(0.134) (0.099) (0.125)
High school or above_Father 0.111 0.313⁎⁎ 0.07
(0.164) (0.128) (0.131)
Family income/person 0.062⁎ 0.104⁎⁎⁎ 0.112⁎⁎⁎
(0.037) (0.035) (0.038)
School location
Provincial capital level 0.138 0.106 0.016
(0.114) (0.095) (0.102)
City level 0.194 0.198 −0.083
(0.143) (0.132) (0.171)
County level 0.006 −0.06 0.109
(0.164) (0.168) (0.144)
Key school = 1 0.269⁎⁎ 0.067 −0.08
(0.110) (0.090) (0.090)
Constant −0.856 −1.447⁎⁎ −1.434⁎⁎
(0.761) (0.562) (0.666)
N 880 1271 1276
Adj-R2 0.112 0.135 0.079

Note: Heteroscedasticity robust standard errors are reported in the parentheses. ⁎⁎⁎, ⁎⁎, and ⁎ denote significance at 1%, 5%, and 10%, respectively.

& Park, 2002; Connelly & Zheng, 2003). The results are displayed in Table 6. Overall, consistent with the whole sample in Table 4,
respectful parenting has significant impact on noncognitive ability measures of both boys and girl except for girls' self-esteem scale.
Compared to girls, it seems respectful parenting has stronger impact on boys' self-esteem and emotion stability. In addition, the
negative impact of disciplinary parenting on locus of control we found in Table 4 is mainly reflected in boys. Interestingly, we find
material-intensive investment positively affects girls' self-esteem but negatively affect their emotion stability. Our results indicate that
compared to boys, girls tend to be more sensitive to their material conditions.

6. Unobserved heterogeneity

Unobserved heterogeneity in our study can arise in several forms. The main concern is that we cannot fully control for all the
factors affecting children's noncognitive ability. For instance, children's heritable endowment is an important unobserved determi-
nant of their human capital development. In addition, historical parenting style and parental investment could also influence current
noncognitive ability. Such omitted variable bias could confound our estimates of the association between parenting style, parental
investment, and noncognitive ability. Another concern lies in the mutual relationship between parents and children. When studying
the influence of parenting style on children, it is necessary to consider how the existing personality and behavior of children affect the
choice of parenting style (i.e., reverse causality).
The direction of endogeneity bias depends on the impact of unobserved characteristics on noncognitive ability as well as the

8
L. Deng and T. Tong China Economic Review 62 (2020) 101477

Table 5
Impact of parenting style and caregiver type on noncognitive ability.
Rosenberg self-esteem Locus of control Emotion stability

Parenting style
Respectful 0.138⁎⁎⁎ 0.204⁎⁎⁎ 0.226⁎⁎⁎
(0.049) (0.044) (0.047)
Disciplinary −0.013 −0.092⁎⁎⁎ −0.002
(0.045) (0.035) (0.035)
Parental investment
Material-intensive 0.112⁎⁎ 0.027 −0.086⁎
(0.050) (0.039) (0.049)
Time-intensive 0.073⁎ 0.078⁎⁎ −0.013
(0.043) (0.038) (0.040)
Caregiver (parents = 1) −0.012 −0.042 −0.049
(0.082) (0.073) (0.072)
Individual characteristics Yes Yes Yes
Family background Yes Yes Yes
School characteristics Yes Yes Yes
N 880 1271 1276
Adj-R2 0.111 0.135 0.078

Note: Individual characteristics include age, education, and Hukou status. Family background variables include parents' age, education, and family
income. School characteristics include school quality and school location. Heteroscedasticity robust standard errors are reported in the parentheses.
, , and ⁎ denote significance at 1%, 5%, and 10%, respectively.
⁎⁎⁎ ⁎⁎

Table 6
Impact of parenting style on noncognitive ability based on subsamples.
Rosenberg self-esteem Locus of control Emotion stability

Girls Boys Girls Boys Girls Boys

Parenting style
Respectful −0.011 0.249⁎⁎⁎ 0.245⁎⁎⁎ 0.164⁎⁎⁎ 0.152⁎⁎⁎ 0.282⁎⁎⁎
(0.074) (0.061) (0.069) (0.055) (0.059) (0.065)
Disciplinary 0.002 −0.006 −0.064 −0.109⁎⁎ −0.084 0.061
(0.063) (0.060) (0.056) (0.045) (0.054) (0.045)
Parental investment
Material-intensive 0.238⁎⁎⁎ −0.009 0.035 0.010 −0.145⁎ −0.047
(0.077) (0.068) (0.054) (0.057) (0.080) (0.058)
Time-intensive 0.067 0.041 0.070 0.085 0.073 −0.100⁎
(0.053) (0.070) (0.044) (0.063) (0.048) (0.061)
Individual characteristics Yes Yes Yes Yes Yes Yes
Family background Yes Yes Yes Yes Yes Yes
School characteristics Yes Yes Yes Yes Yes Yes
N 0.179 0.123 0.119 0.153 0.070 0.128
Adj-R2 396 484 569 702 571 705

Note: Individual characteristics include age, education, and Hukou status. Family background variables include parents' age, education, and family
income. School characteristics include school quality and school location. Heteroscedasticity robust standard errors are reported in the parentheses.
, , and ⁎ denote significance at 1%, 5%, and 10%, respectively.
⁎⁎⁎ ⁎⁎

correlation between parenting style and unobserved characteristics. Positive bias would occur if parenting style is positively related
with unobserved characteristics that also positively affect noncognitive ability. For instance, a high-quality parenting style may be
positively correlated with a warm home atmosphere, which is unobserved but promotes the development of children's noncognitive
ability, resulting an over estimation of the impact of positive parenting style. Negative bias may result if parenting style is negatively
related with unobserved characteristics that positively affect noncognitive ability. For instance, parents may adopt unusual high-
quality parenting styles if their children show some deficits in natural endowment, which is hard to measure, resulting an under-
estimation of the impact of positive parenting style.
The literature has proposed different strategies to solve such endogeneity problem (Todd & Wolpin, 2003, 2007). For instance,
many studies use contemporaneous OLS estimators by including a rich set of control variables, which is the method adopted in our
study in the previous section (Fiorini & Keane, 2014). By incorporating a rich set of controls such as parenting style, parental
investment, and family background information, we attempt to minimize the bias caused by omitted variables. In addition, following
previous literature (Bernal, 2008; Cunha & Heckman, 2007, 2009), we use educational attainment of parents as proxies for ability
endowment. Other commonly used methods include fixed effect specifications, which attempt to difference out unobservables over
time, and value-added specifications, which use lagged terms of dependent and independent variables to control for historical inputs
and heritable endowments. Considering our data structure, we propose the following three methods to deal with endogeneity.

9
L. Deng and T. Tong China Economic Review 62 (2020) 101477

6.1. Heteroskedasticity-based identification

Lewbel (2012) proposed a new identification method in models with endogenous variables, which is especially useful when other
sources of identification such as instrumental variables and repeated measures are not available. As an alternative to IV estimation,
the Lewbel IV estimation method has been widely used in literature (Emran & Hou, 2013; Sabia, 2007). In specific, we use the
following equation to demonstrate the application of Lewbel IV identification strategy in our study:

Y1 = X + Y2 + 1 (5)

Y2 = X + 2 (6)

where Y1 represents noncognitive ability, Y2 represents the endogenous variable, such as parenting style and parental investment
measures, X includes exogenous variables such as age, hukou status, gender, and school locations. The key assumptions for applying
the Lewbel IV estimation are Cov(Z, θ1θ2) = 0 and Cov (Z , 22) 0 , where either Z equals to X or Z is a subset of the elements of X. The
reason for using Z as a subset of X is to conserve degrees of freedom, but with the risk of failing to detect heteroscedasticity in Eq. (5)
(Baum, Schaffer, & Stillman, 2003). Thus, in our study, we choose Z equals to X to increase the power of detecting heteroscedasticity.
After choosing the variables contained in Z, the Lewbel IV estimation can be conducted in two steps. First, we estimate Eq. (6) using
ordinary least squares estimation and obtain estimated residuals 2 . Then, Eq. (5) can be estimated using two stage least squares
estimation method using X and (Z Z ) 2 as instruments, where Z is the sample mean of Z.
In practice, it is difficult to verify whether the key assumptions, that Cov(Z, θ1θ2) = 0 and Cov (Z , 22) 0 , hold. Baum and Lewbel
(2019) discussed three conditions that are sufficient to make the key assumptions hold. However, it is also possible that the as-
sumptions hold even if these conditions are not verified. As a result, the satisfaction of three conditions can be used as a positive
signal but not a mandatory requirement for the application of Lewbel IV estimation.
The first condition is that Y2 is endogenous because it contains an unobserved error component that affects both Y1 and Y2. In our
study, the natural endowment of a child is unobserved, which influences both parenting style and the development of noncognitive
ability, thus the first condition can be satisfied. Second, the residuals in Eq. (5) should be homoscedastic, which, according to Baum
and Lewbel (2019), should be tested using the Pagan-Hall test (Pagan & Hall, 1983), the logic behind which is that the residuals are
homoscedastic if it cannot be predicted by any of the exogenous variables. Results of Pagan-Hall test in Table A2 show that the null
hypothesis of homoscedasticity cannot be rejected in two out of three of noncognitive ability measures, including Rosenberg self-
esteem scale and locus of control. Third, the residuals in Eq. (6) should be heteroscedastic. In our study, we find that the results of the
White heteroskedasticity test (White, 1980) reveal the presence of heteroskedasticity in most of our specifications, as presented in
Table A2. Therefore, three conditions for the key assumptions to hold are satisfied to a large extent, and our results of Lewbel IV
estimation should be valid.
The results of Lewbel IV estimation is presented in Table 7. The over-identification tests based on Hansen J-statistics in Table A2
show that the null hypothesis of exogeneity of the set of instruments cannot be rejected in all specifications. Consistent with previous
estimations in Table 4, Table 7 shows that respectful parenting style positively affects the development of all three noncognitive
ability measures, with coefficients ranging from 0.312 to 0.356. In addition, time intensive investment is beneficial for the devel-
opment of self-esteem and locus of control, which also align with the previous findings in Table 4.

Table 7
Impact of parenting style on noncognitive ability-Lewbel IV estimation.
Rosenberg self-esteem Locus of control Emotion stability

Parenting style
Respectful 0.312⁎ 0.329⁎ 0.356⁎⁎
(0.161) (0.194) (0.174)
Disciplinary −0.066 −0.071 0.048
(0.136) (0.256) (0.238)
Parental investment
Material-intensive 0.181 0.370⁎⁎ −0.093
(0.128) (0.175) (0.159)
Time-intensive 0.093⁎ 0.108⁎ −0.089
(0.053) (0.060) (0.063)
Individual characteristics Yes Yes Yes
Family background Yes Yes Yes
School characteristics Yes Yes Yes
Adj-R2 0.116 0.109 0.018
Observations 795 1159 1164

Note: Individual characteristics include age, education, and Hukou status. Family background variables include parents' age, education, and family
income. School characteristics include school quality and school location. Heteroscedasticity robust standard errors are reported in the parentheses.
, , and ⁎ denote significance at 1%, 5%, and 10%, respectively.
⁎⁎⁎ ⁎⁎

10
L. Deng and T. Tong China Economic Review 62 (2020) 101477

Table 8
Impact of historical inputs on emotion stability.
Specification A. 2016&2014 sample

Full sample Girls Boys

Emotion16 Emotion16 Emotion16

Parenting style
Respectful in 2014 0.082⁎⁎ 0.131⁎⁎ 0.047
(0.039) (0.056) (0.055)
Disciplinary in 2014 −0.044 −0.083 −0.014
(0.036) (0.052) (0.049)
Parental investment
Material-intensive in 2014 0.015 −0.012 0.037
(0.062) (0.065) (0.074)
Time-intensive in 2014 0.062 −0.012 0.116⁎⁎
(0.043) (0.065) (0.059)
Emotion stability in 2014 0.136⁎⁎⁎ 0.159⁎⁎⁎ 0.120⁎⁎
(0.039) (0.061) (0.050)
Adjusted R2 0.032 0.025 0.026
Observations 941 414 527

Specification B. 2016&2012 sample


Parenting style
Respectful in 2012 0.184⁎⁎ 0.249⁎⁎ 0.117
(0.084) (0.113) (0.108)
Disciplinary in 2012 0.179⁎⁎ 0.262⁎⁎⁎ 0.149
(0.077) (0.096) (0.123)
Parental investment
Material-intensive in 2012 0.107 −0.110 0.245⁎⁎
(0.097) (0.158) (0.101)
Time-intensive in 2012 −0.011 −0.001 −0.042
(0.083) (0.116) (0.116)
Emotion stability in 2012 0.003 0.032 −0.027
(0.067) (0.093) (0.090)
Adj-R2 0.081 0.200 0.015
Observations 227 114 113

Note: ⁎⁎⁎, ⁎⁎, and ⁎ denote significance at 1%, 5%, and 10%, respectively. Heteroscedasticity robust standard errors are reported in the
parentheses. To save space, results for individual, family, and school characteristics are not reported. Individual characteristics include age,
education, and Hukou status. Family background variables include parents' age, education, and family income. School characteristics
include school quality and school location.

6.2. Value-added specification

In the value-added specification, when estimating current noncognitive abilities, we control for lagged terms of noncognitive
abilities to control for historical inputs and heritable endowments (NICHD Early Child Care Research Network and Duncan, 2003).
Based on the data availability, we are only able to construct two sets of samples, one with years 2016 and 2014 and another with
years 2016 and 2012. Emotion stability in 2016 is used as the dependent variable because it is the only noncognitive ability measure
available across all three years. Independent variables include previous emotion stability, parenting style, and parental investment in
2014 (i.e., Specification A) and 2012 (i.e., Specification B). The results are displayed in Table 8.
Specification A shows that the coefficient of lagged emotional stability is highly significant in both full sample and subsamples,
implying emotional stability exhibits strong persistence in a three-year time period. The lagged respectful parenting style significantly
affects the current emotional ability for girls, and the lagged time-intensive investment is a significant predictors of current emotion
stability for boys. Specification B indicates that individuals' emotion stability in 2012 does not have a significant impact on the
emotion stability in 2016, suggesting that the persistence in emotion stability disappears with longer time periods. In addition, we
find both respectful and disciplinary parenting styles in 2012 are significantly associated with emotion stability in 2016, especially
for girls. In summary, results in specifications A and B suggest that parenting style and parental investment have a long-lasting impact
on children's noncognitive ability development.

6.3. Coefficient stability

Oster (2017) proposed a new method to assess the degree of omitted variable bias based on the changes in the coefficients and R2
across specifications with different sets of control variables, which has been widely used in recent years (e.g., Alesina, Harnoss, &
Rapoport, 2016; Bosquet, Combes, & García Peñalosa, 2018; Gehring & Schneider, 2018). The method draws on the strategies
proposed by Altonji, Todd, and Christopher (2005), who assumed that the selection on observable variables is informative about the

11
L. Deng and T. Tong China Economic Review 62 (2020) 101477

Table 9
Selection on unobservables.
Rosenberg self-esteem Locus of control Emotion stability

Parenting style
Respectful 7.338 4.935 1.449
Disciplinary 1.523 2.685 4.613
Parental investment
Material-intensive 7.784 2.962 −9.807
Time-intensive 1.731 1.508 −6.628

Note: Numbers in the table are estimated in Oster (2017) based on CFPS-2014 data.

selection on unobservables. They provided a measure of the relative importance of selection on unobservables, which can be written
as = F /( R F ) , where R and F are treatment effects from regressions with restricted set of controls and full set of controls,
respectively. A small gap between R and F means that the effect of selection on observables is small, indicating that selection on
unobservables needs to be stronger to explain the treatment effect in the full regression. Based on , Oster (2017) computed the ratio
= (R2 F R2 R )/(R2max R2 F ) , where R2 F and R2 R are the R2 from full and restricted regressions, and R2 max is the maximum R2
expected in the data. We set R2 max = 1.3 × R2 F . According to Oster (2017), the probability that estimated coefficients driven by
unobserved factors is low when exceeds one or below zero.
Table 9 presents the results of robustness analysis by Oster (2017). We calculate the for each type of parenting style and
parental investment. The restricted regressions are those with only one parenting style or parental investment as the independent
variable, and the full regressions include all the independent variables in Table 4. The values of for all of significant coefficients in
Table 9 are larger than one or smaller than zero, indicating that our estimations of parenting style and parental investment are not
entirely driven by selection on unobservables. Therefore, our previous estimation should reflect a causal relationship between
parenting style, parental investment, and noncognitive ability.

7. Conclusion

We investigate the impact of parenting style and parental investment on children's noncognitive ability development. Using
individual-level data for China, our basic estimation shows that respectful parenting is beneficial for the development of noncognitive
ability, measured by Rosenberg self-esteem scale, locus of control, and emotion stability, while disciplinary parenting negatively
influences children's locus of control. In addition, both material-and time-intensive parental investments benefit the development of
self-esteem. We then evaluate the impact of main caregiver on noncognitive ability and find that, after controlling for detailed
measures of parenting style and parental investment, grandparenting does not significantly influence the development of children's
noncognitive ability. As heterogeneity analysis, we further evaluate the effects between boys and girls, and the results remain
consistent.
To avoid estimation bias caused by the potential unobserved heterogeneity, we employ the Lewbel IV estimation method, which
also shows that the respectful parenting style has significant impacts on children's noncognitive ability. In addition, we use lagged
terms of noncognitive ability and lagged parenting style and parental investment to explain current noncognitive ability. Consistent
with previous estimation, we find evidence that lagged parenting style and parental investment affect current noncognitive ability,
and such impacts, especially from respectful parenting, last for several years. Moreover, based on the method proposed by Oster
(2017), we find evidence that our estimations of parenting style and parental investment are not likely to be driven by selection on
unobservables. Therefore, we reach the conclusion that parenting style and parental investment have a significant causal impact on
noncognitive ability.
The significant impact of parenting style and investment on children's noncognitive ability development, verified by robustness
analysis, has important policy implications. First, given the important role of parents-children interaction on children's noncognitive
ability development, it is critical for parents to have enough effective communication with their children. Thus, government policies
that support part-time or more flexible working practices for parents and policies that encourage parental engagement in home
learning activities could be valuable to the human capital accumulation of children.
Second, family education serves as a foundation for the whole education system. Parents are children's first teachers, and it is
critical for them to know how to educate their children. This study has identified several key factors such as a respectful parenting
style and more time investment that help the development of children's noncognitive ability, which provides some guidelines for
programs that focus on the improvement of parenting skills. Governments can encourage programs that provide educational support
for parents, which help parents to form positive parental attitudes and gain high-quality childcare skills.
Third, considering that noncognitive ability can be affected by parental material investment, the government could take actions to
implement personal income tax special deduction plans, which take households' actual financial situation into consideration. For
instance, taxpayers could pay lower personal income tax if they report the relevant expenditures on children's education to the tax
authorities. In China, starting from Jan 1, 2019, the special expense deductions such as children's education, continuing education,
and caring for the elderly have been introduced to China's individual income tax system. Such policies can help reduce the financial

12
L. Deng and T. Tong China Economic Review 62 (2020) 101477

burden of the family, and allow families to allocate more resources to children, which, together with an effective parenting style, can
be beneficial to the development of children's noncognitive ability.
Our study can be improved in the following ways. First, our study can be extended by incorporating time use data, such as the
media time, social activities time, and study time. An increasing number of studies have indicated that time use plays an important
role in developing children's noncognitive ability (Fiorini & Keane, 2014). Second, parenting style may have a long-lasting impact on
children even when they become adolescent and enter society. It would be interesting to have a panel data with longer time periods
and analyze if parenting style affects individual's future outcomes such as school achievement and career development.

Acknowledgements

This research is supported by the National Natural Science Foundation of China (Grant No. 71903024), Foundation of the
Educational Department of Liaoning Province (Grant No. LN2019Q38), and the National Social Science Fund of China (Grant No.
18CJY034).

Appendix

A.1. Construction of noncognitive ability measures

Noncognitive ability measures

Rosenberg self-esteem (score range: 10–50)

1. On the whole, I am satisfied with myself. (reversed)


2. A times I think I am no good at all.
3. I feel that I have a number of good qualities. (reversed)
4. I am able to do things as well as most other people. (reversed)
5. I feel I do not have much to be proud of.
6. I certainly feel useless at times.
7. I feel that I'm a person of worth, at least on an equal plane with others. (reversed)
8. I wish I could have more respect for myself.
9. All in all, I am inclined to feel that I am a failure.
10. I take a positive attitude toward myself. (reversed)
Answer values for each question are 1 (strongly agree) to 5 (strongly disagree)
Locus of control (score range: 9–45)
1 I pursue my own values instead of following others. (reversed)
2. Some children are born lucky.
3. When I start to do something, I will try my best to finish it. (reversed)
4. I make my own decisions about my life goals. (reversed)
5. Don't spend too much time trying, because things will never prove that it works.
6. Once you do something wrong, it is almost impossible to correct it.
7. The best way to deal with problems is not to think about them.
8. Bad things are going to happen no matter how hard you try to stop them.
9. I believe planning ahead will make things better. (reversed)
Answer values for each question are 1 (strongly agree) to 5 (strongly disagree)
Emotional stability (score range: 6–30)
1. The frequency of feeling upset during the past month
2. The frequency of feeling nervous during the past month
3. The frequency of feeling restless during the past month
4. The frequency of feeling hopeless during the past month
5. The frequency of feeling doing anything is difficult during the past month
6. The frequency of feeling life is meaningless during the past month
Answer values for each question are 1 (almost every day) to 5 (never)

A.2. Diagnostic tests of Lewbel IV estimation.

Rosenberg self-esteem Locus of control Emotion stability

Pagan-Hall test
Respectful 16.645 (0.216) 10.647 (0.640) 36.226 (0.001)
Disciplinary 17.380(0.183) 8.400(0.817) 31.889 (0.002)
Material-intensive 16.869(0.205) 8.851 (0.784) 33.460 (0.002)
Time-intensive 16.474(0.225) 9.435(0.739) 33.030(0.002)
White test
Respectful 1.560(0.211) 2.550(0.099) 0.120(0.728)
Disciplinary 4.430(0.035) 8.030(0.005) 8.230(0.004)
Material-intensive 3.129(0.055) 1.480(0.223) 3.380(0.066)
Time-intensive 103.22(0.00) 96.110(0.000) 93.830(0.000)

13
L. Deng and T. Tong China Economic Review 62 (2020) 101477

Over-identification test
Respectful 14.380(0.277) 8.768(0.722) 9.666(0.645)
Disciplinary 11.959(0.449) 12.775(0.386) 9.946(0.621)
Material-intensive 13.078 0.271) 7.411(0.829) 12.008(0.445)
Time-intensive 10.467(0.575) 16.741(0.159) 10.090(0.608)

Note: Based on the Lewbel IV estimation, we adopt Pagan-Hall Test to examine the homoscedasticity assumption of the residuals in Eq. (5).
Similarly, we use White test to examine the heteroscedasticity assumption of the residuals in Eq. (6). The over-identification test is used to test the
validity of constructed Lewbel IVs.

References

51JOB. A survey on Chinese families' educational investment in children. (2019). Available at: https://www.sohu.com/a/317016925_120098702 (in Chinese) .
Alesina, A., Harnoss, J., & Rapoport, H. (2016). Birthplace diversity and economic prosperity. Journal of Economic Growth, 21(2), 101–138.
Altonji, Joseph G., Todd, E. Elder, & Christopher, R. Taber (2005). Selection on observffed and unobserved variables: Assessing the effectiveness of catholic schools.
Journal of Political Economy, 113(1), 151–184.
Aunola, K., Stattin, H., & Nurmi, J. E. (2000). Parenting styles and adolescents achievement strategies. Journal of Adolescence, 23(2), 0–222.
Baum, C. F., & Lewbel, A. (2019). Advice on using heteroscedasticity based identification. Boston College Working Papers in Economics.
Baum, C. F., Schaffer, M. E., & Stillman, S. (2003). Instrumental variables and gmm: Estimation and testing. Stata Journal, 3(1), 1–31.
Baumrind, D. (1971). Types of adolescent life-styles. Developmental Psychology Monographs, 4(1), 2–21.
Baumrind, D. (1989). Rearing competent children. In W. Damon (Ed.). Child development today and tomorrow. Jossey-Bass: San Francisco.
Baumrind, D. (1991). The influence of parenting style on adolescent competence and substance use. Journal of Early Adolescence, 11, 56–95.
Bengtson, V. L. (2001). Beyond the nuclear family: The increasing importance of multigenerational bonds. Journal of Marriage and Family, 63(1), 1–16.
Bernal, R. (2008). The effect of maternal employment and childcare on Children’s cognitive development. International Economic Review, 49(4), 1173–1209.
Bernal, R., & Keane, M. P. (2011). Childcare choices and children’s cognitive achievement: The case of single mothers. Journal of Labor Economics, 29(3), 459–512.
Black, S. E., Grönqvist, E., & Öckert, B. (2018). Born to lead? The effect of birth order on noncognitive abilities. Review of Economics and Statistics, 100(2), 274–286.
Blau, D. M. (1999). The effect of income on child development. Review of Economics and Statistics, 81, 261–276.
Bosquet, C., Combes, P. P., & García Peñalosa, C. (2018). Gender and promotions: Evidence from academic economists in France. The Scandinavian Journal of
Economics.. https://doi.org/10.1111/sjoe.12300.
Brown, P. H., & Park, A. (2002). Education and poverty in rural China. Economics of Education Review, 21(6), 523–541.
Burton, P., Phipps, S., & C. L. (2002). All in the family: A simultaneous model of parenting style and child conduct. American Economic Association Papers and
Proceedings, 92(2), 368–372.
Caldwell, B. W., & Bradley, R. H. (1979). Home observation for measurement of the environment. Little Rock: University of Arkansas at Little Rock.
Chan, T. W., & Anita, K. (2011). Parenting style and youth outcomes in the UK. European Sociological Review, 27(3), 385–399.
Chan, T. W., & Boliver, V. (2013). The grandparents effect in social mobility: Evidence from British birth cohort studies. American Sociological Review, 78(4), 662–678.
Chang, F., Jiang, Y., Loyalka, P., Chu, J., Shi, Y., Osborn, A., & Rozelle, S. (2019). Parental migration, educational achievement, and mental health of junior high school
students in rural China. China Economic Review, 54, 337–349.
Chen, F., Lin, G., & Mair, C. A. (2011). Intergenerational ties in context: Grandparents caring for grandchildren in China. Social Forces, 90, 571–594.
Cobb-Clark, D. A., Salamanca, N., & Zhu, A. (2019). Parenting style as an investment in human development. Journal of Population Economics, 32(4), 1315–1352.
Connelly, R., & Zheng, Z. (2003). Determinants of school enrollment and completion of 10 to 18 years old in China. Economics of Education Review, 22(4), 379–388.
Cosconati, M. (2009). Parenting style and the development of human capital in children. Bank of Italy: Unpublished Manuscript.
Cunha, F., & Heckman, J. J. (2007). The Technology of Skill Formation. American Economic Review, 97(2), 31–47.
Cunha, F., & Heckman, J. J. (2008). Formulating and estimating the technology of cognitive and non-cognitive skill formation. Journal of Human Resources, 43(4),
738–782.
Cunha, F., & Heckman, J. J. (2009). The economics and psychology of inequality and human development. Journal of the European Economic Association, 7(3), 320–364.
Cunha, F., Heckman, J. J., Lochner, L., & Masterov, D. V. (2006). Interpreting the evidence on life cycle skill formation. In E. A. Hanushek, & F. Welch (Eds.). Handbook
of the economics of education (pp. 697–812). Amsterdam: Elsevier 1(12).
Cunha, F., Heckman, J. J., & Schennach, S. M. (2010). Estimating the technology of cognitive and noncognitive skill formation. Econometrica, 78(3), 883–931.
De Man, A. F., LeDuc, C. P., & Labreche-Gauthier, L. (1992). Parental control in child rearing and multidimensional locus of control. Psychological Reports, 70, 320–322.
Dooley, M., & Jennifer, S. (2007). Family income, parenting styles and child behavioral–emotional outcomes. Health Economics, 16(2), 145–162.
Dornbusch, S. M., Ritter, P. L., Leiderman, P. H., Roberts, D. F., & Fraleigh, M. J. (1987). The relation of parenting style to adolescent school performance. Child
Development, 58(5), 1244–1257.
Emran, M. S., & Hou, Z. (2013). Access to markets and rural poverty: Evidence from household consumption in China. Review of Economics and Statistics, 95(2),
682–697.
Eren, O., & Ozbeklik, S. (2013). The effect of noncognitive ability on the earnings of young men: A distributional analysis with measurement error correction. Labor
Economics, 24, 293–304.
Fan, C. S., Wei, X., & Zhang, J. (2017). Soft skills, hard skills, and the black/white wage gap. Economic Inquiry, 55, 1032–1053.
Fiorini, M., & Keane, M. (2014). How the allocation of children’s time affects cognitive and non-cognitive development. Journal of Labor Economics, 32(4), 787–836.
Gehring, K., & Schneider, S. A. (2018). Towards the greater good? EU Commissioners’ nationality and budget allocation in the European Union. American Economic
Journal: Economic Policy, 10(1), 214–239.
Ginsburg, G. S., & Bronstein, P. (1993). Family factors related to children’s intrinsic/extrinsic motivational orientation and academic performance. Child Development,
64, 1461–1474.
Graham, E., & Jordan, L. P. (2011). Migrant parents and the psychological well-being of left-behind children in Southeast Asia. Journal of Marriage and Family, 73(4),
763–787.
Grönqvist, E., & Lindqvist, E. (2016). The making of a manager: Evidence from military officer training. Journal of Labor Economics, 34(4), 869–898.
Hart, C. H., Newell, L. D., & Olsen, S. F. (2003). Parenting skills and social-communicative competence in childhood, chap. 19: 753–797. Handbook of Communication
and Social Interaction Skills. USA: Routledge.
Heckman, J. J., & Rubinstein, Y. (2001). The importance of noncognitive skills: Lessons from the GED testing program. The American Economic Review, 92(1), 145–149.
Heckman, J. J., Stixrud, J., & Urzua, S. (2006). The effects of cognitive and noncognitive abilities on labor market outcome and social behavior. Journal of Labor
Economics, 24(3), 411–482.
Jæger, M. M. (2012). The extended family and children’s educational success. American Sociological Review, 77, 903–922.
James-Burdumy, S. (2005). The effect of maternal labor force participation on child development. Journal of Labor Economics, 23(1), 177–211.
Khanam, R., & Nghiem, S. (2016). Family income and child cognitive and noncognitive development in Australia: Does money matter? Demography, 53(3), 597–621.
Kuhn, P., & Weinberger, C. (2005). Leadership skills and wages. Journal of Labor Economics, 23, 395–436.
Lewbel, A. (2012). Using heteroscedasticity to identify and estimate mismeasured and endogenous regressor models. Journal of Business & Economic Statistics, 30(1),
67–80.

14
L. Deng and T. Tong China Economic Review 62 (2020) 101477

Lindqvist, E., & Vestman, R. (2011). The labor market returns to cognitive and noncognitive ability: Evidence from the Swedish enlistment. American Economic Journal:
Applied Economics, 3(1), 101–128.
Loken, K., Mogstad, M., & Wiswall, M. (2012). What linear estimators miss: Re- examining the effects of family income on child outcomes. Applied Economics, 4, 1–35.
Lu, Y. (2012). Education of children left behind in rural China. Journal of Marriage and Family, 74(2), 328–341.
Maccoby, E. E., & Martin, J. A. (1983). Socialization in the context of the family: Parent-child interaction. In P. H. Mussen (Vol. Ed.), Handbook of child psychology. Vol.
4. Handbook of child psychology (pp. 1–101). New York: \.
Meng, X., & Yamauchi, C. (2017). Children of migrants: The cumulative impact of parental migration on children’s education and health outcomes in China.
Demography, 54(5), 1677–1714.
Meyerhoefer, C. D., & Chen, C. J. (2011). The effect of parental labor migration on Children’s educational progress in rural China. Review of Economic Household, 9(3),
379–396.
NICHD Early Child Care Research Network, & Duncan, G. J. (2003). Modeling the impacts of child care quality on children’s preschool cognitive development. Child
Development, 74, 1454–1475.
Oster, E. (2017). Unobservable selection and coefficient stability: Theory and validation. Journal of Business & Economic Statistics, 1–18.
Pagan, A. R., & Hall, A. D. (1983). Diagnostic tests as residual analysis. Econometric Reviews, 2(2), 159–218.
Paulson, S. E. (1994). Relations of parenting style and parental involvement with ninth-grade students’ achievement. Journal of Early Adolescence, 14, 250–267.
Pulkkinen, L. (1982). Self-control and continuity from childhood to adolescence. In P. B. Baltes, (Vol. Ed.), Life-span development and behaviour. Vol. 4. Life-span
development and behaviour (pp. 63–105). New York: Academic Press.
Rosenberg, M. (1965). Society and the adolescent self-image. Princeton. NJ: Princeton University Press.
Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs: General and Applied, 80(1), 1–28.
Sabia, J. J. (2007). The effect of body weight on adolescent academic performance. Southern Economic Journal, 871–900.
Spera, C. (2005). A review of the relationship among parenting practices, parenting styles, and adolescent school achievement. Educational Psychology Review, 17(2),
125–146.
Todd, P. E., & Wolpin, K. I. (2003). On the specification and estimation of the production function for cognitive achievement. The Economic Journal, 113(485), F3–F33.
Todd, P. E., & Wolpin, K. I. (2007). The production of cognitive achievement in children: Home, school, and racial test score gaps. Journal of Human Capital, 1(1),
91–136.
Trusty, J., & Lampe, R. E. (1997). Relationship of high-school seniors’ perceptions of parental involvement and control to seniors’ locus of control. Journal of Counseling
& Development, 75(5), 375–384.
Weiss, L. H., & Schwarz, J. C. (1996). The relationship between parenting types and older adolescents’ personality, academic achievement, adjustment, and substance
use. Child Development, 67(5), 2101–2114.
White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48, 817–838.
Zeng, Z., & Xie, Y. (2014). The effects of grandparents on children’s schooling: Evidence from rural China. Demography, 51(2), 599–617.
Zhang, H. L., Behrman, J. R., Fan, C., Wei, X. D., & Zhang, J. S. (2014). Does parental absence reduce cognitive achievements? Evidence from rural China. Journal of
Development Economics, 111, 181–195.
Zhao, Q., Yu, X. H., Wang, X. B., & Glauben, T. (2014). The impact of parental migration on Children’s school performance in rural China. China Economic Review, 31,
43–54.

15

You might also like