Gender differences in early childhood development in rural China: a sibling structure perspective

Abstract

Background

This study examines gender differences in early childhood cognitive development in rural China, focusing on the role of sibling structure. While gender disparities have narrowed in recent decades, concerns remain regarding unequal household resource allocation in low- and middle-income countries, particularly in contexts shaped by traditional son preference.

Methods

Data from 1,320 children aged 3 to 7 years across 11 nationally designated poverty counties in the Qinling Mountain region of western China were analyzed. Cognitive ability was assessed using the Chinese version of the Wechsler Preschool and Primary Scale of Intelligence-Fourth Edition (WPPSI-IV). Descriptive statistics and multivariate regression models were employed to investigate the associations between sibling structure and cognitive outcomes by gender.

Results

A negative association was observed between the number of siblings and cognitive scores, with a substantially larger effect for girls. Gender disparities were also evident across sibling composition and age spacing: girls with older siblings, especially those within a three-year age gap, exhibited lower cognitive scores than boys. With respect to birth order, both boys and girls demonstrated first-born and last-born advantages, though the cognitive benefits were less pronounced for girls. These disparities are likely shaped by resource dilution, reduced parental investment, and lower educational expectations for girls.

Conclusions

Gender differences in early cognitive development persist in rural western China and appear to be significantly influenced by family structure. Traditional norms may continue to affect intra-household resource allocation, often to the detriment of girls. These findings underscore the need for gender-sensitive policy interventions aimed at promoting equitable early childhood development and supporting long-term human capital accumulation.

Trial registration

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-025-23599-y.

Highlights

• Girls face cognitive disadvantages shaped by sibling number, birth order, and gender composition.

• Resource dilution and cultural son preference contribute to gender disparities in cognition.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-025-23599-y.

Background

Gender equality in career and life has always been an important topic at the international level. Differences in the investment and development between boys and girls, in terms of both health and education, have been the focus of many researchers, especially in low- and middle-income countries (LICs and MICs). Empirical evidence has shown that poor, rural households in LICs and MICs prioritize providing for their sons’ health and education over their girls [36, 39, 52, 75]. Specifically, in China, traditional Chinese society has a strong “boy preference,” it is possible that families with numerous children may allocate more financial and emotional support to their sons’ schooling, thereby extending the gender gap [13, 46, 69].

In recent years, China’s government has taken steps to address the disparities. More than three decades of economic reforms, starting in 1980, have lifted China from a low-income country to a middle-income country [14]. China’s government has carried out a series of reforms to improve the quality of education for children in the past years. Along with a gradual increase in education level for all, the educational gender gap continues to narrow [18, 30, 74, 88].

However, evidence suggests that the gender gap in both health and education remains, particularly among rural children. In rural regions of China, there exists a notable disparity between the general physical health of girls and boys. Girls in these areas experience a disproportionate burden of specific health concerns when compared to boys, including but not limited to myopia [70], anemia [56], malnutrition [85], and being underweight [31, 48, 57]. And it is observed that boys absorb more education resources, so in a credit-constrained family in China, an increase in the fraction of siblings who are sisters frees up resources for educating boys [45]. In addition, some studies of Chinese samples have found that girls’ cognitive levels are significantly lower than boys’, especially in rural areas [47, 49].

Simultaneously, several researchers have argued that although girls seem to perform better than boys overall (Bibler, 2020; Baker & Milligan, 2016; Sakata et al., 2022), they are actuarily at a relative disadvantage when taking the sibling size and structure into consideration, suggesting that gender differences are more subtle with changing in family and sibling structure as a result of changes in fertility policies. When considering the quantity and structure of siblings, such as girls from large families or those with younger brothers, girls are frequently at a disadvantage [45, 80].

Several theoretical frameworks explain these associations causally. Resource dilution means that as the number of children increases, the household resources that facilitate cognitive development and educational opportunity (such as parent–child interaction and financial resources) dedicated to each child decrease [20, 40, 68]. The quantity-quality trade-off model treats the number of children in a family and parental investment per child as choice variables that respond to economic forces, acknowledging that parents have to make trade-offs between the quantity and quality of their children due to limited resources [2, 61].

Importantly, these mechanisms may affect boys and girls differently in households with limited resources. In contexts where son preference persists, parents may allocate more cognitive stimulation, time, and educational resources to boys than to girls. Consequently, girls in larger families may experience a dual disadvantage—receiving fewer total resources and facing lower prioritization in intrahousehold investment. This highlights the importance of exploring how sibling structure interacts with gender to shape early cognitive development.

A number of studies suggest that son preference remains a key driver of gender disparities in rural China, influencing how families allocate financial and emotional resources to their children [21, 50, 69]. In resource-constrained households, boys are often prioritized in caregiving and educational investment. This tendency is particularly pronounced in families with low socioeconomic status or those with absent parents, where girls face additional vulnerability to neglect and underinvestment [23, 30, 64].

Currently, in order to adapt to China’s aging and other demographic issues, China has gradually relaxed its one-child policy. In 2013, China launched the implementation of the policy that couples in which one of the parties is only child may have two children. In 2016, China fully implemented the policy that a couple may have two children. In 2021, China implemented the policy that a couple may have three children and the accompanying supportive measures. With the gradual relaxation of fertility policies and the rise in larger families, the presence of multiple siblings introduces competition for access to educational resources within the family. Consequently, the influence of sibling number and gender composition on the allocation of educational investments within families and the attainment of intergenerational equity has garnered increased attention from scholars [51, 83]. Existing studies have found clues to gender differences in terms of sibling number, sibling spacing, birth order, gender composition, etc. For example, Wu [73] found a larger detrimental effect of sibship size tended to be observed on females than on males [73]; Choi & Hwang [16] found that parents expected their children to receive slightly more education and to work in higher-income professions when their first-born child was male [16].

In addressing gendered educational investment, early childhood development (ECD) among children in rural China is of particular concern. Early childhood development (ECD) is central to the future of low- and middle-income countries (LMICs). Early childhood development is an investment with long-term returns for both individuals and nations [3, 4, 29, 32, 33, 41]. Although China is one of the most rapidly developing nations in the world, it is still considered an LMIC [28]. More than 70% of young children are born and raised in rural communities with living standards comparable to those of LMICs [58, 59].

Recent research indicates that infants and children in rural China are at risk for cognitive delay, not reaching their maximum developmental potential due to risk factors in the early home environment and the absence of stimulating parenting practices [24]. Parental investments in the home environment during this early stage of life have been identified as key inputs for early childhood development (ECD) and physical growth outcomes in the short term and improved adult human capital outcomes in the long term [17, 34]. Sustainable Development Target 4.2 of the United Nations’ Sustainable Development Goals highlights the need to give more children a fair chance in life by ensuring that by 2030 “all girls and boys have access to quality early childhood development, care, and preprimary education so that they are ready for primary education”. Further understanding of the role of parental investments, particularly the gendered investment strategy among rural areas, is essential for formulating human capital policies that seek to enhance social mobility and economic development.

Few studies have examined gender differences in cognitive development and parenting investments in early childhood, and even less attention has been paid to sibling structure. To the best of our knowledge, no studies have examined whether or how sibling structure may affect the cognitive development and parenting behaviours of girls and boys during early childhood periods in rural areas. Therefore, it is imperative to reexamine the association of sibling structure with the allocation of educational resources within families, especially the gender differences in educational investment. It may also help to provide evidence of gender bias in early childhood.

In this study, we focus on gender differences in cognitive development rather than academic achievement, as measured by standardized cognitive assessment tools. This study draws on data from an in-the-field randomized controlled experiment of an intervention providing parenting support to caregivers to meet these objectives. The main results of the program are reported by [54, 55], who examined the program’s impacts on parenting behaviour and children’s development [55]. While a prior study using baseline data from this project [80] explored early developmental outcomes among infants aged 6–30 months, our study builds on this by employing follow-up data on the same cohort at preschool age (3–7 years), using a different cognitive assessment tool (WPPSI-IV), and extending the analysis to examine more nuanced sibling structures including birth order, age spacing, and sibling gender configuration. The objective of this study is to comprehensively investigate gender differences in early childhood development by focusing on the influence of sibling structure. The research aims to achieve three primary goals. First, to identify whether gender differences exist during the early stages of children’s development. Secondly, to delve deeper into the topic by analyzing the influence of sibling structure on early childhood development. Factors such as the number of siblings (including only-child status), birth order, age spacing between siblings, and the gender composition of siblings will be examined. Lastly, the potential mechanisms behind these disparities will be investigated by examining the role of family resource allocation and parenting behaviours.

The remainder of the paper is organized as follows. Sect."Method"describes the sample selection, data collection, and statistical approach. Sect."Results"presents the results. Sect."Discussion"discusses the implications of the results. Sect."Conclusions"concludes. Section 6 provides a list of abbreviations.

Method

Sample selection

We conducted our study in 11 nationally designated poverty counties located in the Qinling Mountain Region. The study area primarily comprises the Han Chinese population. In 2013, the average per capita annual income was approximately 6,000 RMB, which is lower than the national average of 8,896 RMB for rural areas in the same year (Shaanxi Provincial Bureau of Statistics, 2014).

For sample selection, we employed a multistage cluster sampling design. Initially, all townships (the middle level of administration between county and village) in each of the 11 counties were selected to participate in the study, except for two categories: townships with small villages (less than 800 residents per village) and the township in each county that hosted the county seat, which is typically more prosperous than the average rural township. These exclusion criteria were implemented to ensure a rural sample and enhance the likelihood of including a sufficient number of children aged 6–11 months. A total of 174 townships met these criteria and were included in the study.

To compile the list of villages in each township, we utilized official government data. From this list, we randomly selected two villages per township in April 2013. Subsequently, we obtained a roster of all registered births in the past 12 months from the local family planning official in each village. In cases where a village had fewer than five children within the specified age range, we randomly selected an additional village within the same township to ensure a total of ten child-caregiver pairs per township. Additional villages were randomly chosen in select townships to meet power requirements. In October 2013, a second cohort of children within our target age range (6–11 months) was enrolled from the same sample villages. Our final baseline sample consisted of 1802 children across 351 villages in 174 townships.

During the period of 2020–2021, the project team conducted a follow-up study on all baseline children and their primary caregivers, with a sample age range of 3–7 years old and a total of 1,320 samples after deleting invalid data (73% of the original cohort). While some attrition occurred between baseline and follow-up, the inclusion of rich baseline covariates helps address concerns about sample representativeness. The data collection included the level of cognitive development of the sample children, as well as their basic demographic characteristics and those of their families.

The data for this study were derived from a randomized controlled trial of an early intervention program for children aged 0–3 years. The results of this study have been published [67], and multiple studies have used this dataset to explore the topic of early childhood development [78–80].

Data collection

Questionnaires

All of the surveys included in this study followed uniform data collection protocols. The data collectors were master’s degree holders in related fields, and all the researchers were trained and passed the assessment before they were allowed to collect the data. There were three types of researchers: Wechsler (one-on-one administration of the WPPSI-IV to the sample children), Questionnaire Worker (one-on-one interview with the primary caregiver using an electronic questionnaire), and Quality Control Worker (quality control of all questionnaires on the day of the survey). In each participating household, the primary caregiver was self-identified as the individual most responsible for the child’s care (typically the child’s mother or grandmother).

A series of steps were carried out to safeguard the quality of data collection. Before the survey, the research team conducted a unified training and assessment for all researchers to ensure the professionalism and standardization of all researchers in the process of administering the survey. During the research process, any questionnaire had to go through “four checks.” The “first check” is a self-check by the researcher after completing the questionnaire, the “second check” is a mutual check by the researchers, the “third check” is a check by the research leader, and the “fourth check” is a check by the quality controller through audio and video recording. Four verifications were made to ensure the accuracy and reliability of the data in this study.

Full-scale intelligence quotient score

To assess the developmental outcomes of the sample children, this study used the Chinese version of the Wechsler Preschool & Primary Scale of Intelligence-Fourth Edition (WPPSI-IV) to measure the cognitive developmental level of the sample children. The Chinese version of the WPPSI-IV was adapted to China’s linguistic environment and cultural practices, and the scale norms for urban and rural areas were developed in 2004 [84]. The WPPSI-IV, having undergone translation and adaptation to align with the Chinese context, has been meticulously revised to accommodate language and environmental specifics relevant to China. Its applicability has been demonstrated across diverse regions in various studies [15, 35]. The WPPSI-IV can be used to assess general intellectual development in children between the ages of 2 years, 6 months, and 7 years, 7 months. This individually administered test comprehensively assesses various domains of intelligence, including verbal comprehension, perceptual reasoning, working memory, and processing speed [27]. The combination of these assessments yields four scores, which are then combined to calculate a Full Scale IQ (FSIQ) score—an inclusive measure of overall cognitive functioning. For all normative samples, the reliability of the total IQ was 0.96, the overall reliability of the scale ranged from 0.85 to 0.94, and the validity ranged from 0.52 to 0.75, with the reliability and validity meeting the standards of the original U.S. version. Since the children in the sample had already entered kindergarten during the preschool years, the WPPSI-IV was administered in a kindergarten setting or in the children’s homes. In order to ensure the accuracy of the data collection results, the project team has made relevant requirements for the administration environment. Based on previous Chinese studies, cognitive raw scores below 85 are considered at risk of lagging [25].

In addition to the WPPSI-IV assessment, we also included a control variable for children’s baseline cognitive ability measured during infancy. Specifically, all children in the study were administered the Bayley Scales of Infant and Toddler Development (BSID) Version I when they were between 6 and 12 months old. The BSID is an internationally recognized tool for assessing cognitive and motor development in infants and toddlers (Bayley, 1993) and has been widely used in early childhood research, including studies in China [7, 89]. The version used in this study was formally adapted to the Chinese context in 1992 and normed based on an urban Chinese population [81].

The assessment was administered in the household using a standardized set of materials and scoring guidelines. The BSID takes into account the child’s age in days and prematurity status. While the BSID is primarily designed to assess early development and is not intended as a predictor of later intelligence, we use it here as a baseline control variable to account for pre-existing differences in early cognitive development prior to preschool age. We acknowledge that the BSID may underestimate developmental delays in some cases [5], and we note this limitation in the discussion.

Sibling structure

The article considered the number of siblings, birth order, gender composition, and age spacing, which were categorized as elements of sibling structure [38, 66]. The number of siblings is the number of siblings in the family except the subject. Birth order is the subject’s ranking among siblings, which in this study was split into dummy variables identifying the oldest of siblings, the youngest of siblings, and the middle of siblings. Gender composition includes whether they have older sisters, younger sisters, older brothers, and younger brothers. Referring to the existing literature (Zhang KZ, Tao DJ, & Jiang QC, 2013), having siblings spaced within three years of age may increase sibling rivalry, and we used the presence or absence of siblings spaced within a three-year age interval to examine the effect of age spacing. The third follow-up study collected basic information about the family members, from which the child’s sibling structural variables were obtained.

Child and family characteristics

Teams of trained enumerators administered a one-on-one structured survey to collect information on the socio-demographic characteristics of each child and their family. This includes child-level variables such as age at the time of data collection (in months, based on the birth certificate), gender, whether the child was born prematurely, and the standardized Bayley Scales of Infant and Toddler Development (BSID) score recorded at baseline. We also collected caregiver-level variables such as the mother’s age, whether the mother was the primary caregiver, the education level of the mother and that of the primary caregiver (both coded as a binary variable: 1 = high school or above, 0 = otherwise).

To capture household socioeconomic status, we constructed a household asset index using polychoric principal component analysis [71, 76, 80] based on the presence of ten durable goods: tap water, toilet, water heater, washing machine, computer, internet access, refrigerator, air conditioner, motorcycle or scooter, and car. The asset index was then categorized into three levels (1 = low, 2 = medium, 3 = high) based on its distribution. Descriptive statistics and definitions of all variables are summarized in Table 1.

Table 1.

Summary statistics

Var Name	Description	N	Mean	SD
FSIQ	Full-Scale Intelligence Quotient (FSIQ) score	1320	95.095	10.509
Female	Females are assigned a value of 1, and males 0	1320	0.491	0.500
Sibling structure
Number of siblings	The number of siblings of the child	1320	0.861	0.613
Only child	Equals 1 if the student is an only child; 0 otherwise	1320	0.256	0.437
Have older brothers	Children with older brothers are assigned a value of 1, otherwise 0	1320	0.237	0.425
Have older sisters	Children with older sisters are assigned a value of 1, otherwise 0	1320	0.287	0.453
Have younger brothers	Children with younger brothers are assigned a value of 1, otherwise 0	1320	0.155	.362
Have younger sisters	Children with younger sisters are assigned a value of 1, otherwise 0	1320	0.135	0.342
Have older brothers within three years	Children with older brothers within three years of age are assigned a value of 1, 2 = have older brother over three years, otherwise 0	1320	0.055	0.227
Have older sisters within three years	Children with older sisters within three years of age are assigned a value of 1; 2 = have older sisters over three years, otherwise 0	1320	0.090	0.287
Have younger brother within three years	Children with younger brothers within three years of age are assigned a value of 1, 2 = have younger over three years, otherwise 0	1320	0.088	0.283
Have younger sisters within three years	Children with younger sisters within three years of age are assigned a value of 1, 2 = have younger sister over three years, otherwise 0	1320	0.074	0.262
The oldest of siblings	The children who are the eldest are assigned a value of 1, otherwise 0	1320	0.231	0.422
The youngest of siblings	The children who are the youngest are assigned a value of 1, otherwise 0	1320	0.460	0.499
The middle siblings	The children who are in the middle are assigned a value of 1, otherwise 0	1320	0.053	0.224
Child and Family Characteristics
Baseline BSID score	Standardized Bailey cognitive score for baseline	1320	−0.009	0.999
Age (month)	Children’s age at the time of data collection (in months)	1320	66.097	6.316
Mother is the primary caregiver	Whether the mother is the primary caregiver	1320	0.572	0.495
Premature birth	Whether the child was born prematurely	1320	0.040	0.198
Age of mother	Mothers’ age	1320	32.273	4.929
Mother completed high school	Children whose mothers have a high school degree or higher are assigned a value of 1; otherwise, 0	1320	0.198	0.399
The primary caregiver completed high school	Children whose primary caregivers have a high school degree or higher are assigned a value of 1, otherwise 0	1320	0.039	0.195
Family asset	The ordinal categorization of the family asset index into three levels based on an average distribution of assets:1 = low 2 = medium 3 = high. The family asset index is constructed using polychoric principal components on the following variables: tap water, toilet, water heater, washing machine, computer, Internet, refrigerator, air conditioner, motor or electric bicycle, and car	1320	2.043	0.815
Parenting/Educational Investment
Read books or look at picture books with child	The caregivers have read books or looked at picture books with child for the past three days are assigned a value of 1, otherwise 0	1320	0.611	0.488
Sing songs with child	The caregivers who have sung with the child for the past three days are assigned a value of 1, otherwise 0	1320	0.462	0.499
Tell stories to child	The caregivers who have told stories to the child for the past three days are assigned a value of 1, otherwise 0	1320	0.433	0.496
Play with child	The caregivers who have played games with the child for the past three days are assigned a value of 1, otherwise 0	1320	.752	.432
Positive parenting behaviours	Whether the caregivers have three positive parenting behaviours or above	1320	2.258	1.347
Cost of purchasing vitamins	Log transformation of cost of purchasing vitamins for children	1320	2.716	2.704
Cost of enrolling in extracurricular classes	Log transformation of cost of enrolling in extracurricular classes for children	1320	.705	2.085
Cost of purchasing toys	Log transformation of cost of purchasing toys for children	1320	4.227	2.045
Cost of purchasing books	Log transformation of cost of purchasing books for children	1320	3.593	2.04
Investment in material resources	Log transformation of cost of purchasing material resources for children	1320	5.628	1.613
Expected level of education	The level of education expected by the caregiver: 1 = junior high school 2 = high school 3 = Technical secondary school, vocational school, secondary technology 4 = university Specialty 5 = University Specialty	1320	0.904	0.295
Project treatment	Children who received the project treatment are assigned a value of 1, otherwise 0	1320	0.535	0.499

Parenting behaviour, educational investment, and expectations

Parenting behaviour was measured through a caregiver-reported module administered during the same household visit. Caregivers were asked whether they had engaged in four types of positive interactions with their child in the past three days: (1) reading books or looking at picture books, (2) singing songs, (3) telling stories, and (4) playing games. Each activity was coded as a binary variable (1 = yes, 0 = no), and a composite index was created to indicate “positive parenting behaviour” if the caregiver reported engaging in at least three of the four activities. These items are widely used in early childhood development literature to capture cognitive stimulation and responsive parenting in the home learning environment [54, 82]. Although parenting practices may vary across developmental stages, all children in our sample were of preschool age (mean: 66 months, approximately 5.5 years). According to principles of positive parenting, caregivers are expected to maintain age-appropriate interactions across early childhood. While the specific content (e.g., books or play types) may change with age, the underlying intent of these behaviours remains constant. Nevertheless, we acknowledge potential age sensitivity in these measures and note this as a limitation.

To measure household educational investment, we included several indicators of material and expectational inputs into children’s development. Specifically, we collected data on caregiver-reported spending in the following areas: (1) cost of purchasing vitamins, (2) cost of enrolling children in extracurricular classes, (3) cost of purchasing toys, (4) cost of purchasing books, and (5) cost of purchasing other material resources. All cost-related variables were log-transformed to reduce skewness and ensure normality in regression analyses. These items are widely used in the developmental economics literature as proxies for household-level investment in early childhood [10, 44, 60].

In addition to material inputs, we also measured parental expectations through a question on the expected level of education for the child. The response was coded on a 5-point scale: 1 = junior high school, 2 = high school, 3 = technical secondary or vocational school, 4 = university (non-specialty), and 5 = university (specialty). This indicator reflects the long-term educational expectations parents hold for their children and is commonly used in studies on human capital formation [11, 65].

After the research was completed, the sample size used for the third evaluation period in this study was 1,320, based on data cleaning conducted during the project. Multivariate analyses were performed using listwise deletion. Missing values in the control variables were addressed by imputing either baseline values or mean values. Overall, the rate of missing control variables was maintained at around 5 percent, minimizing the potential impact on the results. The specific variables used in this study are described in Table 1.

Statistical approach

First, we examined the differences in the level of cognitive development across different sibling structures and gender. We then conducted multivariate analyses to determine the association between the number of siblings and the level of cognitive development. We ran two regression models. The first included the number of siblings as the main explanatory variable, and the second model had four dummy variables for whether there were one to four siblings as the main explanatory variable. Both models used subsample regressions for boys and girls, respectively. We then used multivariate regression analyses to analyze whether gender differences existed across children with different sibling compositions and birth order. Finally, we examined the potential mechanism of the significant differences among girls and boys in different sibling situations and only children.

To analyze the main interest of this study, we employed the ordinary least squares (OLS) regression model to assess whether there are gender differences in children with different sibling statuses.

Y_ij is the FSIQ score of student i in county j, where female_ij is a dummy variable indicating whether the child is female. sib_ij represents the dummy variables indicating the child’s sibling status. Three levels of variables are included: first, including whether had elder brothers, elder sisters, younger brothers, or younger sisters; second, we also include the presence of elder brothers, elder sisters, younger brothers or younger sisters within three years; finally, including whether the child was the oldest, youngest or middle child in the family. X_ij represents a vector of control variables that would be correlated with cognitive development scores. These control variables include cognitive development level in baseline, whether the mother is the primary caregiver, premature birth, age of mother, mother completed high school, primary caregiver completed high school and family asset. ε_ij is a random error term.

We also included a treatment indicator ("whether or not the child was assigned to receive the original parenting intervention") to account for potential confounding from the parent-focused program. To further assess robustness, we conducted additional analyses incorporating interaction terms between sibling structure and treatment status. These interaction terms were not statistically significant, suggesting that the observed associations are unlikely to be driven by differential exposure to the intervention (the complete results are available upon request). This adjustment strengthens our confidence that the estimated effects reflect the net relationship between sibling structure and cognitive outcomes, independent of treatment exposure.

Based on the existing literature on family factors affecting children’s cognitive development, we explore descriptive patterns that may help illuminate potential pathways underlying the associations between sibling structure and cognitive outcomes. Rather than formally testing mediating mechanisms, we aim to assess whether observed empirical differences are consistent with key theoretical frameworks, including resource dilution, quantity–quality trade-offs, and son preference.

The concept of resource dilution suggests that as the number of children in a household increases, the resources—such as parental time, attention, and financial investment—allocated to each child may decrease [40]. This effect may be intensified when siblings are closely spaced in age, which has been linked to heightened sibling rivalry and competition [8, 26]. Drawing on prior research [12, 22, 87], we incorporated both non-material and material dimensions of parental investment into our descriptive analysis. The non-material investment was captured through a composite index of positive parenting behaviours, including reading books or looking at picture books, singing songs, telling stories, and playing games with the child. Material inputs included caregiver-reported expenditures on vitamins, extracurricular classes, toys, and books.

Following the quantity–quality trade-off model, which posits that parental investment per child may decline as family size increases [61], we also examined how parental educational expectations varied with sibling structure. This indicator offers insight into the expectational dimensions of parental investment. Finally, in light of son preference observed in many rural Chinese households, we conducted gender-disaggregated analyses to explore whether parenting behaviours, material investments, and educational expectations differ by child gender.

Although these analyses provide suggestive insights, we emphasize that we do not formally test mediation pathways. These descriptive comparisons are intended to support the interpretation of the findings within established theoretical frameworks. All analyses were conducted utilizing Stata 15.0 (Stata Corp., Texas, United States). All tests were two-sided, and P < 0.05 was considered statistically significant. All regression models include village fixed effects to account for regional-level unobserved heterogeneity, such as differences in local education policies, cultural norms, or resource availability, that may influence both sibling structure and cognitive outcomes.

Results

A questionnaire survey was conducted with 1320 individuals in the Qinling Mountain Region. The average FSIQ score, which represents the child’s cognitive level of all samples, was 95.10. The mean age at the time of the project was 66.10 months, and 49.1% of these individuals were girls. Within the sample, it was observed that a proportion of 25.6% of the children were identified as only children. Furthermore, it was found that 23.7% of the participants in the study reported having elder brothers, while 28.7% reported having elder sisters. Additionally, 15.5% of the participants reported having younger brothers, and 13.5% reported having younger sisters. Table 1 provides summary statistics of the sample.

Table 2 shows the differences in FSIQ scores between boys and girls with different sibling structures. Among all sample children, there was no significant difference in the average FSIQ score between only children and children with siblings. Further breakdown by gender, we found that the FSIQ of girls with siblings was 2.78 points lower than that of only daughters, while there was no significant change for boys. The gender difference in the impact was statistically significant (Table 2, row 7, columns 3 & 4, p < 0.01& p < 0.05 separately).

Table 2.

Differences in children’s cognitive levels by family structure and gender

		1.All sample	2.Male (N = 674) mean (SD)	3.Female (N = 651) mean (SD)	4.Gender differences (Cohen’s d)
(1)	Only child (N = 338)	95.90(11.58)	94.7(11.32)	97.36(11.77)	2.66(0.23)**
(2)	Have siblings (N = 982)	94.82(10.10)	95.07(10.00)	94.57(10.21)	−0.49(0.05)
(3)	Have elder brother (N = 313)	94.32(10.12)	93.18(9.55)	95.49(10.57)	2.31(0.23)**
(4)	Have elder sister (N = 379)	94.40(10.17)	95.73(10.47)	92.97(9.67)	−2.76(−0.27)***
(5)	Have younger brother (N = 204)	94.22(10.12)	94.70(9.18)	93.90(10.73)	−0.8(−0.08)
(6)	Have younger sister (N = 182)	95.17(9.86)	95.88(10.24)	94.63(9.58)	−1.24(−0.13)
(7)	Diff (2)-(1)(Cohen’s d)	−1.07(0.10)	−0.37(0.04)	−2.78(0.26)***	−3.15(0.30)**
(8)	Diff (3)-(1)(Cohen’s d)	−1.57(−0.14)*	−1.52(−0.14)	−1.87(−0.18)	−0.35(−0.03)
(9)	Diff (4)-(1)(Cohen’s d)	−1.50(−0.14)*	1.03(0.09)	−4.40(−0.41)***	−5.42(0.50)***
(10)	Diff (5)-(1)(Cohen’s d)	−1.68(−0.15)*	0.00(0.00)	−3.46(0.30)**	−3.46(−0.31)*
(11)	Diff (6)-(1)(Cohen’s d)	−0.72(−0.07)	1.18(0.11)	−2.72(−0.25)*	−3.91(−0.36)*

Rows (6)-(11) show the differences between children in Multi-child families with different sibling structures and those who are only children. The last column shows the difference between girls and boys

^* p value < 0.1 (marginally significant)

^**p value < 0.05

^***p value < 0.01

Table 2 also shows the differences between children with various sibling structures and children with only one sibling. First, the average FSIQ score for children with elder brothers was 94.32, which was marginally lower by 1.57 points compared to only children (p < 0.10, row 8, column 1). However, subgroup comparisons by gender reveal that the decline for both girls and boys was not statistically significant. Second, children with elder sisters scored 94.40 on average, also marginally lower than only children by 1.50 points (p < 0.10, Table 2, row 9, column 1). The gender-specific results indicate that girls with elder sisters were 4.40 points worse off than only daughters (p < 0.01, Table 2, row 9, column 3), while no significant difference was found among boys. Third, the FSIQ score for children with younger brothers was marginally lower by 1.68 points than for only children (p < 0.10, Table 2, row 10, column 1). Girls in this group showed a statistically significant 3.46-point drop compared to only daughters (p < 0.05, Table 2, row 10, column 3), while no significant difference was observed for boys. The gender difference in decline was marginally significant (p < 0.10). Finally, for children with younger sisters, the overall FSIQ score was not significantly different from that of only children. However, girls with younger sisters had a marginally lower score by 2.72 points compared to only daughters (p < 0.10, Table 2, row 11, column 3), while the score for boys did not differ from that of only sons.

Notably, Table 2 also shows that only daughters had significantly higher FSIQ scores than only sons, with a mean difference of 2.66 points (p < 0.05, Table 2, row 1, column 4), whereas among children with siblings, the gender difference was not significant. This suggests that in the absence of sibling competition and parental gender bias, girls may outperform boys in cognitive development. When girls are the only children in the household, they receive undivided parental attention and investment, potentially offsetting any structural disadvantages commonly faced by girls in multi-child families.

Furthermore, we have added a supplemental bar chart figure (Appendix Figure A2) that visually summarizes key FSIQ score differences by sibling structure and gender.

Table 3 shows the results of differences in cognitive level between children who had siblings within three years and only children. The overall pattern is similar to that in Table 2, but narrower age spacing appears to have a more pronounced negative association with children’s FSIQ.

Table 3.

Differences in children’s cognitive levels by sibling interval and gender

		1.All sample	2.Male (N = 674) mean (SD)	3.Female (N = 651) mean (SD)	4.Gender differences (Cohen’s d)
(1)	Only child (N = 338)	95.90(11.58)	94.7(11.32)	97.36(11.77)	2.66(0.23)**
(2)	Have siblings (N = 982)	94.82(10.10)	95.07(10.00)	94.57(10.21)	−0.49(0.05)
(3)	Have elder brother within 3 years (N = 72)	93.03(8.67)	93.07(9.13)	92.97(8.10)	−0.1(−0.01)
(4)	Have elder sister within 3 years (N = 119)	93.18(11.14)	95.55(12.10)	91.06(9.84)	−4.49(−0.41)**
(5)	Have younger brother within 3 years (N = 116)	94.35(10.25)	94.69(9.10)	94.14(10.97)	−0.55(−0.05)
(6)	Have younger sister within 3 years (N = 98)	94.34(9.70)	94.57(9.51)	94.38(9.90)	−0.19(−0.02)
(7)	Diff (3)-(1)(Cohen’s d)	−2.87(−0.28)**	−1.63(−0.15)	−4.40(−0.39)*	−2.77(−0.25)
(8)	Diff (4)-(1)(Cohen’s d)	−2.72(−0.24)**	0.85(0.07)	−6.30(−0.56)***	−7.15(0.63)***
(9)	Diff (5)-(1)(Cohen’s d)	−1.54(−0.14)	−0.01(0.00)	−3.22(−0.28)*	−3.21(−0.29)
(10)	Diff (6)-(1)(Cohen’s d)	−1.45(−0.13)	−0.13(−0.01)	−2.98(−0.26)*	−2.85(−0.26)

Rows (7)-(10) show the differences between children in Multi-child families with different sibling structures and those who are only children. The last column shows the difference between girls and boys

^* p value < 0.1 (marginally significant)

^**p value < 0.05

^***p value < 0.01

Among children with elder brothers within 3 years, the average FSIQ score dropped to 93.03, a statistically significant decline of 2.87 points compared to only children (p < 0.05, Table 3, row 7, column 1). For girls in this group, the FSIQ score decreased by 4.40 points (p < 0.01), whereas there was no significant change for boys. The gender difference in decline was marginally significant (p < 0.10).

For children with elder sisters within 3 years, the average FSIQ score dropped to 93.18, a statistically significant 2.72-point decline from only children (p < 0.05, Table 3, row 8, column 1). Among girls, the FSIQ score was 6.30 points lower than that of only daughters (p < 0.01), while boys did not show a significant difference. The gender gap in this case was highly significant, with girls’ decline being 7.15 points larger than that of boys (p < 0.01).

For children with younger brothers within 3 years, the overall FSIQ score was not significantly different from that of only children. However, girls in this group experienced a marginally significant drop of 3.22 points compared to only daughters (p < 0.10, Table 3, row 9, column 3), while boys showed no significant difference.

Similarly, children with younger sisters within 3 years did not show a significant difference from only children in overall FSIQ scores. Girls, however, had a marginally lower FSIQ by 2.98 points relative to only daughters (p < 0.10, Table 3, row 10, column 3), while the FSIQ scores for boys were statistically indistinguishable from only sons.

Table 4 illustrates the disparities in cognitive level between children from multi-child families with different birth orders and those who are only children. When children were the eldest in the family, their FSIQ scores were not significantly different from those of only children. Both boys and girls scored lower on average, but the differences were not statistically significant. However, the gender difference in decline was significant, with girls’ scores dropping 3.45 points more than boys’ (p < 0.05, Table 4, row 6, column 4).

Table 4.

Differences in children’s cognitive levels by birth order and gender

		1.All sample	2.Male (N = 674) mean (SD)	3.Female (N = 651) mean (SD)	4.Gender differences(Cohen’s d)
(1)	Only child (N = 338)	95.90(11.58)	94.7(11.32)	97.36(11.77)	2.66(0.23)**
(2)	Have siblings (N = 987)	94.82(10.10)	95.07(10.00)	94.57(10.21)	−0.49(0.05)
(3)	The eldest of siblings (N = 305)	95.77(9.93)	96.19(9.60)	95.41(10.22)	−0.78(−0.08)
(4)	The youngest of siblings (N = 607)	94.89(10.20)	94.96(10.13)	94.81(10.30)	−0.16(−0.02)
(5)	The middle of the siblings (N = 70)	90.07(8.69)	87.07(8.69)	90.79(9.00)	2.97(0.34)
(6)	Diff (3)-(1)(Cohen’s d)	−0.13(−0.01)	1.5(0.14)	−1.95(−0.18)	−3.45(−0.32)**
(7)	Diff (4)-(1)(Cohen’s d)	−1.01(−0.09)_	0.26(0.03)	−2.57(−0.24)**	−2.82(−0.26)*
(8)	Diff (5)-(1)(Cohen’s d)	−5.83(−0.52)***	−6.88(0.62)**	−6.57(−0.59)***	0.31(0.03)

Rows (6)-(8) show the differences between children in multi-child families with different sibling structures and those who are only children. The last column shows the difference between girls and boys

^* p value < 0.1 (marginally significant)

^**p-value < 0.05

^***p value < 0.01

For children who were the youngest in the family, there was no statistically significant difference in FSIQ scores compared to only children overall. Girls who were the youngest had FSIQ scores that were 2.57 points lower than only daughters (p < 0.05), and the decline was marginally greater than the change in boys (difference = 2.82 points, p < 0.10, Table 4, row 7, column 4).

In contrast, middle children had significantly lower cognitive scores than only children. The average FSIQ score for middle children was 90.07, representing a decline of 5.83 points (p < 0.01). Gender-specific results show that both boys and girls in this group had significantly lower scores—by 6.88 points and 6.57 points, respectively (p < 0.01 for both)—with no significant gender difference in the magnitude of the decline.

The effect of sibling number on cognitive level

We conducted a multivariate regression analysis to examine the relationship between the number of siblings and children’s cognitive development, stratified by gender. As shown in Table 5, the number of siblings was significantly negatively associated with FSIQ scores in the full sample (p < 0.01, Table 5, row 2, column 1). Specifically, having one sibling was associated with a 2.74-point reduction in FSIQ, and having two siblings was associated with a 10.95-point reduction (p < 0.05 and p < 0.01, respectively; Table 5, rows 4 and 5, column 4).

Table 5.

Multivariate associations between gender and number of siblings and cognitive development outcomes

FSIQ score
	(1) All sample	(2) Female	(3) Male	(4) All sample	(5) Female	(6) Male
1. Female	0.114			0.116
	(0.539)			(0.539)
2. Number of siblings	−1.344***	−1.790***	−0.703
	(0.474)	(0.667)	(0.685)
3. Number of siblings = 1				−0.676	−1.955**	0.587
				(0.660)	(0.986)	(0.899)
4.Number of siblings = 2				−2.735**	−2.899**	−2.679*
				(1.073)	(1.468)	(1.609)
5.Number of siblings = 3				−10.946***	−12.254***	−9.128
				(3.327)	(4.128)	(5.754)
6.Number of siblings = 4				4.752	4.727	4.394
				(6.965)	(10.073)	(9.786)
4. Constants	94.267***	92.261***	95.520***	93.528***	90.986***	94.968***
	(4.544)	(6.800)	(6.208)	(4.548)	(6.867)	(6.194)
5. Control	yes	yes	yes	yes	yes	yes
R-sq	0.149	0.180	0.135	0.154	0.186	0.144
N	1320	648	672	1320	648	672

All regressions control for village fixed effects. All regressions control for Age at the time of data collection, Mother is primary caregiver, Premature birth, Age of mother, Mother completed high school, Primary caregiver completed high school, Family asset, and Project treatment

Columns (1) and (4) use the full sample, columns (2) and (5) use the sample of females, and columns (3) and (6) use the sample of males. N is the total number of observations in each regression

Standard errors are in parentheses

The coefficient for “number of siblings = 4” is based on a small subgroup and is not statistically significant. It should be interpreted with caution due to potential sample limitations

^* p value < 0.1 (marginally significant)

^**p value < 0.05

^***p value < 0.01

Gender-specific results further indicate that these effects were concentrated among girls. For female children, the number of siblings was significantly associated with a decrease of 1.79 points in FSIQ (p < 0.01, Table 5, row 2, column 2), whereas no significant effect was found for boys (Table 5, row 2, column 3). Among girls, the estimated reductions in FSIQ associated with having one, two, and three siblings were 1.96, 2.90, and 12.2 points, respectively (p < 0.05 for one and two siblings; p < 0.01 for three siblings; Table 5, column 5). For boys, the FSIQ score declined by 2.68 points when they had two siblings, but this result was marginally significant (p < 0.10, Table 5, row 4, column 6), and the effects for other sibling numbers were not significant. These findings suggest that sibling-related resource constraints may disproportionately affect girls’ cognitive outcomes, although the magnitude and statistical strength of the effects differ by gender and sibling number.

Associations between sibling status and cognitive development

Table 6 presents the results of multivariate regressions examining the associations between sibling status and cognitive development, as well as gender differences in these associations. Overall, the regression results are consistent with the descriptive patterns observed earlier. In particular, having elder sisters was associated with a 3.49-point lower FSIQ score for girls compared to boys (p < 0.05, Table 6, row 5, column 3). All regressions control for village fixed effects. All regressions control for Age at time of surgery (month), Mother is primary caregiver, Premature birth, Age of mother, Mother completed high school, Primary caregiver completed high school, Family asset, and Project treatment. Columns (1) and (2) additionally control for having older sister, having a younger brother, and having younger sister. Columns (3) and (4) additionally control for having older brother, having younger brother, and having younger sister. Columns (5) and (6) additionally control for having older brother, having older sister and having younger sister. Columns (7) and (8) additionally control for having older brother, having older sister and having younger brother Only children serve as the reference group Column (1)(2) uses the sample of only children and children with older brothers, column (3)(4) uses the sample of only children and children with older sisters, column (5)(6) uses the sample of only children and children with younger brothers, and column (7)(8) uses the sample of only children and children with older sisters. N is the total number of observations in each regression Standard errors in parentheses

Table 6.

Multivariate associations between siblings’ status and cognitive level

FSIQ score	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Characteristics	Have older brothers		Have older sisters		Have younger brothers		Have younger sisters
1. Female	2.170*	2.169*	2.023*	2.014*	2.161*	2.163*	2.108*	2.110*
	−1.127	−1.126	−1.111	−1.11	−1.14	−1.142	−1.145	−1.147
Characteristics (Only child. As Reference Group)
2. Characteristics 1(have older brothers/older sisters/younger brothers/younger sisters)	−1.167		1.724		−0.591		1.354
	−1.227		−1.124		−1.447		−1.488
3. Characteristics 2 (have older brothers/older sisters/younger brothers/younger sisters with 4 years)		0.047		1.377		−0.822		0.927
		−1.796		−1.576		−1.775		−1.995
4. Characteristics 3 (have older brothers/older sisters/younger brothers/younger sisters over 4 years)		−1.546		2.061*		−0.295		1.72
		−1.375		−1.235		−1.948		−1.881
5. Female* characteristics 1	−0.919		−3.493**		−2.191		−2.923
	−1.637		−1.567		−1.945		−1.977
6. Female* characteristics 2		−4.814*		−5.297**		−1.997		−2.732
		−2.736		−2.218		−2.341		−2.504
7. Female* characteristics 3		0.137		−2.629		−2.45		−2.942
		−1.749		−1.704		−2.627		−2.645
Constants	90.541***	90.969***	96.133***	97.244***	96.349***	96.442***	93.754***	93.957***
	−6.886	−6.92	−6.205	−6.241	−7.844	−7.88	−7.773	−7.809
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	651	651	717	717	542	542	516	516
R-sq	0.155	0.159	0.166	0.171	0.169	0.169	0.152	0.152

All regressions control for village fixed effects. All regressions control for Age at time of surgery (month), Mother is primary caregiver, Premature birth, Age of mother, Mother completed high school, Primary caregiver completed high school, Family asset, and Project treatment. Columns (1) and (2) additionally control for having older sister, having a younger brother, and having younger sister. Columns (3) and (4) additionally control for having older brother, having younger brother, and having younger sister. Columns (5) and (6) additionally control for having older brother, having older sister and having younger sister. Columns (7) and (8) additionally control for having older brother, having older sister and having younger brother

Only children serve as the reference group

Column (1)(2) uses the sample of only children and children with older brothers, column (3)(4) uses the sample of only children and children with older sisters, column (5)(6) uses the sample of only children and children with younger brothers, and column (7)(8) uses the sample of only children and children with older sisters. N is the total number of observations in each regression

Standard errors in parentheses

^* p value < 0.1 (marginally significant)

^**p value < 0.05, ***p value < 0.01

We further examined whether the age spacing between siblings played a role. The results suggest that girls with elder brothers within three years scored 4.81 points lower than boys (p < 0.10, Table 6, row 6, column 2), and girls with elder sisters within three years scored 5.30 points lower (p < 0.05, Table 6, row 6, column 4). In contrast, when the age gap between siblings was wider (i.e., more than three years), the gender difference in FSIQ was no longer statistically significant. This suggests that narrow age spacing may exacerbate gender disparities in early cognitive development, although the strength of the association varies across sibling configurations.

Associations between birth order and cognitive development

Table 7 reports the multivariate associations between birth order and cognitive development, with a focus on gender differences. The results indicate that birth order is related to differential cognitive outcomes by gender. Specifically, being the eldest child was associated with a 3.09-point lower FSIQ score for girls compared to boys (p < 0.10, Table 7, row 3, column 1). Similarly, being the youngest child was associated with a 2.51-point lower score for girls relative to boys (p < 0.10, Table 7, row 3, column 2). These results are marginally significant and suggest that both ends of the birth order hierarchy may be more challenging for girls in terms of cognitive development.

Table 7.

Multivariate associations between gender and the order in family and cognitive development outcomes

FSIQ	(1)	(2)	(3)
characteristics	the oldest	the youngest	middle
1.female	2.175*	2.092*	2.077*
	(1.125)	(1.099)	(1.169)
2.characteristics(the child is the eldest, the youngest and middle of siblings)	1.318	2.527	−1.910
	(3.410)	(1.641)	(8.463)
3.female*characteristics	−3.092*	−2.514*	−0.059
	(1.631)	(1.374)	(3.188)
Constants	96.302***	91.604***	91.737***
	(7.024)	(5.426)	(8.859)
Controls	Yes	Yes	Yes
R-sq	0.139	0.151	0.178
N	643	945	408

All regressions control for village fixed effects. All regressions control for the number of siblings, Age at the time of data collection, Mother is the primary caregiver, Premature birth, Age of mother, Mother completed high school, Primary caregiver completed high school, Family asset, and Project treatment

Only children serve as the reference group

Column (1) includes only children and the eldest child among siblings; Column (2) includes only children and the youngest child; Column (3) includes only children and children in middle birth positions

Standard errors in parentheses

^* p value < 0.1 (marginally significant)

^**p value < 0.05

^***p value < 0.01

Patterns of parenting and investment by sibling structure and gender

This section presents a descriptive examination of family-level characteristics associated with differences in children’s cognitive outcomes across sibling structures in rural China. Drawing on frameworks such as resource dilution, the quantity–quality trade-off, and son preference, the analysis considers whether observed variation in parenting behaviours, material investments, and educational expectations aligns with these conceptual perspectives. No formal mediation analyses are conducted, and the results should be interpreted as suggestive associations rather than definitive mechanisms.

As shown in Table 8, panels A and B, families with multiple children report slightly fewer positive parenting behaviours than families with only one child with marginal statistical significance. In addition, children with siblings receive significantly lower levels of material investment—particularly in books, toys, and extracurricular classes—than only children. These findings are consistent with the effects of resource dilution.

Table 8.

Breakdown of mechanism variables by gender and sibling status

	All sample			Female			Male
Variable	(1) Have siblings (N = 982)	(2) Only child (N = 338)	(3) Mean difference	(4) Have siblings (N = 496)	(5) Only child (N = 152)	(6) Mean difference	(7) Have siblings (N = 486)	(8) Only child (N = 186)	(9) Mean difference
Panel A
Positive parenting behaviours	0.457	0.512	−0.055*	0.474	0.533	−0.059	0.440	0.495	−0.054
	(0.248)	(0.251)		(0.250)	(0.251)		(0.247)	(0.251)
1. Read books or look at picture books with child	0.599	0.645	−0.046	0.619	0.645	−0.026	0.578	0.645	−0.067
	(0.240)	(0.230)		(0.236)	(0.231)		(0.244)	(0.230)
2. Sing songs with child	0.432	0.435	−0.003	0.476	0.480	−0.004	0.387	0.398	−0.011
	(0.246)	(0.246)		(0.250)	(0.251)		(0.238)	(0.241)
3. Tell stories to child	0.450	0.497	−0.047	0.454	0.520	−0.066	0.447	0.478	−0.032
	(0.248)	(0.251)		(0.248)	(0.251)		(0.248)	(0.251)
4. Play with child	0.741	0.784	−0.043	0.740	0.796	−0.056	0.743	0.774	−0.031
	(0.192)	(0.170)		(0.193)	(0.163)		(0.191)	(0.176)
Panel B
Material investment	5.527	5.920	−0.392***	5.496	6.010	−0.514***	5.560	5.846	−0.286**
	(2.582)	(2.547)		(2.831)	(2.128)		(2.331)	(2.890)
1. Cost of purchasing vitamins	2.592	3.078	−0.486***	2.495	3.256	−0.761***	2.691	2.933	−0.242
	(7.219)	(7.434)		(7.189)	(6.789)		(7.246)	(7.954)
2. Cost of enrolling in extracurricular classes	0.615	0.968	−0.353***	0.900	1.387	−0.486**	0.323	0.626	−0.303**
	(3.856)	(5.704)		(5.387)	(7.458)		(2.133)	(4.042)
3. Cost of purchasing toys	4.121	4.536	−0.416***	3.847	4.539	−0.693***	4.400	4.533	−0.133
	(4.248)	(3.876)		(4.558)	(3.568)		(3.785)	(4.148)
4. Cost of purchasing books	3.515	3.820	−0.305**	3.538	3.972	−0.434**	3.492	3.696	−0.204
	(4.109)	(4.260)		(4.090)	(3.515)		(4.136)	(4.857)
Other characteristics
Educational expectations of child	4.147	4.246	−0.099**	4.149	4.263	−0.114*	4.144	4.231	−0.087
	(0.519)	(0.352)		(0.584)	(0.354)		(0.453)	(0.352)

Standard errors are in parentheses

Results indicate that only children receive significantly more material investment than children with siblings, especially among girls. However, differences in positive parenting behaviours are relatively small and mostly insignificant. Additionally, parents of only children tend to hold slightly higher educational expectations for their children, though the magnitude of this difference is modest

Positive parenting behaviours is a dummy variable in which parents are considered to be invested in positive parenting behaviours if they have more than three of the following positive parenting behaviours. Material investment is the sum of all the following material inputs

^* p value < 0.1 (marginally significant)

^**p value < 0.05

^***p value < 0.01

As shown in the last row in Table 8, parental expectations for children’s educational attainment appear to be higher among only children relative to those with siblings. This finding may reflect the classic quantity–quality trade-off, whereby parents with fewer children are able to devote more attention to setting ambitious long-term goals for their child’s education. Although no formal mediation analysis is conducted, the observed association aligns with the theoretical expectation that educational expectations are an integral component of early parental investment strategies.

Gender-specific patterns emerge in how parenting inputs vary by sibling structure as well. Among girls, having siblings is associated with significantly reduced investment in vitamins, extracurricular classes, toys, and books, whereas among boys, significant reductions are observed only for extracurricular classes. Moreover, when disaggregated by gender, the data reveal that this pattern is more pronounced for girls: educational expectations are significantly lower for girls with siblings compared to only daughters, while no such difference is observed among boys.

Taken together, these patterns suggest that sibling structure interacts with gender norms in shaping the allocation of resources and expectations within households. However, these results are descriptive in nature and do not establish causal mediation, highlighting an avenue for further research.

Finally, we provide additional descriptive evidence on differences in sibling structure by gender, which may indirectly reflect traditional son preference. As shown in Table 9, boys are more likely to be only children, have fewer siblings, and be the youngest child in the household. In contrast, girls are more likely to be the oldest or middle child, suggesting that girls are more likely to have younger brothers. This pattern is consistent with previous findings indicating that parents may continue to have children after the birth of a daughter in hopes of having a son [9]. While not definitive, these descriptive differences offer suggestive support for the presence of son preference in family formation decisions.

Table 9.

The difference of sibling status between male and female

Variable	Male	Female	Mean difference
	N = 672	N = 648
Only child	0.277	0.235	0.042*
	(0.200)	(0.180)
Number of siblings	0.818	0.904	−0.086**
	(0.357)	(0.393)
The oldest of siblings	0.207	0.256	−0.049**
	(0.164)	(0.191)
The middle of siblings	0.025	0.082	−0.056***
	(0.025)	(0.075)
The youngest of siblings	0.491	0.427	0.064**
	(0.250)	(0.245)

^* p value < 0.1 (marginally significant)

^**p value < 0.05

^***p value < 0.01

Discussion

This study investigated gender differences in cognitive development and parenting behaviours across various sibling structures—namely, sibling number, birth order, gender composition, and age spacing—using cross-sectional data from rural western China. While some previous studies suggest that girls in only-child families may outperform boys in cognitive development (Bibler, 2020; Baker & Milligan, 2016; Sakata et al., 2022), our findings indicate that this advantage tends to diminish as family size increases. One potential explanation is that, under resource constraints, parents may allocate more attention and investment to sons, particularly in contexts where son preference is culturally embedded [9, 37].

Interestingly, we found that only daughters significantly outperformed only sons in FSIQ scores. This finding suggests that when girls grow up without sibling competition and receive undivided parental investment, they may exhibit stronger cognitive development outcomes than boys. It underscores the role of family context—rather than inherent gender differences—in shaping early cognitive outcomes. This pattern aligns with previous studies showing that girls may have an advantage in early cognitive performance in supportive and low-competition environments (Bibler, 2020; Baker & Milligan, 2016; Sakata et al., 2022), particularly when they are the sole recipients of household attention and stimulation. We found significant gender differences in cognitive development associated with sibling structure. Regarding sibling number, girls’ cognitive development was more negatively affected by having siblings than boys’. This finding was robust across multiple regression models and consistent with previous literature [73, 80].

Regarding sibling composition, girls with siblings- except those with elder brothers- had significantly lower cognitive scores than only daughters. Girls with elder sisters were particularly disadvantaged compared to boys. Notably, the gender gap in cognitive outcomes varied by sibling composition. For example, girls with elder sisters tended to perform worse than girls with elder brothers. This pattern may reflect complex intra-household dynamics in rural Chinese families. One possibility is that when the first-born is a boy, families may perceive the arrival of a daughter as completing the ideal “son-and-daughter” family composition, which may reduce gender bias in subsequent investment [1, 19]. In contrast, girls with elder sisters may face cumulative disadvantage due to both sibling competition and lower perceived marginal returns—especially in families still hoping for a son [53, 68]. In addition, elder sisters are often expected to take on caregiving responsibilities for younger siblings, which may reduce the time and attention available for their own learning and cognitive stimulation [45]. Moreover, narrower age spacing between siblings was particularly detrimental for girls with elder brothers, likely due to intensified competition for limited resources [38, 86].

With respect to birth order, middle children—especially girls—had lower cognitive scores than their first-born and last-born peers, echoing previous studies [43, 77]. First-born children of both genders exhibited cognitive advantages, with stronger effects among boys. This could reflect a temporary period of undivided parental attention or, in the case of girls, additional caregiving responsibilities that may offset birth order advantages [26, 45]. While last-born children also showed higher scores than middle children, the advantage was less pronounced for girls.

To explore potential explanatory pathways, we descriptively examined patterns in positive parenting behaviours, material investments, and educational expectations. The observed results were consistent with the resource dilution hypothesis [72], whereby more children in the household reduce per-child inputs. Similarly, parents of children with siblings reported lower educational expectations than parents of only children, a pattern aligned with the quantity–quality trade-off framework [42]. We also found gender differences in these inputs: girls with siblings received significantly fewer material resources and lower educational expectations than boys, particularly in families with narrow sibling spacing or specific sibling gender compositions.

Additionally, descriptive evidence from sibling structure patterns suggests that boys are more likely to be only children or the youngest child, whereas girls are more often found in older sibling positions. This asymmetry is consistent with previous studies suggesting that families may continue childbearing after the birth of a girl in hopes of having a son [9], indirectly reflecting cultural son preference. While not definitive, these findings provide suggestive evidence that sibling structure and resource allocation may be influenced by gendered fertility preferences.

While this study does not seek to disentangle the relative contributions of biological and social factors, the variation in gender disparities across different sibling structures suggests that environmental influences play a substantial role. In particular, the findings point to the importance of gendered parental expectations and intra-household resource allocation as key drivers of the observed cognitive differences. If biological differences were the primary determinant, we would expect more uniform patterns across sibling configurations. Instead, the gender gap appears most pronounced in contexts associated with greater competition for limited resources, supporting the argument that early cognitive development is highly sensitive to social and familial environments.

Despite these contributions, the study has several limitations. First, the sample is drawn from rural western China, limiting the generalizability of the findings to urban or more economically developed contexts. Second, the cross-sectional nature of the data precludes causal inference. Third, although this study explores several potential explanatory factors, it does not conduct formal mediation or mechanism analyses, and these patterns should be interpreted as suggestive rather than confirmatory. Future studies are needed to examine these pathways more rigorously. A further limitation is the lack of direct measures of parental attitudes or cultural preferences, which may simultaneously influence both family size decisions and children’s development. While we include rich socio-demographic controls and village fixed effects to mitigate this concern, we acknowledge that some residual confounding may persist.

These early cognitive disparities may have implications for children’s longer-term academic and developmental trajectories. Prior research has found that early childhood cognitive development is associated with later educational attainment and labor market outcomes [29, 33]. In this context, the relatively lower cognitive scores observed among girls in larger, resource-constrained households highlight the importance of monitoring potential developmental vulnerabilities and ensuring early support where needed. While our cross-sectional design does not permit conclusions about long-term outcomes, the findings highlight the potential value of early, gender-sensitive support strategies to help promote more equitable opportunities for children from diverse family backgrounds.

These findings raise important questions regarding broader theoretical and policy implications. The observation that only daughters outperform only sons does not necessarily contradict the wider narrative of gender disadvantage in rural China. Instead, it may reflect that, in the absence of sibling competition, girls are better able to realize their developmental potential—highlighting the influence of contextual and structural factors rather than inherent gender differences. This interpretation aligns with international literature emphasizing the role of social environments, family dynamics, and cultural expectations in shaping developmental trajectories [62, 63].

These patterns may also be relevant to the context of China’s evolving fertility landscape. As larger families become more common following the relaxation of birth restrictions, characteristics such as birth order, age spacing, and sibling gender composition may increasingly influence child development outcomes. While our findings do not establish causality, they suggest that greater attention to intra-household dynamics may be warranted in future policy design. In particular, parenting support and early childhood investments could consider family structure and gender composition to help foster more inclusive developmental environments for children across diverse family structures.

Although our study focuses on rural China, the underlying mechanisms—such as resource allocation, gendered investment strategies, and sibling competition—are applicable to other low- and middle-income countries as well. As such, the results contribute to a growing international literature on how family structure intersects with gender to influence early childhood development [36].

Nonetheless, this study contributes in several ways. It provides new empirical evidence on how sibling dynamics interact with gender in shaping early cognitive development. It also highlights the importance of sibling characteristics—composition, age spacing, and birth order—which are often overlooked in early childhood research. Finally, the use of the WPPSI-IV provides a standardized and objective measure of cognitive outcomes, reducing potential bias from caregiver reports and improving the reliability of developmental assessments [6].

Conclusions

Using cross-sectional data from rural western China, this study examined gender disparities in children’s cognitive development within varying sibling structures. The results revealed that having more siblings is negatively associated with cognitive outcomes, particularly for girls. These gendered patterns were most pronounced among girls with elder sisters, girls with elder brothers who were close in age, and girls who were the oldest or middle child in the family.

Our findings offer descriptive support for theoretical frameworks such as resource dilution, quantity–quality trade-offs, and cultural son preference, which may jointly contribute to observed gender disparities in early development. While the results do not establish causality, they highlight the importance of considering sibling dynamics and gendered expectations in designing equitable early childhood interventions. Future research using longitudinal or experimental designs is needed to unpack these pathways further.

From a policy perspective, the findings suggest a need for gender-sensitive support strategies that ensure equitable access to developmental resources for all children. Interventions that target families with multiple children—particularly girls—may be crucial in reducing early disparities and promoting more inclusive human capital development in rural areas.

Furthermore, our findings underscore the urgency of tailoring policy responses to mitigate the cumulative disadvantages faced by girls in multi-child households. Policymakers should consider targeted early childhood support for girls, such as subsidized preschool access, parent training on equitable caregiving, and public awareness campaigns addressing unconscious gender bias in resource allocation. These interventions may be especially effective when integrated into broader strategies aligned with China’s three-child policy, ensuring that expanded family size does not exacerbate gender gaps in cognitive development.

In terms of future research, longitudinal tracking of affected children into adolescence could help reveal how early disadvantages compound over time, while mixed-methods studies may shed light on family dynamics and parental beliefs. Extending these inquiries beyond the Chinese context could also contribute to a global understanding of gendered childhood development.

Abbreviations

ECD

Early childhood development

LICs and MICs

Low- and middle-income countries

Authors’ contributions

H.G. contributed to the conceptualization methodology and drafted the original manuscript. X.C. contributed to the methodology interpretation of the data and drafted the original manuscript. L.Z. contributed to the interpretation of the data and substantively revised the study. Y.Z. contributed to the interpretation of the data and substantively revised the study. Y.D. contributed to the methodology and interpretation of the data. A.Y. designed the study and contributed to substantively revising the study. All authors read and approved the final manuscript.

Funding

This work was supported by the National Social Science Foundation of China (grant number 22BGL212), the 111 Project (Grant No. B16031, 2015), and the HuPan Modou Foundation.

Data availability

The dataset used and/or analyzed during the current study is available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The studies involving humans were approved by the Biological and Medical Ethics Committee, Minzu University of China (ECMUC2020027CO). The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation in this study was provided by the participants’ legal guardians/next of kin.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.