Introduction

Development during the first three years of life postpartum has profound short- and long-term effects on a child’s later school readiness, academic performance, employment, and income (Currie and Almond, 2011; Grantham-McGregor et al., 2007; Jeong et al., 2021; Walker et al., 2007). The literature has shown that delays in either cognitive or non-cognitive development of children before age three significantly increase the likelihood of adverse financial and life outcomes in adulthood (Attanasio et al., 2020; Cunha and Heckman, 2009; Heckman, 2008; Heckman et al., 2010; Knudsen et al., 2006). Importantly, widespread, persisting developmental delays can affect the quality of a country’s labor force and impede the growth and prosperity of the economy (Guo and Qu, 2019; Heckman et al., 2010).

When investigating the pathways of the cognitive and non-cognitive development before age three, previous studies have demonstrated that children often experience dynamic development that are influenced by either or both developmental domains (cognitive and non-cognitive) (Briggs-gowan et al., 2006; Feldman and Eidelman, 2009; Haapsamo et al., 2012; Intusoma et al., 2013; Sansavini et al., 2014). Some children transit from normal to delayed development in either or both domains, while others exhibit improvements in or deterioration of their levels of development (Sansavini et al., 2014). A U.S. study using nationally representative longitudinal data, for instance, identified four development trajectories based on children’s cognitive developmental outcomes during the period of 9 to 24 months old: normally developed, delayed resolved, newly delayed, or persistently delayed. 20% of the sample children that experienced cognitive delays at 9 months of age continued to experience delays at 24 months; 80.3% of the children resolved their developmental delay during this period (turning from delayed to normal); and 85% of children that displayed cognitive delays at 24 months old showed normal levels of cognitive development when they were 9 months old (turning from normal to delayed) (Cheng et al., 2014). Another study examining the pathways of cognitive or non-cognitive development among Finnish children found that non-cognitive development delay decreased from 10% to 5% in the period between 18 and 36 months of age (Haapsamo et al., 2012).

Research on pathways of cognitive or non-cognitive development indicates that children’s later cognitive and non-cognitive development are potentially shaped by different pathways of development in early childhood (Campbell et al., 2006; Cheng et al., 2014; Dekker et al., 2007; Stålnacke et al., 2019; van Beek et al., 2021). Campbell et al. (2006) found that, in a sample of U.S. children, whether a child experienced stable or decreasing non-cognitive development between toddlerhood and 9 years old predicted their non-cognitive development when they reached age 12. Most of these studies, however, focused on the pathways identified in middle childhood or adolescence. Only a few studies have examined how developmental pathways before age three affect later developmental outcomes (Cheng et al., 2014; Wang et al., 2021). The limited existing research on this topic shares the consensus that differences in early childhood developmental pathways affect children’s development at preschool age. Unfortunately, it is not clear how they continue to influence cognitive and non-cognitive development as children reach primary school (i.e., at 9 or 10 years old).

Since different pathways of development at early life have different impacts on later life outcomes, the literature also has identified factors that can influence the different pathways of cognitive or non-cognitive development of children (Cheng et al., 2014; Planalp and Braungart-Rieker, 2015; Wang et al., 2021). According to these studies, children’s cognitive and non-cognitive developmental pathways are correlated with children and households’ certain demographic characteristics, such as the child’s gender and household socioeconomic status (SES). For example, girls were found to be more likely to experience improving developmental outcomes (improving pathways) when their development pathways were compared to those of boys (Cheng et al., 2014). Research also has demonstrated that children from households with lower SES are at higher risk of suffering from persistent developmental delays (Cheng et al., 2014; Wang et al., 2021).

For young children in low- and middle-income countries (LMICs), many of whom frequently experience developmental delays (Engle et al., 2007; Grantham-McGregor et al., 2007), a better understanding of how cognitive and non-cognitive developmental pathways before age three affect developmental outcomes into primary school years has great practical value. According to the findings of Lu et al. (2016), 45 million of all children in LMICs who experience under-development come from rural China. Recent research on young children in the region also revealed concerning levels of cognitive and non-cognitive developmental delays: 39% to 49% of young children experience cognitive developmental delays, and 34% to 58% non-cognitive developmental delays (Emmers et al., 2021; Wang et al., 2019).

A nascent but growing body of literature in recent years has examined pathways of development among children under three years old (Luo et al., 2019) and the role these pathways continue to play in children’s development when they reach preschool (Wang et al., 2021, 2022, 2024). In line with the international studies discussed previously, these works support that differences in early non-cognitive and cognitive developmental pathways similarly affect the subsequent development of children in rural China. These studies, however, only examined the associations between different pathways of either one of the developmental domains (either that of the pathway of cognition or the pathway of non-cognition) and the development of children at preschool age. Importantly, how both the cognitive and the non-cognitive pathways of development concurrently affect the development of those children over the longer-term (i.e., when the children are at primary school age) has not been investigated in the setting of rural China or any other developing country.

The current study

This study aims to investigate the associations between the different pathways of cognitive and non-cognitive development before age three and children’s levels of development at primary school age in rural China. We pursue this objective by addressing the following three specific questions:

  1. a)

    How do children’s pathways of cognitive and non-cognitive development change before the age of three?

  2. b)

    Are different pathways of cognitive and non-cognitive development before age three associated with later child developmental outcomes in primary school?

  3. c)

    What factors influence the different pathways of development of children before the age of three?

In line with suggestions from previous literature (Cheng et al., 2014; Wang et al., 2021, 2022, 2024; Witt et al., 2009), we classified children in our sample into four pathways based on how their cognitive and non-cognitive development trajectories changed between 6–12 months and 22–30 months of age. The four pathways include: “never delayed”— children that were never delayed in either domain of cognitive or non-cognitive development; “persistently delayed”— children that were delayed in either or both domains for the entire study period, or between 6 to 12 months old and 22 to 30 months old; “improving”— children that transited from delayed to normal in either or both domains; and “deteriorating” — children that transited from normal to delayed in either or both domains. We hypothesize that the sample children that were on different pathways when they were below the age of three had different levels of developmental outcomes when they reached primary school age. We also expect that the children’s categorization into the difference pathways is predicted by their own demographic characteristics and those of their families.

Methods

Participants

This study draws on three waves of longitudinal data collected between 2013 and 2022 from children and households selected from 11 nationally designated poverty counties in northwestern China. We followed a multistage cluster sampling design. After the initial baseline in 2013, six follow-up surveys were administered by the research team to sample children and their households. The current study uses data from the first wave (collected in 2013, when the children were between 6 and 12 months old), the fourth wave (collected in 2015, when the children were between 22 and 30 months old), and the seventh wave (collected in 2022, when the children entered primary school or were between 9 and 10 years old). A total of 1,087 pairs of children and their primary caregivers completed all three aforementioned surveys and were included in the current study.

Data collection for this study was first approved by the Stanford University Institutional Review Board (Protocol ID 25734) on October 26, 2012, with approval renewed annually. Before data collection, enumerators obtained informed consent from each caregiver for both their and their child’s participation in the study. The research team also informed participants of the purpose of the study, their rights, and efforts to safeguard their personal information. The principles of the declaration of Helsinki were observed throughout the procedures of the study.

Measures

Child cognitive development

Child cognitive development was assessed through two measures at different survey waves to ensure age-appropriateness. In the first and fourth wave, since the sample children were under 30 months old, the first version of the Bayley Scales of Infant Development (BSID) was used to measure the cognitive skills of each child. Widely regarded as the international benchmark for measuring cognitive skills among children under 30 months old (Bayley, 1974; Hamadani et al., 2010), the BSID has been translated and scaled to adapt to a Chinese population. In the Chinese version of the scale (Yi et al., 1993), the cognitive development measure (Mental Development Index; MDI) has been found to have an expected mean of 100 and a standard deviation (SD) of 16. As a result, we define an MDI score below 84, or one SD below the expected mean, as indicating cognitive developmental delay.

We used the Wechsler Intelligence Scale for Children-Fourth Edition (WISC-IV) to measure cognitive development in the seventh wave when the sample children were at primary school age, because the BSID is less effective for assessing development beyond three years of age. The WISC-IV computes a composite score, called the Full-Scale Intelligence Quotient (FSIQ), which encompasses a child’s cognitive abilities across a diverse set of domains. In line with previous literature (Wechsler, 2008), we categorize children in our sample who scored below 85, or 1 SD below the expected mean, in FSIQ as experiencing cognitive developmental delays. First translated and scaled for use in Chinese populations in 2008 according to a sample from both rural and urban China, the WISC-IV has been recognized internationally as an effective standardized measure for cognitive development among children aged between 6 and 16 years (Wechsler, 2003).

Child non-cognitive development

Similarly, we only used the Ages and Stages Questionnaire: Social-Emotional (ASQ:SE) to measure non-cognitive skills in the first and fourth survey waves, since the ASQ:SE has been validated and widely used to measure non-cognitive skills of children between 3 and 66 months old (Squires et al., 2002). The ASQ:SE has been culturally modified and normalized in China since 2017 (Bian et al., 2017). The primary caregiver, typically the child’s parent or paternal grandmother if the parents were working and living away from home, reported if the child exhibited a series of behaviors “most of the time,” “sometimes,” or “never”. We assigned a score of 0 to behaviors the child displayed most of the time, 5 to behaviors children sometimes displayed, 10 to behaviors children never displayed, and added an extra 5 points for behaviors that were a concern. We computed a total score by summing scores from each item, and a higher total score represents worse levels of non-cognitive development of the sample children. According to age-based cutoff scores set by Squires et al. (2002), we identified children who were experiencing delays in non-cognitive development.

In the seventh wave when the children were 9 to 10 years old, we used the Strengths and Difficulties Questionnaire (SDQ) to assess non-cognitive development. The SDQ is a caregiver-reported questionnaire which produces a Total Difficulties score and a Prosocial Behavior score from 25 items (Goodman, 2001). Caregivers reported the extent to which a series of strengths and difficulties applied to their child (0 = not true, 1 = somewhat true, and 2 = certainly true). In general, higher Total Difficulties scores indicate lower levels of non-cognitive development, while higher Prosocial Behavior scores represent higher development of non-cognitive skill. In 2008, the SDQ sale was adapted to the Chinese context (Du et al., 2008).

Academic performance

Math and reading standardized tests were used to measure child academic performance during the seventh wave of the survey when the children were 9–10 years old. From the Trends in International Mathematics and Science Study (TIMSS) data bank, one of the most common instruments for measuring academic performance in mathematics for primary school students in the world (Mullis, Martin, Foy, et al., 2012) and in China (Zhao et al., 2014), we selected standardized math test items such that all questions were grade-appropriate and within the children’s math curriculum.

Sample children in the seventh wave of the survey were also given a reading test. We selected standardized reading test items from a pool of questions from the International Reading Literacy Study (PIRLS) reading test with the help of local education officials and experts. The PIRLS is a testing scale that provides international measures of reading comprehension that is widely used throughout the world (Mullis, Martin, Minnich, et al., 2012). The research team validated the psychometric properties of the reading tests to ensure good reliability and validity, using data from pilot testing among fourth- and fifth-grade children. Through the pilot tests, we also confirmed that the test materials, time limits, and the difficulty of the questions were appropriate.

Socio-demographic characteristics

We administered a survey on child and household demographic characteristics to all primary caregivers in the sample. Caregivers reported the child’s gender, age (in months), whether the child had siblings, and whether the child was born prematurely. Household characteristics included the primary caregiver’s relationship to the child (e.g., mother or grandmother), maternal age, and educational level. As a proxy of household assets, we recorded whether the household had access to tap water, a toilet with running water, a water heater, a washing machine, a refrigerator, air conditioning, a computer, the Internet, a motorcycle or electric scooter, and a car or truck. Based on ownership of the items, we used polychoric principal component analysis to compute a family assets index.

Procedure

Prior to survey administration in 2013, 2015, and 2022, enumerators recruited from local universities completed a training that lasted a week and included more than two days of field practice. Surveys were administered through visits to individual households. The enumerators introduced the purpose and scope of the study upon arrival at each household and obtained consent for participation of the sampled caregiver-child dyad from the caregiver. The enumerators then administered the tests on the cognitive development (BSID and WISC-IV) and academic performance of the sampled children. When interviewing the primary caregivers, data on non-cognitive development (ASQ:SE and SDQ) and socio-demographic characteristics were collected.

Statistical analysis

Analyses took place in five stages. First, we used chi-squared tests to test the homogeneity of the three waves of data collected in 2013, 2015, and 2022. Second, we presented descriptive statistics of the current study variables. The cutoffs of the developmental delays followed the manuals of the scales (Squires et al., 2002; Wechsler, 2008; Yi et al., 1993). In line with the non-parametric standardization method developed by Rubio-Codina et al. (2016), we separately standardized MDI, FSIQ, ASQ:SE, and SDQ raw scores internally for each survey wave. Third, following the previous studies (Cheng et al., 2014; Wang et al., 2021, 2022; Witt et al., 2009), we identified the trends in different developmental pathways using the data collected between 6 and 12 months of age (the first wave) and between 22 and 30 months of age (the fourth wave). Fourth, linear regression analysis was used to examine whether different developmental pathways predicted children’s cognitive development, non-cognitive development, and academic performance in primary school. Because a child’s age, gender, (Baillargeon et al., 2007) whether they were born prematurely (van Beek et al., 2021), and the SES of their primary caregiver (including maternal age and education, family asset index) (Attanasio et al., 2022; Mazza et al., 2016; Skopek and Passaretta, 2021) have all been found to affect early development, we further conducted logistic regressions to investigate demographic characteristics that predicted a child’s entry into different developmental pathways. In linear regression models for the seventh wave when the children were at primary school age, we also included these factors as control variables, in addition to controlling for levels of cognitive and non-cognitive development measured during the first wave and time fixed effect. All standard errors were clustered at the village level. All statistical analyses in the current study were two-sided and conducted using STATA 16.0. Following the literature (Attanasio et al., 2020; Carneiro et al., 2021; Sylvia et al., 2022; Bo et al., 2024; Kaushik et al., 2024; Xu et al., 2023), we considered a P < 0.10 as statistically significant.

Results

Descriptive characteristics

The demographic characteristics of the sample children, primary caregivers, and their families at the time of the seventh survey are presented in Table 1. 62.3% of all primary caregivers in our sample were 43 years old or older. Only 13.5% of them had completed more than 12 years of schooling. Mothers were the primary caregivers for 65.1% of the children. Paternal grandmothers were the second most common primary caregivers, caring for 19.2% of the children, followed by fathers (8.6%). The rest of the primary caregivers were either paternal grandfathers, maternal grandparents, uncles, aunts, or older siblings. Slightly more than half of all sample children (52%) were male. 82.2% of the children have siblings, and only a small proportion of them (4.7%) were born prematurely. Detailed participant demographic characteristics at the time of the first survey are provided in Appendix Table 1.

Table 1 Descriptive statistics when the sample children were 9–10 years old (N = 1087).
Full size table

Child developmental outcomes

The children’s cognitive and non-cognitive scores, as well as the rates of developmental delay across the three survey waves are presented in Table 2. We found lower levels of cognitive development in our sample than the average of healthily developed population (Bayley, 1974; Yi et al., 1993; Wechsler, 2008), as indicated by means below the expected average of 100 in both the MDI and the FSIQ scales. Further, we found that 20% of the children experienced cognitive developmental delay (defined as scoring more than 1 SD below the normed mean of a healthy population) between 6 and 12 months of age. The rate of cognitive delay increased to 54% when the sample children were 22 to 30 months old and persisted (48%) when they were 9 to 10 years old. When contrasting cognitive delay between children with different caregivers, we found that children whose primary caregivers were not their mothers experienced higher rates of delay at all three time waves of the study than their peers cared for by mothers.

Table 2 Cognitive and non-cognitive development of sample children at three waves (N = 1087).
Full size table

Table 2 reports the ASQ:SE scores which indicate the children’s non-cognitive development at 6 to 12 months and 22 to 30 months old, respectively, and the SDQ scale which reflects the children’s non-cognitive development at 9 to 10 years old. Due to the lack of cutoff values for the SDQ for identifying the non-cognitive delays at 9 to 10 years old, the rates of non-cognitive developmental delays were only assessed using the ASQ:SE scores and reported in Table 2. We found that an estimated 43% of all sample children experienced non-cognitive developmental delay at 6 to 12 months old; As they grew older, the rate of non-cognitively delay rose to 61% at 22 to 30 months old. Similar to the case of cognitive developmental delay, children primarily cared for by mothers experienced lower rates of non-cognitive developmental delay at both time points in early childhood compared to their peers who were cared for by other family members.

Pathways of cognitive and non-cognitive development in children before age three

Table 3 presents the frequencies and percentages of the four pathways of cognitive and non-cognitive development before age three (i.e., never delayed, improving, deteriorating, and persistently delayed). As indicated, of all 1087 sampled children, only 127 children (11.7%) were in the pathway of never delayed. In contrast, 261 children (24%) were observed in the pathway of persistently delayed. Only a small share of children (N = 130; 12%) demonstrated improvements in either or both domains of cognitive and non-cognitive development; On the other hand, the results show that there is a large share of children (N = 569; 52.3%) was in the pathway of deteriorating, meaning that these children were transiting from normally developed to developmentally delayed in either or both domains of development in the period between 6 to 12 months and 22 to 30 months age.

Table 3 Pathways of cognitive and non-cognitive development of children in the course of 6–12 months and 22–30 months old (N = 1087).
Full size table

Associations between different pathways and the development of children at primary school age

Table 4 presents the associations between four different pathways of development before age three and children’s developmental outcomes at primary school age (including their cognitive skills, non-cognitive skills, and math and reading performance). Children who were either persistently delayed (\(\beta\) = −0.24, p < 0.05) or deteriorating (\(\beta\) = −0.54, p < 0.01) had significantly lower FSIQ scores than children who were in the pathway of never delayed, indicative of lower levels of cognitive skills. No significant differences in the FSIQ scores, however, were found between children in the pathway of improving and children in the pathway of never delayed. On the other hand, compared to children who never experienced developmental delay before the age of three, children who were persistently delayed (\(\beta\) = 0.20, p < 0.10) or children whose development deteriorated (\(\beta\) = 0.25, p < 0.01) before the age of three exhibited significantly higher Total Difficulties scores when they reached primary school age, indicating lower levels of non-cognitive skills. Additionally, Prosocial Behavior scores did not differ significantly for children in any of the development pathways.

Table 4 Associations between pathways of the cognitive and non-cognitive development of children before age three and development at primary school age (N = 1087).
Full size table

The results also show that the pathway of deteriorating development predicted significantly lower math scores of the sample children at primary school age (\(\beta\) = −0.28, p < 0.01), when compared to those in the pathway of never delayed. No significant association, however, was found in any of the three other pathways. In addition, compared to the pathway of never delayed, both pathways of persistently delayed (\(\beta\) = −0.20, p < 0.05) and deteriorating development (\(\beta\) = −0.41, p < 0.01) predicted significantly lower reading scores for the children. No significant association, however, was observed between either the pathway of never delayed or the pathway of improving development and the reading scores of the sample children.

Associations between demographic characteristics and different pathways

Table 5 presents the estimated associations between the demographic characteristics of children and households and the different pathways of the cognitive and non-cognitive development before age three. The results of the analysis show that if a child was born prematurely, they were more likely to be on the pathway of improving (OR = 2.44, p < 0.05) and were at lower risk of deteriorating development (OR = 0.56, p < 0.10). In addition, compared to children from families that are in the bottom tercile of the sample in family asset index, children from families in the top tercile were more likely to be never delayed (OR = 1.84, p < 0.05) and at lower risk of deteriorating development (OR = 0.68, p < 0.05). We also found that, compared to children that had primary caregivers with less than 12 years of schooling, children whose primary caregivers had 12 years or more of schooling were more likely to have normal levels of development (OR = 1.76, p < 0.05) and were less likely to experience deteriorating development throughout the first three years of life (OR = 0.65, p < 0.05).

Table 5 Associations between demographic characteristics and the different pathways (N = 1087).
Full size table

Discussion

Using three waves of data from a longitudinal study in rural China, this study examined associations between four distinct pathways of cognitive and non-cognitive development between 6–12 months and 22–30 months and children’s developmental outcomes at 9 to 10 years of age, including cognitive development, non-cognitive development, math performance, and reading skills. Results suggest that the sample children could be grouped into four pathways based on trajectories of cognitive and non-cognitive development before the age of three: never delayed, persistently delayed, improving, or deteriorating. Importantly, after defining these four pathways, it was shown that children in the different pathways had different associations with their levels of development as they reach primary school age. Children with either persistent developmental delay or deteriorating development had significantly lower levels of development than their peers who never experienced delay, as measured by the FSIQ scale, the Total Difficulties scale, and math and reading tests. On the other hand, we found no significant differences between the pathways of never delayed and improving development in any of the developmental outcomes. Additionally, whether the child was born prematurely, maternal education attainment, and family assets all predicted a child’s entry into different pathways.

The high rates of developmental delay we observed in our sample align with previous reports of prevalent cognitive and non-cognitive developmental deficits in early childhood in rural China (Emmers et al., 2021; Wang et al., 2019). Our study found that 20% to 54% of the sample experienced cognitive developmental delays in the first three years of life, while 43% to 61% experienced non-cognitive delays. This consensus with previous research indicates that prevalent developmental delay persists in rural China.

The literature that has shown that there are dynamic pathways of development that a child may experience in the first years of life (McCormick et al., 2020; Orri et al., 2021; Stålnacke et al., 2019; van Beek et al., 2021). Our study goes further to show that entry into different pathways of development between 6–12 and 22–30 months is not arbitrary. In our sample, only small proportions of children never experienced delay (around 12%) or were on a pathway of improved development (12%), while a much larger proportion of them suffered from persistent delay (24%) or deteriorating development (52%). Our finding adds to a long line of existing works that found the development of children in rural China to stagnate or even deteriorate as they age through the first three years of life (Luo et al., 2019; Wang et al., 2019, 2021, 2022). Taken together, they highlight the importance of identifying interventions that aim to improve the conditions for healthy development in early childhood.

Our study is among the first to reveal associations between cognitive and non-cognitive developmental trajectories before age three, cognitive and non-cognitive development when children reached primary school age, and academic performance at the time. Even though no study has investigated the pathways of development in both cognitive and non-cognitive domains at the same time, our findings echo existing works that examined the long-term effects of either cognitive or non-cognitive development pathways individually (Stålnacke et al., 2019; van Beek et al., 2021; Wang et al., 2021, 2022). We provide evidence that developmental delays during early childhood (the first three years of life) can have long-lasting negative impacts on later cognitive and non-cognitive development as well as levels of academic attainment, as shown by the significantly lower FSIQ, ASQ:SE, SDQ, math, and reading scores at 9 to 10 years old among children who experienced persistent delay or deteriorating development in early childhood. The possible reasons for these effects appear to be that, on one hand, many foundational skills needed for later development form during the first three years of life (Attanasio et al., 2020; Carneiro et al., 2021; Houmark et al., 2024). Children who experience delays during this critical period often struggle in school and in life when they are older (Attanasio et al., 2020; 2022; Bevilacqua et al., 2021). In addition, cognitive and non-cognitive development also appear to be interlinked from an early age and continue influencing each other over time (Chen et al., 2012; Prado et al., 2021; Yoshikawa et al., 2013). Deficiencies in either domain of child development at an early age, thus, can negatively affect both domains as well as academic performance as the child reaches older ages (primary school and older).

Additionally, our study draws on previous studies’ suggestion that socio-economic factors shape the developmental pathways of young children (Bevilacqua et al., 2021; Gutman, Codiroli McMaster, 2020; Wang et al., 2021) to examine which factors specifically predict developmental pathways in a rural Chinese context. Our finding that a child was more likely to be on an improving pathway and was less likely to be on a deteriorating pathway if he or she was born prematurely suggested that the primary caregivers of those children might pay more attention to take care of their children due to their initial vulnerable conditions thus leading to an improving pathway of development when the child is young. This study’s results also demonstrated that children from families with higher SES were more often never experiencing delay and at lower risk of deteriorating development. As observed in previous works (Wang et al., 2021, 2022), maternal education and household resources profoundly shaped the children’s developmental trajectory. One possible explanation could be that caregivers with higher SES have a stronger awareness of the importance of engaging in interactions with their children due to having higher average levels of education and more resources. These stimulating interactions foster better child developmental outcomes (Emmers et al., 2021; Sylvia et al., 2022; Wang et al., 2019).

Our findings have significant implications for scholars and health policy makers in China and beyond. First, the persisting high rates of developmental delay suggest that actions to improve the development of children in rural China when they are at a young age should be a priority of the Chinese government. Successful experiences from other countries across the world have already shown that the implementation of parenting training programs can effectively improve the early childhood development (Attanasio et al., 2020; Carneiro et al., 2021; Jeong et al., 2021). Hence, the government should draw on these examples to design and implement such parenting training programs in rural China. Importantly, policy makers need to design and implement training sessions regarding parental beliefs and knowledge about early childhood development so that rural caregivers can begin to improve their caregiving at home. Considering the pathways of the development before the age of three can have big impact on the developmental outcomes in later life, it is of great value to identify and monitor child development trajectories throughout the first three years of life. This enables swift actions to reduce the risk of children falling into either the pathway of being persistently delayed or the pathway of deteriorating development. On the other hand, since children spend most of their day at schools after they reached primary school age, the negative effects of the primary caregivers at home can be partially avoided if the government invest in training teachers to provide quality care and improving the conditions of primary schools.

Conclusion

This study is the first to concurrently examine the pathways of both cognitive and non-cognitive development of children under age three and the influence of these pathways on subsequent developmental outcomes at primary school age. Using rigorous and multi-dimensional scales of child development from three waves across ten years, we address a significant gap in the field and provide strong evidence for the predictive power of developmental pathways in the first three years of life on later developmental outcomes. Additionally, this study identifies specific factors associated with different developmental pathways before the age of three, laying a foundation for the design of targeted interventions to improve developmental outcomes among children at risk of delay at a young age.

Several limitations also warrant mention. Despite that the sample was randomly selected with a large sample size, the conclusions drawn from our sample might not apply to all rural areas in China. Future research should extend these investigations to other rural regions nationwide. Second, this study relied on caregiver reports to measure children’s non-cognitive development, which might lead to inaccuracies in the responses and is subject to reporting bias. Future research needs to find observational measures of child non-cognitive development.

Our investigation of associations between developmental pathways before age three and child developmental outcomes at primary school age highlight the need for timely, sustained interventions to address early developmental delays among children in rural China. By identifying key factors associated with different developmental pathways, the findings provide a foundation for designing effective policies and targeted programs that are feasible in rural and low-resource settings and can improve the developmental outcomes of children vulnerable to developmental delay.