Cognitive and Socioemotional Caregiving in Mothers, Fathers, and Children’s Other Caregivers from 51 Low- and Middle-Income Countries

SYNOPSIS Objective. Cognitive and socioemotional caregiving practices are both important for child development, but little is known about the extent to which children’s different caregivers engage in the two types of practices or their relative effects on child development, especially in low- and middle-income countries (LMICs). Design. The current study investigates how often mothers, fathers, and children’s other caregivers in 159,959 families from 51 LMICs engage in cognitive versus socioemotional caregiving practices, associations between these caregiving practices, and how the balance between these practices predicts child development. Results. Caregivers reportedly engage in more socioemotional than cognitive caregiving practices in all LMICs examined at all levels of national development. The more mothers, fathers, and other caregivers reportedly engage in cognitive caregiving practices, the more they engage in socioemotional parenting practices. Engaging in cognitive caregiving practices is the strongest predictor of early childhood development when considering cognitive caregiving, socioemotional caregiving, and the balance between the two types of caregiving. Conclusions. Promoting increased caregiver use of cognitive caregiving and integration of cognitive and socioemotional caregiving could close the gap between the number of cognitive and socioemotional caregiving activities parents engage in and potentially promote child development in LMICs.


INTRODUCTION
Of all the people in a child's life, parents are typically the most powerful influencers of early childhood development (Bornstein & Rothenberg, 2022a).Two types of caregiving practices have been established as especially important for early childhood development: cognitive caregiving and socioemotional caregiving (Bornstein & Putnick, 2012;Bornstein et al., 2022a).Broadly, cognitive caregiving includes the strategies parents enact in stimulating children to engage and understand the environment by describing, demonstrating, and providing children with opportunities to learn, and socioemotional caregiving includes the diversity of parental activities that engage children in interpersonal interactions.
Several studies have demonstrated the benefits of both types of caregiving for early childhood development in LMICs (McCoy, Seiden, Cuartas, Pisani, & Waldman, 2022;Tran, Luchters, & Fisher, 2017).Both cognitive and socioemotional caregiving predict better early childhood development even after controlling for a host of demographic and socioeconomic factors in LMICs (Bornstein, Rothenberg, & Putnick, 2022), and both have been included in indexes of nurturing care that predict better child development (McCoy, Seiden, Cuartas, Pisani, & Waldman, 2022).However, existing literature has largely investigated cognitive and socioemotional caregiving practices separately, leaving three unanswered questions.Specifically, it is not known: (1) how relatively often different caregivers engage in cognitive and socioemotional caregiving, (2) how these two caregiving practices are associated with one another, and (3) how differences in the amounts of cognitive versus socioemotional caregiving provided by caregivers are associated with child development.

Question 1: How Often Do Different Caregivers Engage in Cognitive Versus Socioemotional Caregiving Practices?
In considering the first unanswered question, it is important to recognize that parenting children is time-limited.Parents cannot and do not engage in all types of caregiving activities at all times.Moreover, children are in the care of mothers, fathers, and other nonparental caregivers at different times.Therefore, it is useful to compare the distributions of cognitive and socioemotional practices different caregivers engage in.Understanding more about these practices is additionally important because, when included together as predictors, cognitive caregiving activities emerge as stronger predictors of childhood development than socioemotional caregiving activities in LMICs (Bornstein, Rothenberg, & Putnick, 2022).Yet, mothers in LMICs generally engage in more socioemotional than cognitive caregiving practices, even when mothers in LMICs with higher levels of national development engage in more cognitive caregiving and socioemotional caregiving practices overall (Bornstein & Putnick, 2012).Still, it is not yet known whether differences in amounts of cognitive and socioemotional practices caregivers engage in differ significantly.Similarly, it is not known whether the relative balances in cognitive versus socioemotional caregiving practices are similar in mothers, fathers, and children's other caregivers.
Finally, it is not known whether the gap between cognitive and socioemotional caregiving changes at different levels of national development.In nations with higher levels of national development, parents may be more likely to obtain the skills (e.g., reading and writing) and materials (e.g., books, writing utensils) necessary to engage in more cognitive caregiving activities, and therefore reduce the socioemotional-cognitive caregiving gap.Supporting this hypothesis, existing work using UNICEF data reveals that children in upper middle-income countries are 3.5 times more likely to receive adequate maternal stimulation, almost 2 times more likely to receive adequate other caregiver stimulation, and almost 8 times more likely to have adequate learning materials in the home compared to children from low-income countries (McCoy, Seiden, Cuartas, Pisani, & Waldman, 2022).
The current study attempts to answer these questions by examining differences between cognitive and socioemotional caregiving practices in LMICs in mothers, fathers, and children's other nonparental caregivers, and compare those gaps across different levels of national development.

Question 2: How are Cognitive and Socioemotional Caregiving Practices Associated with One Another?
In considering the second unanswered question, two different hypotheses emerge.One hypothesis proposes that, as parents engage more in one set of caregiving practices (i.e., cognitive or socioemotional), they also engage in more of the complementary set of caregiving practices.This positive association is typical of many positive parenting behaviors that stimulate children in similar ways as the cognitive and socioemotional caregiving practices examined in this study.For instance, parents who are warmer and more affectionate with their child are also more likely to be sensitive to their child's needs and to support them in difficult moments (McKee, Colletti, Rakow, Jones, & Forehand, 2008;Rothenberg, 2019).Indeed, like-kind behaviors are often so highly associated that they can be combined in scales that capture generally "positive" parenting (McKee, Colletti, Rakow, Jones, & Forehand, 2008).An alternative hypothesis proposes that parents engage in more of either cognitive or socioemotional parenting practices at the expense of the other.For instance, given that parent time and energy are limited, it may be easier for parents who have lower levels of literacy or education to play with their child or take them outside (socioemotional caregiving practices) as opposed to read with their child or name/count/draw with their child (cognitive caregiving practices that require higher levels of literacy and education).Indirect evidence for this hypothesis emerges from existing work that finds mothers from LMICs engage in more socioemotional than cognitive caregiving practices (Bornstein & Putnick, 2012), and direct evidence comes from studies that show modularity among different domains of maternal behavior and even different types of maternal responsiveness (Bornstein, Tamis-LeMonda, Hahn, & Haynes, 2008;Bradley & Caldwell, 1995;Skinner, Johnson, & Snyder, 2005).
Examining support for these two opposing hypotheses has implications for promoting both cognitive and socioemotional caregiving practices in LMICs.If cognitive and socioemotional parenting practices are positively associated, then it might be most useful to encourage parents to use the two types of parenting practices simultaneously (e.g., playing with the child by reading a book or naming and counting while playing outside).If cognitive and socioemotional parenting practices are unassociated, then it might be most useful to engage parents in separate interventions to promote each set of caregiving practices.If cognitive and socioemotional parenting practices are negatively associated, then it might be most useful to assess a parent's balance between their cognitive and socioemotional parenting practices and consider how to replace one set of caregiving practices with another to ensure parental participation in both types of caregiving.

Question 3: How is the Balance Between Cognitive and Socioemotional Caregiving Associated with Early Childhood Development?
Considering the "optimal" balance between cognitive and socioemotional caregiving practices leads to addressing the third unanswered question.Research indicates that caregivers engaging more in both cognitive and socioemotional caregiving predicts more optimal child development (Bornstein, Rothenberg, & Putnick, 2022).However, it may be that, in addition to the number of cognitive or socioemotional parenting behaviors that a parent engages in, some optimal balance between these sets of activities might be associated with optimal child development.This balance may be analogous to that found in the classic parenting styles literature.There, for certain ethnicities, an authoritative parenting style, which balances high parental warmth with moderate levels of parental control, produces the most optimal child outcomes, whereas parenting styles that have an imbalance between warmth and control predict less optimal child outcomes (Baumrind, 1989;Pinquart & Kauser, 2018).It may be that a similar delicate balance needs to be struck between cognitive and socioemotional caregiving practices to promote childhood development.
The current study examines this question by building on existing work in two specific ways.First, the current study examines whether the balance struck between cognitive and socioemotional caregiving practices is associated with child development outcomes above and beyond the absolute number of cognitive and socioemotional caregiving activities caregivers engage in (because previous research has already demonstrated greater use of cognitive and socioemotional caregiving in LMICs is associated with more optimal child development; Bornstein, Rothenberg, & Putnick, 2022).Second, past work only examined associations between cognitive and socioemotional caregiving on the one hand, and child development on the other, at the family level (i.e., examining whether any caregiver engaged in these activities; Bornstein, Rothenberg, & Putnick, 2022).The current study examines whether the balance between cognitive and socioemotional caregiving struck by each individual caregiver (i.e., mothers, fathers, and other caregivers) is associated with better child development.Examining these caregiver-specific associations is important, given that different types of caregivers provide different levels of adequately stimulating caregiving in LMICs.For instance, about 40% of mothers, 12% of fathers, and 21% of other caregivers provide such stimulating parenting, with estimates varying slightly across studies (Cuartas et al., 2020;Jeong, McCoy, Yousafzai, Salhi, & Fink, 2016;McCoy et al., 2018;McCoy, Seiden, Cuartas, Pisani, & Waldman, 2022).Given these caregivers' varying levels of participation in caregiving, it is possible that striking a balance between cognitive and socioemotional caregiving might matter differently depending on the caregiver.

The Current Study
The current study examines 159,959 families with children ages 3-5 years old living in 51 LMICs to answer the foregoing three research questions.First, the study explores how often different caregivers engage in cognitive compared to socioemotional caregiving.Second, the study explores whether and how the two caregiving practices are associated with one another.Third, the study explores how the balance between cognitive and socioemotional caregiving provided by caregivers is associated with early childhood development.

Data Source
Data used in this study were collected as part of the UNICEF Multiple Indicator Cluster Surveys, rounds 4 (MICS4; UNICEF, 2009) and 5 (MICS5; UNICEF, 2013) collected from 2009-2017.To participate in the MICS, each country designed and selected a representative probability sample.Personnel from each country, trained by UNICEF, conducted surveys with families, and children's primary caregivers participated after giving their consent (UNICEF, 2013).Surveys were provided in participants' native language by field workers from the same nation (Bornstein et al., 2022b;UNICEF, 2009UNICEF, , 2013)).

Participants
Altogether 159,959 families with children ages 3-5 years old living in 51 LMICs provided data (refer to Supplemental Table S1 for a list of countries).UNICEF identified nations as LMICs based on World Bank classification standards that measure per capita gross national income (World Bank, 2015).To obviate within-family or within-household variance, one child was randomly selected from families with more than one child in the target age range.Children were, on average, 47.31 months old (SD = 6.89), and half (49.30%) were girls.In the dataset used in this study, primary caregivers reported on caregiving practices and child development.Primary caregivers were female caregivers who were the main caregivers of the children in the household, regardless of their biological relation to the child.Therefore, most caregivers were biological mothers, but primary caregivers might include some adoptive mothers, stepmothers, aunts, grandmothers, and foster mothers (Bornstein et al., 2022b).

Measures
Cognitive and Socioemotional Caregiving Practices.The MICS4/5 asked primary caregivers whether the mother, father, or any other caregiver over 15 years of age engaged in each of six different caregiving practices over the past 3 days.Primary caregivers were always asked to respond separately about the caregiving practices of mothers, fathers, and other caregivers in the household.The six practices were: (1) read books or looked at picture books with the child, (2) told stories to the child, (3) sang songs to or with the child, including lullabies, (4) took the child outside the home, compound, yard, or enclosure, (5) played with the child, and (6) named, counted, or drew things with the child (UNICEF, 2009(UNICEF, , 2013)).These items were selected for inclusion in the MICS via a 3-stage process that included, first, identification by a panel of 25 experts and, second, field testing in eight LMICs to examine their psychometric soundness, clarity, relevance, and applicability.Third, an independent panel of 27 experts selected the highest performing items for inclusion in the MICS3 round of data collection starting in 2003 (Kariger et al., 2012).These items have continued to be included in MICS until the present day and were adopted as an indicator of Sustainable Development Goal 4 (ratified by 195 nations) by UNESCO (see Kariger et al., 2012, for further detail).Separate scores were calculated for each of the three caregivers as 0 = the caregiver did not engage in the practice and 1 = the caregiver did engage in the practice.
In accord with previous work (Bornstein, Rothenberg, & Putnick, 2022;Sun, Liu, Chen, Rao, & Liu, 2016), the six items were combined into two separate subscales for each caregiver.The cognitive caregiving subscale includes reading, storytelling, and naming/counting/drawing.These items are grouped together as cognitive caregiving items because each stimulates children to understand the environment and provides children with opportunities to learn (Bornstein & Putnick, 2012;Sun, Liu, Chen, Rao, & Liu, 2016).The socioemotional caregiving subscale includes playing with the child, singing songs, and taking the child outside.These items are grouped together as socioemotional caregiving items because each engages children in interpersonal interaction and communication (Bornstein & Putnick, 2012;Sun, Liu, Chen, Rao, & Liu, 2016).
These measures of caregiving practices can be best understood as indexes as they are composed of formative or causal indicators (Bornstein et al., 2022a).The set of indicators used to compose a given index are placed together, not because they are thought to derive from a common cause or latent factor, but because they are assumed to produce (or cause) a common circumstance or outcome.To be more concrete as regards the caregiving practices examined in this study, a variety of parental actions may potentiate socioemotional wellbeing in children (e.g., singing songs for a child, playing with a child, taking a child outdoors), or cognitive well-being in children but the issue is not that they derive from a common source (e.g., a parent's personality, the quality of interpersonal relationships within the family, living in a context of relative affluence).Rather, the reason for bundling them into a single index is that each action is thought to have a similar effect on the child (i.e., they help support the child's socioemotional or cognitive well-being), and exposure to more of such actions is assumed to be better than exposure to each one separately.
Both the cognitive (α Mother = .73,α Father = .68,α OtherCaregiver = .71)and socioemotional (α Mother = .69,α Father = .65,α OtherCaregiver = .72)caregiving subscales demonstrated adequate reliability according to the Kuder-Richardson -20 reliability estimates (DeVellis, 2016).Moreover, in previous work, both subscales as a whole demonstrated comparable or higher positive correlations with early child development scores than did any individual item, which evidences the enhanced construct validity gained by combining items into these subscales (e.g., Bornstein, Rothenberg, & Putnick, 2022).Each set of three items was summed to create each subscale, wherein 0 = the caregiver completed no activities in the last 3 days to 3 = the caregiver completed all three activities in the last 3 days.Average cognitive and socioemotional caregiving scores for the 51 LMIC are reported in Supplemental Table S1.
Additionally, the current study sought to examine whether the balance between cognitive and socioemotional caregiving scores predicted early childhood development.To operationalize a measure of balance, cognitive caregiving scores were subtracted from socioemotional caregiving scores, and the absolute value of the difference was taken.The absolute value of this difference was calculated because otherwise the balance score would just be a linear combination of the cognitive and socioemotional caregiving scores, and therefore could not be included in linear regression models.The balance score can be interpreted as the measure of how large the gap is between the number of cognitive and socioemotional caregiving practices a caregiver engages in (but does not indicate whether that gap favors cognitive or socioemotional parenting practices).
To ensure that the balance measure is not highly co-linear with the measures of cognitive and socioemotional caregiving, correlations between the balance score and cognitive and socioemotional caregiving scores were calculated for each caregiver.The balance measure was correlated with cognitive caregiving and socioemotional caregiving scores in mothers (r cognitivecaregiving = .29,r socioemotional caregiving = .02,ps < .01),fathers (r cognitivecaregiving = .22,r socioemotional caregiving = .59,ps < .01),and other caregivers (r cognitivecaregiving = .56,r socioemotional caregiving = .03,ps < .01)with effect sizes ranging from small to large (Cohen, 1988).However, even in the largest correlation, the balance measure shared only 35% of variance with the socioemotional caregiving score.Therefore, it does not appear that collinearity between the balance measure and the cognitive and socioemotional caregiving measures is a concern.Additionally, in regression analyses reported below that included cognitive caregiving practices, socioemotional caregiving practices, and balance measure together, variance inflation factors for the balance measure were below 10 in every model run, indicating multicollinearity was not a problem (SAS, 2015).
Early Childhood Development Index.Ten questions in the MICS4/5 Questionnaire for Children under Five measured four domains of early childhood development in children ages 36 months to 59 months, including literacy/numeracy development (e.g., knowing and using numbers/letters/ words), socioemotional development (e.g., recognizing and expressing thoughts and feelings), physical health (e.g., developing gross and fine motor abilities), and approaches to learning (e.g., following directions and independently completing tasks).These 10 questions together comprise the Early Childhood Development Index (ECDI;McCoy et al., 2016;Miller, Murray, Tomson, & Arbour, 2016;UNICEF, 2017).These items were selected for inclusion in the MICS via a three-stage process that included, first, identification of items by a panel of experts and, second, field testing in three LMICs to examine their psychometric soundness, clarity, relevance, and applicability (UNICEF, 2017).Third, a second independent panel of experts, with input from caregivers in these LMICs, selected the highest performing items for inclusion in the MICS4 round of data collection (UNICEF, 2017).These items have continued to be included in MICS until the present day and have been adopted as an indicator of Sustainable Development Goal 4 (ratified by 195 nations) of the United Nations (UNICEF, 2017).Multiple independent psychometric evaluations have demonstrated the validity and cross-national comparability of these indices in the MICS4/5 samples (UNICEF, 2017), and ECDI scores have been used by multiple independent investigation teams using MICS data (e.g., Bornstein, Rothenberg, & Putnick, 2022;Sun, Liu, Chen, Rao, & Liu, 2016).To complete the ECDI, the primary caregiver of the child responded either "Yes" or "No" to questions of whether the child displayed a particular behavior related to child development (UNICEF, 2017).
Following UNICEF guidelines (McCoy et al., 2016; UNICEF, 2017), answers that indicated positive child development were then summed to compute a total Early Childhood Development Index (0-10 scale; higher scores indicated more advanced child development).See Supplemental Table S1 for ECDI scores in each LMIC examined.
National Development.To measure national development, the United Nations Human Development Index (HDI; UNDP, 2014) was used.The HDI includes measures of circumstances present in countries that reflect the extent to which conditions in the country promote the development of people and their capabilities (UNDP, 2014).HDI has three major components: (1) life expectancy, (2) education (composed of adult literacy rate and combined gross enrollment in primary, secondary, and tertiary school), and (3) gross national income per capita.HDI scores range from 0 to 1. HDI  Child age in months and child gender (0 = female, 1 = male), primary caregiver education (0 = no formal education, 1 = primary education, 2 = secondary education, and 3 = higher education), the number of children under age 18 years old who lived in the household, and country of residence were characteristics controlled in analyses.Child age was controlled to ensure that differences between, and associations among, caregiving practices emerged above and beyond those expected by normative child development over time.Child gender was controlled because past work has indicated that caregivers differ in some caregiving practices depending on their child's gender, although gender differences have not been found in past work examining cognitive and socioemotional caregiving practices in LMICs (Bornstein, Rothenberg, & Putnick, 2022).Primary caregiver education was controlled because socioeconomic status is often among the most powerful predictors of child development (Bornstein et al., 2022b), and this investigation endeavored to explore the unique effects of cognitive and socioemotional caregiving after accounting for those associations.The number of children under 18 in the home was controlled to account for the possibility that in homes with more children caregivers may not be able to provide as many caregiving practices to any single child (Juhn, Rubinstein, & Zuppann, 2015;Zajonc, 2001).

Analytic Plan
First, to examine whether a particular caregiver (i.e., mothers, fathers, or other caregivers) was reportedly using more cognitive or socioemotional caregiving practices we utilized generalized linear mixed models with a Poisson link using the "proc genmod" function in SAS (Cameron & Trivedi, 2009;SAS Institute Inc, 2015).In this model, the number of cognitive and socioemotional caregiving practices used by each caregiver were compared after controlling all covariates.Nesting of caregivers within families was accounted for in the modeling process by including family ID as a repeated (i.e., random) effect (Cameron & Trivedi, 2009).Standardized mean differences (SMDs) were computed using the Coxe, West, and Aiken (2009) online calculator to examine three comparisons: the average differences between (1) mothers' mean cognitive and socioemotional caregiving practices, (2) fathers' mean cognitive and socioemotional caregiving practices, and (3) other caregivers' mean cognitive and socioemotional caregiving practices.These SMDs were also examined across levels of HDI, and non-overlapping 95% confidence intervals across HDI levels were interpreted as significant differences.
Second, to examine associations between cognitive and socioemotional caregiving practices correlations between cognitive and socioemotional caregiving practices were calculated for mothers, fathers, and other caregivers separately.
Third, linear regression analyses examined the extent to which mothers', fathers', and other caregivers' cognitive and socioemotional caregiving practices each predicted childhood development.Specifically, in separate models for mothers, fathers, and other caregivers, cognitive caregiving practice scores, socioemotional caregiving practice scores, and the balance measure between cognitive and socioemotional caregiving practice scores were used to predict Early Childhood Development Index scores separate and apart from child age and gender, primary caregiver education, number of children under 18 living in the home, and national development.
On account of the large sample sizes involved, statistically significant results can emerge even with small effect sizes.A statistically significant result does not necessarily signify a result that is also practically meaningful.For this reason, effect size metrics are reported, and results are distinguished by statistical and practical significance.Classic rules-ofthumb proposed by Cohen (1988) were adopted to interpret effect sizes.When interpreting effect sizes, SMDs of 0.2, 0.5, and 0.8, rs of 0.1, 0.3, and 0.5, and R 2 s of .02,.13,and .26are considered small, medium, and large, respectively.

Question 1: How Often Do Different Caregivers Engage in Cognitive Versus Socioemotional Caregiving Practices?
Mothers, fathers, and other caregivers were all reported to engage in more socioemotional than cognitive caregiving practices whether families resided in low, medium, high, or very high HDI countries (Table 1).In the whole sample, these differences are statistically significant, even after controlling for several covariates and nesting of caregivers within families but are small in effect size.Notably, the gap between cognitive and socioemotional caregiving practices decreases in mothers and other caregivers as HDI increases (as evidenced by the decreasing SMD effect sizes and non-overlapping 95% confidence intervals seen at each level of HDI in mothers and other caregivers; Table 1).The one exception to this pattern is seen in mothers from high HDI and very high HDI countries, where the gaps between cognitive and socioemotional caregiving are the same size (as evidenced by overlapping 95% confidence intervals).The pattern for fathers is different; the gap between cognitive and socioemotional caregiving practices is widest in high HDI, then equal in medium HDI countries and very high HDI countries (as evidenced by overlapping 95% confidence intervals), and smallest in low HDI countries.
In sum, in these 51 LMICs mothers, fathers, and other caregivers each participated in more socioemotional than cognitive parenting practices.This pattern was consistent across different contexts: It emerges regardless of the type of caregiver examined and regardless of the level of country HDI.However, for mothers and other caregivers, this gap narrows as country HDI increases.As seen in Table 2, across mothers, fathers, and other caregivers cognitive and socioemotional caregiving practices are positively associated with one another.
Correlations between cognitive and socioemotional caregiving were above r = .60across all caregivers, amounting to large effect sizes.No matter the caregiver, in these 51 LMICs if caregivers participated in a greater number of socioemotional caregiving practices, then they were more likely to also participate in a greater number of cognitive caregiving practices, and viceversa.

Question 3: How is the Balance Between Cognitive and Socioemotional Caregiving Associated with Early Childhood Development?
A greater imbalance between cognitive and socioemotional caregiving practices is associated with lower Early Childhood Development Index scores for mothers (Table 3), fathers (Table 4), and other caregivers (Table 5).These associations emerged as significant even after measures of cognitive and socioemotional practices and other demographic covariates known to be powerful predictors of early childhood development (e.g., child age, primary caregiver education, national development; Bornstein & Rothenberg, 2022a) were controlled.However, the balance measure did not predict practically significant amounts of unique variance in early childhood development scores in any regression model.Balance measures explained only 0 or 0.1% of unique variance in each caregiver model (Tables 3-5).Socioemotional caregiving scores explained a similar proportion of unique variance in each caregiver model.In contrast, cognitive caregiving scores explained 3-5 times as much variance as balance measures or socioemotional caregiving practice scores depending on the caregiver model examined (Tables 3-5).
In other words, the imbalance between cognitive and socioemotional caregiving practices emerged as a statistically significant predictor of worse childhood development even after controlling for overall cognitive and socioemotional caregiving practices.However, the effect sizes associated    with this imbalance indicate that this effect may have little practical significance after other powerful predictors are controlled.In all caregivers, the number of cognitive caregiving practices a caregiver engaged in emerged as the most powerful of the three caregiving measures in predicting Early Childhood Development Index scores.The greater the number of cognitive caregiving activities a caregiver engaged in, the better were children's Early Childhood Development Index scores.However, when examining effect sizes, even the effects of cognitive caregiving practices were much smaller than those of other demographic covariates like parent education, child age, and national development (Tables 3-5).

DISCUSSION
The current study fills gaps in existing literature by demonstrating that, in nearly 160,000 children in more than 50 LMICs, caregivers engage in more socioemotional compared to cognitive caregiving practices, caregivers tend to engage in more cognitive caregiving practices if they also engage in more socioemotional caregiving practices (and vice-versa), and an imbalance between caregiver engagement in cognitive and socioemotional caregiving practices predicts worse child development (but this effect is of questionable practical significance due to its small effect size).Finally, cognitive caregiving practices positively affect early childhood development.

Answering Question 1: Caregivers Engage in Socioemotional Caregiving Practices More Often Than Cognitive Caregiving Practices
Regardless of level of national development, mothers, fathers, and other caregivers all engaged in more socioemotional than cognitive caregiving activities.These differences are small in effect size, but statistically significant after controlling for covariates.One explanation for this gap is that it may be easier for caregivers to engage in socioemotional, compared to cognitive, caregiving activities.Cognitive caregiving included reading books or looking at picture books with the child, telling stories to the child, and naming/ counting/drawing with the child.Given that 27.9% of primary caregivers in the current sample reported receiving no formal education, and a further 28.4% reported receiving only primary education (Bornstein et al., 2021), substantial portions of caregivers in this sample may be illiterate, making it challenging to engage in reading as a cognitive caregiving practice.Additionally, caregivers in LMICs may experience difficulty obtaining learning or drawing materials, making it difficult to engage in reading and drawing with their children (Bornstein & Rothenberg, 2022a).In contrast, the socioemotional caregiving practices (playing with the child, singing songs, and taking the child outside) are freely available to caregivers regardless of their level of literacy or availability of learning materials in their home.
Existing evidence from developmental and intervention science indicates that cognitive caregiving activities are powerful predictors of child development (Bornstein & Putnick, 2012;Bornstein & Rothenberg, 2022a), so finding ways to promote caregivers' cognitive caregiving practices in LMICs should be a priority.It seems especially important to "close the gap" between the cognitive and socioemotional caregiving practices in mothers and other caregivers from LMICs with low levels of national development for two reasons.First, current results reveal that the gap between cognitive and socioemotional caregiving activities is higher in LMICs with low levels of national development (as measured by the HDI) for mothers and other caregivers.Second, national development has been found to the one of the most powerful predictors of child development (Bornstein et al., 2021).However, its effects appear to be mediated in part by parenting practices, and the mediating effects of parenting practices are especially strong in countries with lower levels of national development (Tran, Luchters, & Fisher, 2017).Consequently, promoting more cognitive caregiving practices in LMICs with lower levels of national development may offer an especially powerful protective factor against the pernicious effects of low national development on early child development.

Answering Question 2: Cognitive and Socioemotional Caregiving Practices are Positively Associated with One Another
In these LMICs, if caregivers engage in more socioemotional caregiving practices, they are also likely to engage in more cognitive caregiving practices.It does not appear that caregivers engage in one set of parenting behaviors at the expense of another.
From an interventionist's perspective, this encouraging news suggests that, if parents already engage in socioemotional caregiving practices, if they are taught cognitive caregiving practices, they may likely incorporate such practices in their daily routines (and vice-versa).Put another away, the gap between the number of cognitive and socioemotional caregiving practices parents engage in may be able to be closed by making parents aware of simple cognitive caregiving practices they could incorporate into their daily lives (e.g., reading with, telling stories to, or naming/counting/ drawing with their child).An example of closing this gap emerges from an intervention program designed to teach Wolof-speaking Sengalese to verbally interact and talk with their children more (Weber, Fernald, & Diop, 2017).Long-standing cultural traditions in some parts of rural Africa discourage parents from talking with their children.However, after participating in the intervention, Wolof-speaking Sengalese parents improved their verbal communication compared to parents of controls, and 1 year later children of these Wolof-speaking Sengalese parents demonstrated improved language development compared to Wolof-speaking children of controls (Weber, Fernald, & Diop, 2017).Although this particular intervention was successful, it may be difficult to introduce wholesale changes to parent-child interactions in some cultures.These results provide support for another, more universally applicable intervention that can be incorporated by a variety of cultures: providing more access to schooling so that literacy levels (and accompanying cognitive caregiving activities like reading) can be increased across LMICs (Bornstein & Rothenberg, 2022a).Past investigations of MICS data have demonstrated the power of parental education in promoting child development, as higher paternal and maternal education both predict greater paternal and maternal stimulating care and more books in the home, each of which subsequently predicts more optimal early child development (Jeong, McCoy, & Fink, 2017).

Answering Question 3: An Imbalance Between Cognitive and Socioemotional Caregiving is Associated with Worse Child Development, but the Association is Very Small
The current study endeavored to identify whether striking an equal balance between cognitive and socioemotional caregiving scores is associated with better child development.Findings were somewhat equivocal.It appears that an imbalance between cognitive and socioemotional caregiving practices in mothers, fathers, and other caregivers was associated with worse child development scores, even after accounting for absolute levels of these parenting practices.However, these significant effects did not explain meaningful amounts of variance in early childhood development.Therefore, it appears uncertain whether balancing cognitive and socioemotional caregiving practices impacts child development.Future work investigating this question in other samples is needed.
Nonetheless, regression models clarified that, among balance measures of socioemotional caregiving practices and cognitive caregiving practices, cognitive caregiving practices emerged as the more powerful predictor of better Early Childhood Development Index scores regardless of whether the model examined mothers', fathers', or other caregivers' caregiving.This finding aligns with prior work using the MICS4/5 data set that identified aggregate measures of cognitive caregiving across all caregivers in the family environment as one of the five most powerful predictors of early childhood development (Bornstein & Rothenberg, 2022a;Tran, Luchters, & Fisher, 2017).It also aligns with meta-analyses of literacy interventions in LMICs that have confirmed positive effects on cognitive caregiving practices and language growth in children (Kim, Lee, & Zuilkowski, 2020).

Strengths and Limitations
The current study's strengths include the use of nationally representative samples and data, inclusion of reports about mothers, fathers, and other caregivers, and investigation of different kinds of caregiving in a large sample from a large number of typically understudied LMICs.However, the current study also has several limitations.Primary caregivers reported on all caregivers' caregiving practices and therefore may not capture the totality of all caregivers' parenting.Additionally, only three cognitive and three socioemotional caregiving practices were queried, and therefore the current study does not capture the full range of cognitive or socioemotional caregiving practices (Bornstein & Rothenberg, 2022a).Families in these LMICs live in varied cultural, social, economic, political, educational, and healthcare contexts that shape caregiving practices and adaptive child development, and a parenting practice that is adaptive in one community may not be in another.Although the caregiving activities included in the MICS were selected and specifically designed to be generalizable and comparable across the diverse contexts of LMICs, and have done so effectively in a mass of burgeoning research, expectations such as daily play with one's child or taking one's child outside assume that in different contexts (1) playing/going outside is an important activity, (2) caregivers have the time and resources to engage in that activity, and (3) it is safe to engage in that activity in one's immediate context.None of these is universally true.Finally, "other caregivers" are a heterogenous mix and could include teenage siblings, grandparents, other relatives, et al.Future work could collect multi-informant measures of a greater diversity of specific caregiving practices of specific "other" caregivers.

Conclusion
Despite these limitations, this study adds to our currently poor understanding of diverse caregivers' caregiving in LMICs.Mothers, fathers, and other caregivers report engaging in more socioemotional than cognitive caregiving practices in all LMICs examined here at all levels of national development.Mothers, fathers, and other caregivers were also reported to engage in more socioemotional caregiving practices, the more they engaged in cognitive parenting practices.Finally, it appears that engaging in more cognitive caregiving practices is the strongest predictor of early childhood development when considering cognitive caregiving, socioemotional caregiving, and the balance between these two types of caregiving.Promoting increased caregiver use of cognitive caregiving activities and deeper integration of cognitive and socioemotional caregiving activities could close the gap between the number of cognitive and socioemotional caregiving activities caregivers engage in and potentially promote child development in LMICs.

IMPLICATIONS FOR PRACTICE, APPLICATION, AND POLICY
In the 51 LMICs examined here, a sizable gap appears to characterize caregiver socioemotional and cognitive caregiving practice use.It also appears that increasing cognitive caregiving practices may be the best way to promote early childhood development.One way to accomplish this goal might be to promote the use of cognitive caregiving practices while parents are engaged in the socioemotional caregiving activities that they appear to practice more on a daily basis.For instance, parents could be taught to name or count with their children while taking their child outside (e.g., counting the number of steps to familiar places, counting the number of berries picked from a bush), or to build and read with their child while playing.Caregiver engagement in more of one set of cognitive or socioemotional parenting practices promotes greater parent engagement in the other set of caregiving practices.Consequently, emphasizing the integration of cognitive and socioemotional activities might help parents increase the use of cognitive caregiving practice while ensuring that they still participate in the socioemotional caregiving practices they are more likely to regularly engage in with their children.This suggestion is supported by systematic reviews that find that programs designed to cognitively stimulate children can also improve children's interpersonal skills and self-esteem (Barlow, Smailagic, Ferriter, Bennett, & Jones, 2010).

Age and Gender, Primary Caregiver Education, Number of Children in the Household, and Country of Residence.
scores ≤ .550indicate low national development, .550-.699 indicate medium national development, .700-.799 indicate high national development, and ≥ .800indicate very high national development.The average household HDI score across the entire LMIC sample in this study was .60 (SD = .13),indicating that on average households were living in nations in the "medium" human development level.

Table 1 .
Mother, father, and other caregiver cognitive and socioemotional caregiving scores and differences between reporters' cognitive and socioemotional caregiving scores.Note.HDI = Human Development Index.Low HDI = HDI < .550.Medium HDI = HDI .550-.700.High HDI = HDI .700-.800.Very High HDI = HDI > .800.Standardized mean difference scores can be interpreted as if the second scale listed is the scale that is being compared to.For instance, in the first Mother Cognitive-Socioemotional Standardized Mean Difference scores column, in the first row, that number can be interpreted as "Mothers have a cognitive caregiving score that is .22standard deviations smaller compared to their socioemotional caregiving scores."All standardized mean differences reported here account for nesting of caregivers within families and control for child age and gender, number of children in the home, parent education, and country membership.

Table 2 .
Correlations between cognitive and socioemotional caregiving scores in mothers, fathers, and other caregivers.

Table 3 .
Linear regression analyses comparing the simultaneous associations of mother cognitive caregiving, socioemotional caregiving, the balance between cognitive and socioemotional caregiving and demographic characteristics with child development.Note.∆R 2 = Proportion of variance uniquely added to model by variable standardized parameter.

Table 4 .
Linear regression analyses comparing the simultaneous associations of father cognitive caregiving, socioemotional caregiving, the balance between cognitive and socioemotional caregiving and demographic characteristics with child development.
2 = Proportion of variance uniquely added to model by variable standardized parameter.

Table 5 .
Linear regression analyses comparing the simultaneous associations of other caregiver cognitive caregiving, socioemotional caregiving, the balance between cognitive and socioemotional caregiving and demographic characteristics with child development.
Note. ∆R 2 = Proportion of variance uniquely added to model by variable standardized parameter.
W. Andrew Rothenberg, Duke University Center for Child and Family Policy, 302 Towerview Road, Durham, NC, 27708.EMAIL: war15@duke.edu.Marc H. Bornstein is at the Eunice Kennedy Shriver National Institute of Child Health and Human Development, UNICEF, and the Institute for Fiscal Studies.