Pathways to improved development for children living in poverty

Abstract

Conditional cash transfer programs (CCTs) have shown mixed effects on child development outcomes in the context of poverty. Direct parenting support integrated with CCTs may improve the effectiveness of CCTs for children’s development, and benefits could occur via improvements in parenting practices or the home environment. Here, we use data from a randomized effectiveness trial to examine the pathways connecting parenting support and child development. The parenting program (EI: Educación Inicial) was implemented at scale among beneficiaries of Prospera (a CCT, previously Oportunidades and Progresa). Participants included children between the ages of 3 and 5 years old (n = 1,362) from 91 rural communities in three Mexican states. Communities were stratified by indigenous classification and randomized to one of three arms: (T₀) Comparison group (CCT benefits only); (T₁) CCT benefits plus availability of EI in the community; or (T₂) CCT benefits plus promotion and encouragement by the CCT program to participate in EI. Findings were that participation in the T₂ arm of the study was associated with a 13% increase in the number of play activities that parents engaged in with their children, particularly shared storybook reading and singing. Parents in T₂ showed nearly two times greater odds of reading daily to their children. In mediation analyses, the amount of play activities and shared book reading explained up to 32% of the effects of the EI parenting program on child development outcomes. In this study, collaboration and integration of two social programs was critical for program impact, which occurred through pathways relating to parental engagement in reading and play.

Keywords

childhood development developing country intervention program mechanisms of change parenting behaviors

To reach their developmental potential, children must have stimulating home environments and engaged parents who actively promote learning through behavioral modeling, teaching, and providing new opportunities (Walker et al., 2011). For example, parents’ verbal engagement with their children, through activities such as reading books and telling stories, promotes cognitive development (Bradley & Corwyn, 2002) and the acquisition of language skills (McCartney, Dearing, Taylor, & Bub, 2007). Poverty can negatively affect parenting and the home environment, creating less stimulating environments for children, and thus compromising early childhood physical, cognitive, and social-emotional development (Vernon-Feagans, Garrett-Peters, Willoughby, Mills-Koonce, & The Family Life Project Key Investigators, 2012).

Parents living in poverty often lack resources, knowledge, and capabilities to promote the optimal development of their children (Johnson, Berdahl, Horne, Richter, & Walters, 2014; Linver, Brooks-Gunn, & Kohen, 2002). As a consequence, parents living in poverty are less likely to read or share storybooks with their children, they engage less in play with their children, and they have fewer age- and development-appropriate play materials in the home (Bradley, Corwyn, McAdoo, & García Coll (2001; Duncan, Brooks-Gunn, & Klebanov, 1994). The quality of the home environment and child development outcomes are worse when mothers have fewer years of formal education or less knowledge relating to child development (Harding, Morris, & Hughes, 2015). In a study of 28 low- and middle-income countries (LMICs), fewer than 25% of mothers had read to their children in the previous 3 days, and only about a third had told stories to their children in that same time frame (Bornstein & Putnick, 2012). As a result of poverty and its downstream effects, including insufficient stimulation, hundreds of millions of children under the age of 5 years worldwide fail to meet their developmental potential.

Several types of interventions—such as conditional cash transfer programs (CCTs), home-visiting programs, responsive parenting promotion, or parenting support groups—have attempted to reduce the effects of poverty on child development by encouraging parents to make greater investments in their children’s health and development (Reynolds, 2004; Engle et al., 2011; Grantham-McGregor, Fernald, Kagawa, & Walker, 2013). Studies from Ecuador, Nicaragua, and Peru have shown mixed effects of cash transfers on child development outcomes (Fernald, Gertler, & Hidrobo, 2012). These findings argue that just providing cash to poor families is not sufficient, and promoting home stimulation to improve the development of children from poor families in low-income countries is also needed. Early childhood interventions that include direct parent support and/or education have shown the largest effects on child cognitive development in comparison to interventions that are only focused on nutrition or poverty alleviation (Engle et al., 2011). However, there is a demonstrated need to integrate interventions focused on promotion of child development within existing services and sectors, such as health, education and social services (Grantham-McGregor et al., 2013).

The study described here adds to the existing literature by reporting on mechanisms at work in an evaluation of the integration of two at-scale programs (a conditional cash transfer program and a parenting support program), which provide parenting education in a group setting. This study expands our understanding of the findings on child development outcomes by examining effects of the combination of at-scale programs on parenting practices and the home environment, measured using family care indicator (FCI) subscales and play activity frequency measures.

There are two objectives of the current study. The first objective is to determine the effects of randomization to a scaled-up, cluster-randomized, group-based parenting program, Educación Inicial, integrated with a conditional cash transfer program (Prospera), on parenting behaviors and the home learning environment in rural Mexico. The second objective is to examine whether changes in parenting practices and the home environment resulting from participation in the program act as mediators of the impact of the program on child development outcomes.

Methods

Description of interventions

The Mexican government has created several social “compensatory” programs to address disparities in nutrition, health, primary school attendance, and early childhood education, as well as poverty alleviation. Among the first of these programs was Mexico’s conditional cash transfer program (CCT) Prospera (first called Progresa, then Oportunidades), which provides cash payments to parents if they comply with certain conditions, such as taking their children to preventative health check-ups and/or making sure their school-aged children attend school at least 80% of the time. Details of the program have been described previously (e.g., Fernald, Gertler, & Neufeld, 2008). Prospera is among only a handful of CCTs that have examined effects on child development outcomes. Prospera has shown small but positive effects on child behavior, as well as better cognitive, language, and motor development in children whose households received greater cash transfers (Fernald et al., 2008; Fernald, Gertler, & Neufeld, 2009).

Another large scale program in Mexico, Educación Inicial, has operated since the early 1990s, with the goal to improve the knowledge and parenting practices of caregivers in poor, rural, and underserved communities, and promote the optimal development of at-risk children (Miguel et al., 2010). To date, the program has reached more than 23,000 communities and 400,000 families (Silva, Ulloa, & Ramirez, 2009). Educación Inicial targets pregnant women and caregivers of children aged 0–4 years, operating weekly sessions for 9 months out of the year (Miguel et al., 2010). The program is delivered by local community educators, “promotoras,” using a well-developed, evidence-based curriculum that addresses safety and hygiene, nutrition, and early childhood stimulation at age-specific intervals. Each session is approximately 2 hours long, with an average of 14 mothers attending the sessions (Instituto National de Salud Publica, 2012).

Study design

This study was designed as a randomized controlled effectiveness evaluation. There were two intervention arms testing two different levels of promotion of Educación Inicial in comparison with a third arm, which did not receive the EI parenting program. Other available social services continued to operate as normal in all three treatment arms. The treatment arms are described as follows:

The comparison arm (T₀) received the usual benefits of Prospera, the CCT program, which includes a combination of interventions in health, nutrition, and education, and the receipt of cash transfers. Communities assigned to this arm did not have access to the EI parenting program.

The first treatment arm (T₁) had access to the EI parenting education program in their community in addition to receiving the benefits of Prospera, but Prospera staff did not promote or give logistical support to the parenting program.

The second treatment arm (T₂) had access to the EI parenting education program in their community, in addition to receiving the benefits of Prospera. In this arm, Prospera provided additional promotion and support to encourage participation in the parenting program. The original intention was to make attendance in Educación Inicial one of the conditions of the cash transfer in this arm, but enforcement was not feasible.

Sampling

A sample of 204 eligible communities was stratified by indigenous designation (102 indigenous and 102 non-indigenous) and then randomly assigned within each stratum to one of the three study arms. Community designation as “indigenous” or “non-indigenous” was based on the prominence of the indigenous population in the community (at least 80% speaks an indigenous language) according to the National Institute of Statistics and Geography (INEGI) census classifications. Eligibility criteria for inclusion of communities in the study were: (1) being located in a rural area (population < 2,500) in the Mexican states of Chiapas, Puebla, or Oaxaca; (2) having a minimum of 15 families with children aged 0–2 years; (3) having at least 70% of families in the community receiving benefits from Prospera; and (4) having no current or prior operation (within the past 5 years) of Educación Inicial. Participation in other social programs was not an exclusion criterion.

Data collection

All Prospera beneficiary families in the eligible treatment and comparison communities were invited to participate in the study, with baseline data collection occurring in 2008, 1–2 months before program implementation. Baseline data were collected on 2,472 children aged 0–18 months and their families through in-home interviews with the mother or caregiver of each child. Investigators returned in 2012 to collect follow-up data from the same children assessed at baseline. Due to budgetary limitations, a smaller sample of children (N = 1383) was selected for follow-up at that time. In general, the children excluded from follow-up were from the most remote villages, due to the costs associated with travel. Of those interviewed, we lost some (n = 19) because the Family Care Indicators (FCI) were not applied, and others (n = 2) because they were missing data on two or more of the subscale components, resulting in a sample size of n = 1,362 for the FCI subscales analyses. For our analyses of program impact on the frequency with which parents engaged their children in specific play activities, we included only those caregivers who reported that they engaged in any of the play activities listed (singing, reading, and talking and playing), resulting in sub-samples for singing (n = 560), shared book reading (n = 717), and talking and playing analyses (n = 742). Finally, we tracked implementation of the Educación Inicial program in all enrolled communities to monitor contamination. Of the communities in the final sample, some in the treatment arm failed to implement the program (n = 1 in T₁, n = 7 in T₂); and some in the comparison arm (n = 2 in T₀) had actually implemented Educación Inicial.

Measurement of parental engagement and home environment

Family Care Indicators (FCI) are a set of indicators that measure the quality and quantity of stimulation available to a child in his or her home environment. The FCI was originally derived from the Home Observation for Measurement of the Environment (HOME) Inventory (Caldwell & Bradley, 1984). The FCI is validated and used worldwide in the UNICEF Multiple Indicator Cluster Survey (Kariger et al., 2012) and is associated with early child development (Hamadani et al., 2010). The FCI consists of several sets of yes/no questions, including the sources of play materials (household objects or things from outside, homemade toys, or toys purchased from a toy store), the variety of play materials (items to make/play music, items to use to draw and/or write, picture books for children, objects to stack/construct/build, things for moving around, objects for learning shapes and colors, and items for pretend play), and parent–child play activities in the previous 3 days (reading/looking at picture books; telling stories; singing songs; taking child outside the home; playing with toys; and spending time naming things, drawing and counting). The last part of the FCI is a count of the number of books in the household, capped at 10 or more. Finally, separate questions were asked about the frequency with which parents engaged in singing, reading, and talking and playing with their child. These questions were asked of parents who reported that they engaged in the play activities, but relevant activities were not restricted to the previous 3 days. The response categories for the frequency question were not conducive to a linear examination of frequency (everyday, 2–6 times per week, 1 time per week, and less than 1 time per week). Therefore, we examined these variables as “parents who engaged in the activity on a daily basis” vs. “parents who engaged in the activity less than daily.”

Our analyses focus on four key outcomes: 1) the variety of play materials in the household, 2) the variety of parent–child play activities, 3) the presence of books in the household, and 4) the frequency of parental engagement in specific play activities. We hypothesized that these characteristics would be the most likely to change as a result of exposure to the program and thus would likely mediate the associations between Educación Inicial and child development outcomes.

Less than 1% of the follow-up analysis sample was missing data on the FCI subscales. For children missing only one subscale component (n = 14 for play activities and n = 12 for play materials), a conservative imputation of zero was applied to retain a score for the total subscale. Subscale scores missing more than one indicator measure were excluded (n = 2).

Measurement of child development

Child development was measured using the McCarthy Scales of Children’s Abilities (MSCA), which is a psychological test that can be administered to children aged 2–8 years to assess their abilities in five domains; Verbal Ability, Perceptual-Performance (non-verbal reasoning), Quantitative, Memory, and Fine and Gross Motor Skills (McCarthy, 1972). The Verbal, Perceptual, Quantitative, and Memory scales of the MSCA were applied in this study. The General Cognitive Index (GCI) is a sum of the Verbal, Perceptual, and Quantitative Scale scores. The MSCA was translated and adapted for the local context by one of the authors [LS], who also trained field staff that applied the instrument. Test scores were converted to age-adjusted standard scores using 2-month age intervals. The standard scores have a mean of 100 and standard deviation of 15. Each 2-month age category included between 10% and 13% of the sample except for the categories at the two extremes (youngest category: 9%, oldest category: 6%).

Other household, parent and child variables

Data were collected on household socio-economic status and demographic structure, characteristics of household members, state of residence (Chiapas, Oaxaca, or Puebla), and whether or not the community was indigenous or non-indigenous. 70% of residents in the study communities were recipients of Prospera, a welfare program, and thus designated as poor. Other relevant variables included child age (in 6-month categories) and sex (boy, girl), parent education (kindergarten or less, completed primary school, completed secondary school), whether the father was present in the household, and household wealth. Household wealth was measured using an inventory of household assets (21 items, e.g., refrigerator, TV, iron). The average number of household assets owned in the sample was four. A principal components analysis was used to consolidate this index, and we retained the first component (Falkingham & Namazie, 2002; Jolliffe, 2002). This first component was then rotated using the varimax rotation and logged to account for rightward skew (Vyas & Kumaranayake, 2006). Wealth was included in the model as a continuous variable. In addition to the above, variables for household composition (number of adults and children in the house) and household crowding (number of people in the house/number of rooms), as well as the presence of electricity and piped water in the home, were used to assess baseline differences among the treatment and comparison arms in the follow-up group. Fewer than 1% of follow-up observations were missing data for mother’s education, father present, or household wealth, while 4.8% of observations were missing data for father’s education. No data were missing for child age or sex. Where demographic data were missing, values were imputed using the community mean.

Statistical analysis

Our specific statistical approach for addressing each of our stated objectives is described below. For all of the three objectives described, we included indicator variables for state of residence (Chiapas, Oaxaca, and Puebla) and clustered standard errors at the community level. We assessed the robustness of the findings from our parsimonious models by comparing these results to the results from models in which we controlled for potential confounding by including the following variables: child age and sex, parent education, whether the father was present in the household, whether the community was indigenous or non-indigenous, and household wealth. We conducted our statistical analyses using STATA 13 (STATA Corporation, College Station, TX, USA).

To address our first objective, which was to examine the effectiveness of Educación Inicial on parenting behaviors and the home-learning environment, we analyzed the variety of play materials and parent–child play activities using Poisson regressions, and then examined the play activities items individually using logistic regression. In these models, assignment to the EI parenting program was the independent variable, and each measure of parenting behavior was the dependent variable in repeated independent models. The number of books in the household (two categories: 0 & ≥ 1), and the frequency of play activities (every day/less than every day) conditional on engaging in the play activities to avoid a problem of excess zeros and over dispersed data, were also analyzed using logistic regression. All regressions were run incorporating both treatment groups as dummy variables within the same regression (Johnson & Wichern, 2012). We did not adjust for multiple testing as it was not justified by the design of the study or the intentions of the analyses (Bender & Lange, 2001).

To address our second objective, which was to examine whether changes in parenting practices and the home environment resulting from participation in Educación Inicial acted as mediators of the impact of the program on child development outcomes, we used mediation analyses (paramed command in Stata) (Emsley & Liu, 2013). In these models, the independent variable was randomized assignment to Educación Inicial, the dependent variable was child development scores (MCSA), and the mediators were the Family Care Indicators and frequency of play activities that were found to be significantly influenced by the EI parenting program in the first objective. For the mediation analyses we only examined the T₂ group, because assignment to T₁ did not have statistically significant effects on the mediators. To examine mediation, we used parametric regression techniques (Valeri & VanderWeele, 2013), which require four key assumptions: 1) the exposure–outcome relationship and 2) the exposure–mediator relationship are not confounded conditional on the covariate set; 3) the mediator–outcome relationship is not confounded conditional on the covariate set and the exposure; and 4) the exposure does not affect the mediator–outcome confounders (VanderWeele & Vansteelandt, 2009). Because our exposure was randomized, we assume the exposure–outcome and exposure–mediator relationships are not confounded. To address the third assumption, we controlled for child age and sex, parents’ education, father presence, household wealth, indigenous community, and state of residence. Finally, the fourth assumption was met because the mediator–outcome confounders were present and measured before initiation of the program and therefore could not have been influenced by randomization to Educación Inicial.

In a linear mediation model, the total effect of the intervention is the overall average change in the outcome in T₂ compared with T₀; this total effect can then be decomposed into the direct and indirect effects (Valeri & VanderWeele, 2013, VanderWeele, 2014). The percent of the effect of randomization to Educación Inicial on child development scores that was mediated by the family care variables was calculated by dividing the indirect effect by the total effect. The indirect effect is the effect of the mediator holding treatment constant (Pearl, 2012). A separate model was run for each mediator. We used bootstrapping (1,000 samples) to ensure accuracy of the inferences due to the smaller sample sizes.

Results

Descriptive statistics

Baseline characteristics between the followed and not followed groups were not significantly different for child and parental characteristics, and attrition did not differ between study arms (Supplementary Table 1). However, there were a greater proportion of households with piped water in the follow up group (74% followed vs. 63% not followed) and less crowding (3.08 followed vs. 3.41 not followed), and households had greater average wealth (0.81 followed vs. 0.68 not followed).

At follow-up, the three groups (T₀, T₁, T₂) were well balanced across a range of child, parent, and household characteristics, suggesting that despite the loss to follow up, randomization at baseline was successful (Supplementary Table 2). There was wide variability in the types of play materials owned at follow-up, as well as the play activities that parents engaged in with their children (Supplementary Table 3). For example, the majority of parents reported owning toys for pretend play, such as dolls (> 85%), but only about 12% of parents reported having toys for building, such as blocks. Over 75% of parents reported taking their children outside the home within the last 3 days, whereas only 33% or fewer reported telling stories to their children in the same time period. More than 50% of households owned over 10 books in total, but these books were not limited to children’s picture books, and could have included schoolbooks, bibles, or other adult books. Despite high book ownership, fewer than half of all parents reported engaging in shared storybook reading with their children.

Effect of Educación Inicial on parenting practices and home environment

Parents in T₂ were more likely than parents in T₀ to participate in play activities with their children (3.37 vs. 2.85 activities); the magnitude of the coefficient decreased but remained significant in adjusted analyses (Table 1). Parents in T₂ were also more likely to engage in shared storybook reading (52% vs. 39%) and singing (49% vs. 38%) than parents in T₀, but these differences became non-significant with the inclusion of covariates. Spending time drawing, or counting and naming things displayed a non-significant trend when comparing T₂ and T₀ (data not shown). There were no differences in play activities, reading, or singing between T₁ and T₀.

Table 1.

Unadjusted and adjusted analyses: Effects of randomization to Educación Inicial on Family Care Indicators (n = 1,362).

	Unadjusted		Adjusted
	Conditional cash transfer program with Educación Inicial available vs. Conditional cash transfer program only¹	Conditional cash transfer program with promotion and encouragement of participation in Educación Inicial among participants vs. Conditional cash transfer program only¹	Conditional cash transfer program with Educación Inicial available vs. Conditional cash transfer program only¹	Conditional cash transfer program with promotion and encouragement of participation in Educación Inicial among participants vs. Conditional cash transfer program only¹
Family Care Indicator subscales²	Log count difference (CI)	Log count difference (CI)	Log count difference (CI)	Log count difference (CI)
Play activities	0.05 (−0.09, 0.18)	0.16 (0.03, 0.30)*	0.04 (−0.07, 0.15)	0.12 (0.002, 0.24)*
Individual play activities³	Odds ratio (CI)	Odds ratio (CI)	Odds ratio (CI)	Odds ratio (CI)
Shared book reading	1.01 (0.69, 1.47)	1.60 (1.06, 2.40)*	1.00 (0.69, 1.44)	1.50 (0.98, 2.30)
Singing	1.35 (0.91, 2.01)	1.52 (1.01, 2.29)*	1.35 (0.91, 1.93)	1.44 (0.96, 2.15)
Play activity frequencies⁴	Odds ratio (CI)	Odds ratio (CI)	Odds ratio (CI)	Odds ratio (CI)
Daily reading	1.89 (1.08, 3.31)*	1.89 (1.04, 3.42)*	1.90 (1.08, 3.33)*	1.90 (1.05, 3.45)*

Note. All analyses compare treatment groups to control (Educacion Inicial available vs. CCT only; Educacion Inicial available and participation encouraged vs. CCT only) using indicator variables. Unadjusted analyses combine indigenous strata, adjust for community level clustering and control for state fixed effects. Adjusted analyses combine indigenous strata, adjust for community level clustering and control for state fixed effects, child age and sex, mother and father education, father present, household wealth, and indigenous community. ¹The conditional cash transfer program only group is the comparison group. ²Family Care Indicator subscales are sums of number of parent–child play activities (6 items). ³Individual play activities are measured as yes/no. ⁴Play-activity frequencies are measured as every day/less than every day. Sample sizes for individual play-activities are: singing = 1,357; and for play activity frequencies are: reading = 717. *p < .05.

Parents in both treatment groups (T₁ and T₂) had significantly greater odds of reading with their children every day than parents in the comparison group T₀ (T₁ and T₂ > 11% vs. 7% T₀). The difference in frequency of reading remained significant with the inclusion of covariates. There were no additional significant differences between T₂ or T₁ and T₀.

As a test of sensitivity, unadjusted and adjusted analyses were repeated using inverse probability censored weights.¹ After including weights, the differences in the total FCI play activities and the odds ratios for the individual play activity frequencies remained nearly unchanged with the inclusion of weights (Supplementary Table 4). However, the unadjusted odds ratio for singing comparing T₂ to T₀ and the log count difference comparing T₂ to T₀ in the adjusted model became non-significant.

Mediation analysis: Home environment and parenting practices

Mediation was considered significant if the indirect effect (the effect of the mediator on the outcome holding treatment constant) was significant at the 0.05 level. The FCI play activities subscale was a significant mediator for the General Cognitive Index (GCI) and all subscales of the McCarthy. Shared storybook reading significantly mediated the program effect on the GCI and the Perceptual Scale. No other mediators reached statistical significance. The percent of the total effect of Educación Inicial that was mediated through changes in parenting behaviors (as measured by the FCI play activities subscale) was 17% for the Verbal Scale, 32% for the Quantitative Scale, 25% for the Perceptual Scale, 12% for the Memory Scale, and 22% for the GCI (Table 2). One-fifth (20%) of the effect of the program on the GCI was mediated by engagement in shared storybook reading. Shared storybook reading also mediated 32% of the effect on Perceptual Scale scores. There was no mediation by parents singing with or frequency of reading to their children. The percent of the mediated effect cannot be compared across models, as the mediators are not mutually exclusive (e.g., shared book-reading and singing are components of the FCI play activities subscale, and frequency of reading is a measure for parents that engage in shared book-reading), and because each model does not account for the effects of the other mediators.

Table 2.

Mediation analysis: The percent of the total effect of Educación Inicial on McCarthy Scores mediated by Family Care Indicators, and the decomposed direct and indirect effects, comparing the conditional cash transfer program with promotion and encouragement of participation in Educación Inicial among participants group and the conditional cash transfer program only group¹.

McCarthy Score	Mediator	% of TE that is mediated²	Direct effect (CI)			Indirect effect (CI)
General Cognitive Index³	Play activities subscale⁴	22	2.20	(0.26, 4.48)	*	0.63	(0.18, 1.34)	*
	Shared book reading⁵	20	2.18	(0.28, 4.39)	*	0.54	(0.06, 1.13)	*
	Singing⁵	7	2.62	(0.63, 4.87)	*	0.19	(−0.13, 0.69)
	Daily reading⁶	2	3.23	(0.50, 6.33)	*	−0.06	(−0.68, 0.30)
Verbal
	Play activities subscale	17	2.33	(0.37, 4.40)	*	0.49	(0.12, 1.13)	*
	Shared book reading	12	2.43	(0.46, 4.47)	*	0.32	(−0.10, 0.83)
	Singing	8	2.61	(0.70, 4.84)	*	0.22	(−0.08, 0.74)
	Daily reading	6	2.93	(0.40, 5.79)	*	−0.17	(−0.74, 0.21)
Quantitative
	Play activities subscale	32	1.32	(−0.86, 3.43)		0.61	(0.17, 1.40)	*
	Shared book reading	20	1.50	(−0.74, 3.75)		0.37	(−0.08, 0.95)
	Singing	10	1.72	(−0.54, 3.94)		0.20	(−0.14, 0.70)
	Daily reading	6	1.90	(−0.90, 4.91)		0.13	(−0.37, 0.71)
Perceptual
	Play activities subscale	25	1.63	(−0.29, 3.86)		0.55	(0.12, 1.23)	*
	Shared book reading	32	1.38	(−0.58, 3.66)		0.67	(0.14, 1.24)	*
	Singing	3	2.08	(0.03, 4.27)		0.06	(−0.32, 0.52)
	Daily reading	<1	3.08	(−0.03, 6.28)		0.004	(−0.56, 0.45)
Memory
	Play activities subscale	12	2.65	(0.72, 4.74)	*	0.38	(0.01, 0.95)
	Shared book reading	6	2.81	(0.89, 4.88)	**	0.17	(−0.23, 0.77)
	Singing	6	2.76	(0.84, 4.78)	**	0.19	(−0.12, 0.69)
	Daily reading	6	2.67	(−0.05, 5.69)		−0.15	(−0.81, 0.23)

Note. Coefficients are the difference in McCarthy Scores. 95% CI are bootstrapped with 1,000 samples. All models adjust for child age and sex, parent’s education, father present, household wealth, indigenous community, and state fixed effects. ¹The conditional cash transfer program only group is the comparison group. ²Percent of Total Effect (TE) mediated is calculated using the indirect effect divided by the total effect. ³The General Cognitive Index is a sum of the Verbal, Perceptual and Quantitative Scale scores. ⁴Play Activities Subscale is the sum of Family Care Indicator parent-child play-activities (6 items). ⁵Shared Book Reading and Singing are measured as yes/no. ⁶Play- activity frequencies (Daily Reading) are measured as every day/less than every day. ** p < .01 * p < .05.

Discussion

In this study, we have shown that participation in a weekly, group-based, parenting education program in Mexico (Educación Inicial) had positive effects on parenting practices and the home environment. Even after adjustment for covariates, there was a 13% increase (calculated by exponentiating the log count difference) in the number of different play activities parents engaged in with their children, and nearly two times greater odds of parents reading daily with their children in the T₂ intervention group. Using mediation analyses, we also showed that these changes in parenting behavior were partially responsible for explaining the positive effects that the EI program had on child development outcomes, with up to 32% of the effect explained. The treatment effects were significant only when comparing the T₂ treatment arm with T₀.

The effect size of the T₂ intervention on FCI play activities (d = 0.29) is slightly smaller than that found on HOME scores (from which the FCI is derived) in two similar but smaller studies of group based parenting programs in Bangladesh (d = 0.34–0.38) (Aboud, 2007; Aboud & Akhter, 2011) and substantially smaller than effects on HOME-based parenting measures in programs that include home visiting in Bangladesh (d = 0.55–0.68) (Aboud, Singla, Nahil, & Borisova, 2013) and Jamaica (d = 0.50). It is likely that effect sizes reported here are smaller than those found in other studies because the Mexican program was group-based and not as intensive as existing parenting support programs.

All of the results reported here occurred in the context of an existing CCT (Prospera). It is possible that in the T₂ arm, the collaboration and support from a large and well-established social program aided Educación Inicial in achieving its effects on parenting through increased awareness of and participation in the program. There may also have been increased community and participant buy-in of the new program due to community familiarity with Prospera. It is possible that parents may have believed that Educación Inicial was in fact a part of Prospera, or that attendance and participation in sessions was a condition of their cash transfer benefits as a result of the collaboration between the two programs in the T₂ arm, in spite of the fact that there were no monetary consequences or benefits related to attendance at the group-based Educación Inicial sessions.

There were no effects on parenting practices, the home environment or child development in the T₁ arm, which did not receive institutional support from Educación Inicial by Prospera. Qualitative reports suggest that recruitment of caregivers to participate in the EI parenting program was more difficult in the T₁ arm when compared with T₂ because Prospera program promotoras encouraged mothers to attend Educación Inicial in the T₂ arm, but there was no support from Prospera in the T₁ arm (Instituto National de Salud Publica, 2012). These findings suggest that a CCT may be an effective platform through which a parenting program can reach particularly vulnerable children.

Our findings have encouraging implications for resource-limited settings. For example, we found that parenting support programs can be successful simply by teaching parents to use what they have available and to be creative in their developmentally supportive play; this technique has been used previously by the well-known Jamaican home-visiting program (Grantham-McGregor, Powell, Walker, & Himes, 1991), among others. There is good evidence that combining nutritional and child development activities, or integrating child development support into existing systems and structures, can be beneficial for young children (Grantham-McGregor et al., 2013). This existing evidence suggests that integrating parenting support into an existing conditional cash transfer program may be a cost-effective and at-scale approach to improving child development in the context of poverty.

The primary limitation of this study is the lack of complete follow-up data for the original sample. Due to budget constraints, children who were the most difficult to reach were not followed. Our sensitivity analyses indicate that poorer, more crowded households, and households without piped water were less likely to be followed; their non-inclusion affects the generalizability of the results found especially given that these are documented developmental risk factors in multiple countries (Walker et al., 2011). This caveat is especially important as the intention of the program is to target the most poor, rural and marginalized children who are at the greatest nutritional and psychosocial risk. This limitation does not appear to affect the internal validity of the study, as balance was maintained among the three treatment arms at follow-up. An additional limitation is that independent effects of the two programs cannot be disentangled, as it was not possible to have a control group that did not receive Prospera, since the program exists throughout Mexico.

This study demonstrates that a well-designed group parenting program such as Educación Inicial, implemented in coordination with an existing CCT program can have a positive effect on parenting behaviors and the home environment, which are known to improve early childhood development. Collaboration and integration of social programs may provide cost savings through economies of scale, and may be critical for program impact, as suggested by our finding that Educación Inicial had significant effects on parenting and the home environment only when the program received promotional support from Prospera. The results reported here have implications for the design of parenting education programs in LMICs. Group-based, weekly, parent education programs can be an effective solution for resource-limited contexts where early education services or home-visiting programs are not available, but their effects may be enhanced when working in collaboration with an established social infrastructure.

Footnotes

Acknowledgements

We acknowledge the significant contributions of our late colleague Patrice Engle, who contributed greatly to the initiation and design of the study described here. We gratefully acknowledge the contributions of Concepcion Steta during the planning of this project.

Funding

The author(s) declared receipt of the following financial support for the research, authorship, and/or publication of this article: This study was implemented as part of the external evaluation of the Programa de Desarrollo Humano Oportunidades (now PROSPERA Programa de Inclusión Social), with financial loans from the World Bank and Interamerican Development Bank. The authors received support for the data analysis and preparation of this manuscript from the UBS Optimus Foundation.

Note

Supplemental material

The online supplementary material is available at

References

Aboud

F. E.

(2007). Evaluation of an early childhood parenting programme in rural Bangladesh. Journal of Health, Population, and Nutrition, 25(1), 3–13.

Aboud

F. E.

Akhter

(2011). A cluster-randomized evaluation of a responsive stimulation and feeding intervention in Bangladesh. Pediatrics, 127(5), e1191–e1197.

Aboud

F. E.

Singla

D. R.

Nahil

M. I.

Borisova

(2013). Effectiveness of a parenting program in Bangladesh to address early childhood health, growth and development. Social Science & Medicine, 97: 250–258.

Bender

Lange

(2001). Adjusting for multiple testing—when and how? Journal of Clinical Epidemiology, 54(4), 343–349.

Bornstein

M. H.

Putnick

D. L.

(2012). Cognitive and socioemotional caregiving in developing countries. Child Development, 83(1), 46–61.

Bradley

R. H.

Corwyn

R. F.

(2002). Socioeconomic status and child development. Annual Review of Psychology, 53(1), 371–399.

Bradley

R. H.

Corwyn

R. F.

McAdoo

H. P.

García Coll

(2001). The home environments of children in the United States part I: Variations by age, ethnicity, and poverty status. Child Development, 72(6), 1844–1867.

Caldwell

B. M.

Bradley

R. H.

(1984). Home observation for measurement of the environment: Administration manual. Tempe, AZ: Family & Human Dynamics Research Institute, Arizona State University.

Duncan

G. J.

Brooks-Gunn

Klebanov

P. K.

(1994). Economic deprivation and early childhood development. Child Development, 65(2), 296–318.

10.

Emsley

Liu

(2013). PARAMED: Stata module to perform causal mediation analysis using parametric regression models. Statistical Software Components. Available at: https://ideas.repec.org/c/boc/bocode/s457581.html

11.

Engle

Fernald

L. C.

Alderman

Behrman

O’Gara

Yousafzai

… Ertem

(2011). Strategies for reducing inequalities and improving developmental outcomes for young children in low-income and middle-income countries. The Lancet, 378(9799), 1339–1353.

12.

Falkingham

Namazie

(2002). Measuring health and poverty: A review of approaches to identifying the poor. London, UK: DFID Health Systems Resource Centre.

13.

Fernald

L. C. H.

Gertler

P. J.

Hidrobo

(2012). Conditional cash transfer programs: Effects on growth, health, and development in young children. In Maholmes

King

R. B.

. (Eds.), The Oxford handbook of poverty and child development (pp. 569–600). New York, NY: Oxford University Press.

14.

Fernald

L. C. H.

Gertler

P. J.

Neufeld

L. M.

(2008). Role of cash in conditional cash transfer programmes for child health, growth, and development: An analysis of Mexico’s Oportunidades . The Lancet, 371(9615), 828–837.

15.

Fernald

L. C. H.

Gertler

P. J.

Neufeld

L. M.

(2009). 10-year effect of Oportunidades, Mexico’s conditional cash transfer programme, on child growth, cognition, language, and behaviour: A longitudinal follow-up study. The Lancet, 374(9706), 1997–2005.

16.

Grantham-McGregor

S. M.

Powell

C. A.

Walker

S. P.

Himes

J. H.

(1991). Nutritional supplementation, psychosocial stimulation, and mental development of stunted children: The Jamaican study. The Lancet, 338(8758), 1–5.

17.

Grantham-McGregor

S. M.

Fernald

L. C.

Kagawa

Walker

(2013). Effects of integrated child development and nutrition interventions on child development and nutritional status. Annals of the New York Academy of Sciences, 1308, 11–32.

18.

Hamadani

J. D.

Tofail

Hilaly

Huda

S. N.

Engle

Grantham-McGregor

S. M.

(2010). Use of family care indicators and their relationship with child development in Bangladesh. Journal of Health, Population, & Nutrition, 28(1), 23–33.

19.

Harding

J. F.

Morris

P. A.

Hughes

(2015). The relationship between maternal education and children’s academic outcomes: A theoretical framework. Journal of Marriage and Family, 77(1), 60–76.

20.

Instituto Nacional de Salud Pública. (2012). Evaluación de impacto del programa de Educación Inicial no Escolarizada de CONAFE (PEI-CONAFE) en beneficiarios de Oportunidades. Cuernavaca, Morelos: Mexico. Available at: http://www.conafe.gob.mx/libroblanco/Documents/2-%20Anexos%20Libro%20Blanco%20Programa%20de%20Educación%20Inicial/Anexo%2098/Producto_4_Informe_final_PEI-CONAFE_6Agosto2012%20(2).pdf

21.

Johnson

B. D.

Berdahl

L. D.

Horne

Richter

E. A.

Walters

M.-g.

(2014). A parenting competency model. Parenting, 14(2), 92–120.

22.

Johnson

R. A.

Wichern

D. W.

(2012). Applied multivariate statistical analysis. Englewood Cliffs, NJ: Prentice Hall.

23.

Jolliffe

(2002). Principal component analysis. New York, NY: Springer-Verlag New York, Inc.

24.

Kariger

Frongillo

E. A.

Engle

Britto

Sywulka

S. M.

Menon

(2012). Indicators of family care for development for use in multicountry surveys. Journal of Health, Population, and Nutrition, 30(4), 472–486.

25.

Linver

M. R.

Brooks-Gunn

Kohen

D. E.

(2002). Family processes as pathways from income to young children’s development. Developmental Psychology, 38(5), 719–734.

26.

McCarthy

(1972). McCarthy Scales of Children’s Abilities. New York, NY: Psychologial Corporation.

27.

McCartney

Dearing

Taylor

B. A.

Bub

K. L.

(2007). Quality child care supports the achievement of low-income children: Direct and indirect pathways through caregiving and the home environment. Journal of Applied Developmental Psychology, 28(5), 411–426.

28.

Miguel

M. E.

Mendez

A. L.

Galicia

S. M.

Lazarin

R. I. V.

Morales

L. M. V.

Silva

S. V.

(2010). Modelo de Educación Inicial del CONAFE [ Model of Educación Inicial of CONAFE ]. Mexico: Consejo Nacional de Fomento Educativo. Available at: http://www.conafe.gob.mx/educacionincial/pdf/documentos-normativos/ModeloEducacionInicial.pdf

29.

Pearl

(2012). The causal mediation formula—a guide to the assessment of pathways and mechanisms. Prevention Science, 13(4), 426–436.

30.

Reynolds

A. J.

(2004). Research on early childhood interventions in the confirmatory mode. Children and Youth Services Review, 26(1), 15–38.

31.

Silva

Ulloa

Ramirez

(2009). Análisis ejecutivo de resultados de la evaluación de impacto de Educación Inicial no Escolarizada 2008 [Executive analysis of results of the impact evaluation of Educacion Inicial 2008]. Mexico City, Mexico: Secretario de Educación Pública (Secretary of Public Education).

32.

Valeri

VanderWeele

T. J.

(2013). Mediation analysis allowing for exposure–mediator interactions and causal interpretation: Theoretical assumptions and implementation with SAS and SPSS macros. Psychological Methods, 18(2), 137–150.

33.

VanderWeele

T. J.

Vansteelandt

(2009). Conceptual issues concerning mediation, interventions and composition. Statistics and its Interface, 2, 457–468.

34.

VanderWeele

T. J.

(2014). A unification of mediation and interaction: A 4-way decomposition. Epidemiology, 25(5), 749–761.

35.

Vernon-Feagans

Garrett-Peters

Willoughby

Mills-Koonce

, & The Family Life Project Key Investigators. (2012). Chaos, poverty, and parenting: Predictors of early language development. Early Childhood Research Quarterly, 27(3), 339–351.

36.

Vyas

Kumaranayake

(2006). Constructing socio-economic status indices: How to use principal components analysis. Health Policy and Planning, 21(6), 459–468.

37.

Walker

S. P.

Wachs

T. D.

Grantham-McGregor

Black

M. M.

Nelson

C. A.

Huffman

S. L.

… Lozoff

(2011). Inequality in early childhood: Risk and protective factors for early child development. The Lancet, 378(9799), 1325–1338.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.36 MB

0.35 MB

0.36 MB

0.43 MB