1.1. Parental employment and child nutritional outcomes

To the best of our knowledge, no studies to date have examined the associations between parental occupation and child development in LLMICs. Instead, numerous studies have examined the association between maternal employment and child nutritional status and yielded mixed results. In some cases, maternal employment is associated with improved child nutritional outcomes [6–8], whereas in other cases maternal employment is associated with poor child nutritional outcomes [9–14]. Still other studies have found no significant associations between maternal employment and child nutritional outcomes [15–17]. Similarly, findings with respect to maternal occupation type are also mixed. In some contexts, maternal agricultural and manual employment are associated with poor child nutritional outcomes [9], whereas in others, maternal non-agricultural employment is associated with poor child nutritional outcomes [18].

Importantly, none of these studies in LLMICs considered the associations between paternal employment or occupation type and child outcomes. This is despite prior work highlighting that fathers are often the primary breadwinner of the family, providing substantial financial resources that can support positive outcomes for all household members [19, 20]. Only a few known studies have investigated the role of both maternal and paternal employment in child outcomes [21, 22]. The emerging evidence is largely inconclusive and none of the studies considered child development as an outcome. For example, one study from the United Kingdom found that employment of either parent was associated with improved child nutritional outcomes, compared to both parents being unemployed [21]. Another study from China found that associations differed depending on which parent was employed: paternal unemployment was negatively associated with child health, while maternal unemployment was positively associated with child nutrition and health [22]. However, this nascent literature from upper-middle and high-income countries may not be generalizable to LLMICs, where employment opportunities are more informal and of lower wages [23] and where agricultural employment is the predominant type of labour (49% in LLMICs vs. 21% in upper-middle income countries vs. 3% in high-income countries [5]).

These mixed findings are not surprising given the complexity of the relationships between parental occupation and child nutritional outcomes, varying study methodologies, and the competing mechanisms through which parental employment and occupation type can influence child outcomes (see Section 1.2). Given that family and ecological caregiving environments are similar for child nutrition and development [24], associations between parental occupation and child development are plausible and equally important to unpack.

1.2. Hypothesised mechanisms

Parental occupation can influence child development through five potential mechanisms: (1) household income, (2) women’s empowerment, (3) childcare, (4) parental physical and mental health, and (5) pesticide exposure. The relative importance of these mechanisms likely varies by occupation type. In LLMICs, agricultural employment is common with 60% of working adults (63% of women, 57% of men) in low-income countries employed in agriculture and 38% (42% of women, 36% of men) in lower-middle income countries [5].

First, parental employment increases household income [11, 25], and thus the availability of additional resources to allocate towards child health, nutrition, and development [26, 27]. Families with higher income may have more and better access to child healthcare services in times of child illness [28] and financially provide more enriching opportunities for early learning like preschool fees and toys [1].

Second, parental employment can improve child development by increasing women’s empowerment. Extensive empirical literature from LLMICs has demonstrated that women’s employment is associated with greater women’s empowerment [29–31]. In fact, women’s employment, seasonality of employment, and type of remuneration (cash vs. in-kind) are often included as indicators in composite measures of women’s empowerment [27, 32, 33]. Further, evidence suggests that women’s non-agricultural employment is associated with greater women’s empowerment relative to agricultural employment [34–36], likely because non-agricultural employment allows women to learn non-farm skills and exposes them to knowledge and information which can improve their household decision-making [34]. More empowered women allocate more resources towards their children, which in turn support better child development [27, 37]. Of note, prior literature on women’s empowerment has largely been in the context of women’s employment, despite scholars highlighting the need for research on employment and empowerment to include both men and women [38].

Third, parental employment can influence child development by impacting childcare arrangements, such as reducing the amount of time parents spend with their children and/or increasing reliance on alternative caregivers within (e.g., other adults or children) or outside (e.g., preschool) the household [7, 11, 39]. Given the morning and seasonal nature of farming, parents may have limited time with their children and require alternative non-parental childcare (e.g., supervision by older siblings). Evidence indicates different time patterns in childcare between working and non-working parents [6, 40–43]. However, the relationship between parental childcare time patterns and child development is complex and can vary depending on how parents manage time trade-offs and negotiate caregiving activities [43]. Studies of women’s time use in agriculture show that women in agricultural settings face more severe time constraints and stricter trade-offs than women in non-agricultural settings [44, 45], which could adversely affect child development. Finally, considering that fathers are the primary breadwinners in many LLMICs contexts and can positively influence child development [46], there may also be interactions between maternal and paternal occupation with respect to childcare. However, evidence considering both maternal and paternal work and childcare is lacking from LLMICs.

Fourth, employed women have to balance paid work and unpaid childcare and household work, which can compromise their physical and mental health and, in turn, affect their ability to care for themselves and their children [40]. The potential adverse effects of employment on parental health are likely not unique to women, though they are probably exacerbated relative to men given employed women’s dual burden of occupational and household work [47]. Stark gender inequities in the distribution of other household responsibilities remain, with women largely responsibility for all household chores, including physically intensive chores like collecting water and firewood [40]. Moreover, parental health is likely worse among parents employed in agriculture, which is generally more physically taxing than other occupations.

Lastly, pesticide exposure is a unique mechanism through which agricultural work can influence child development. Pesticides are widely used in agricultural contexts in LLMICs, with higher exposure due to continued use of harmful pesticides banned in high-income countries and unsafe handling and application practices [48]. Children of parents engaged in agriculture can be exposed to pesticides by: (1) spending time on the farm where they may inhale pesticides directly during spraying or ingest them by touching objects and putting them or their fingers in their mouths; or (2) consuming contaminated foods or water [49, 50]. Extensive evidence has linked pesticide exposure to suboptimal child development, largely through inhibition of acetylcholinesterase activity [51, 52]. Pesticide exposure can also adversely affect parental health [53] and compromise parents’ ability to care for their child.

1.3. Current study

Given the high proportion of developmentally off-track children in LLMICs and employed parents, unpacking the relationships between parental occupation and child development is crucial. Understanding the strength and direction of these relationships can help inform the design and targeting of interventions to support working parents with young children. Therefore, in this paper, we used nationally representative data from nine Demographic and Health Surveys (DHS) to investigate the associations between parental employment and child development, and the role of childcare and women’s empowerment as potential mechanisms. Given the high proportion of adults employed in agriculture in LLMICs, we compare agricultural to non-agricultural employment. Based on our review of the literature and mechanisms (Sections 1.1 and 1.2), in this analysis, we tested five exploratory hypotheses. Due to the limited literature considering both maternal and paternal employment, we had no a priori hypotheses about parental sex and occupation type. Compared to non-agricultural employment, we hypothesised that parental agricultural employment is associated with:

1. poorer child development (with the poorest outcomes among families where both parents are employed in agriculture relative to both parents employed in non-agriculture)
2. lower women’s empowerment
3. less child stimulation
4. less adequate supervision
5. less attendance in early childhood care and education programmes (ECCE)

We build on existing literature by examining child development as our primary outcome and considering both maternal and paternal occupation as determinants.

2.1. Data and study population

We pooled data from all DHS that collected information on parental employment and child development and were publicly available as of February 2022. All DHS collect data on women’s employment in the last 12 months in the woman’s interview. For a random sub-sample of households, an adult man is also interviewed, and the same employment questions are asked during the man’s interview. A couples’ file is then generated by the DHS Program pairing women and men who are married or cohabitating. With respect to child development, this optional module is applied to the youngest child aged 36–59 months, and collects data on children’s attainment of developmental milestones, child stimulation, child supervision, and ECCE attendance. Since the child development module is optional, we could only include surveys which collected this module. However, within a DHS survey, the sub-samples of households completing the child development module and the man’s interview are not always overlapping. Therefore, not all DHS surveys collecting data on child development contributed data to our analysis. In total, we included nine DHS spanning 2011–2020: Benin, Burundi, Cambodia, Congo, Haiti, Rwanda, Senegal, Togo, and Uganda (S1 Table). These were all the countries with an overlapping sub-sample of households with data on child development and paternal employment.

2.2. Measures

We created two variables for parental employment and occupation in the last 12 months. First, we created a binary variable for whether each parent was employed in the last 12 months (employed vs. unemployed). Second, for those employed in the last 12 months, we created a categorical variable for whether one or both parents were employed in agriculture (occupation): (1) mother employed in agriculture, father employed in non-agriculture, (2) mother employed in non-agriculture, father employed in agriculture, (3) both parents employed in agriculture, and (4) both parents employed in non-agriculture. Non-agricultural occupations included: clerical, sales, services, professional/technical/managerial, household and domestic, skilled manual, and unskilled manual occupations. These groupings are pre-specified by the DHS Program [54].

Child development was assessed using the Early Childhood Development Index (ECDI), a population-based measure designed to assess four domains of development in children aged 36–59 months: cognitive, socio-emotional, literacy-numeracy, and physical [55]. The child’s mother reports on whether the child can perform each of ten developmental milestones. Per the ECDI guidelines, we first created binary indicators for whether children attained each one of the ten developmental milestones. We then constructed binary indicators for whether children were developmentally on-track in each domain and for whether children were overall developmentally on-track (on-track in at least three out of the four domains). We also constructed the count ECDI score as the number of milestones achieved by each child, range 0–10 [55].

Child stimulation was assessed using three indicators for the number of stimulation activities (range 0–6) provided by the mother, father, or other household members in the last three days (all based on maternal report) [56]. The six activities were: (1) reading books/looking at pictures, (2) telling stories, (3) naming/counting/drawing, (4) singing, (5) taking the child outside, and (6) playing with the child. The adequacy of child supervision over the last week was assessed using three indicators: (1) child was not left alone for >1 hour, (2) child was not left under the supervision of another child for >1 hour, and (3) child was not left alone or under the supervision of another child (referred to as “adequate supervision” for brevity) [56]. ECCE attendance was assessed using a single indicator for whether the child attended an organised learning or ECCE programme.

We assessed three dimensions of women’s empowerment: (1) access to and control over resources, (2) decision-making, and (3) attitudes towards wife beating. Factor scores for each dimension were derived from a form-invariant model using confirmatory factor analysis. We also calculated a total women’s empowerment score as the sum of the three dimensions’ factor scores. The only indicator related to women’s employment included in the derivation of the “access to and control over resources” dimension was an indicator for seasonality of employment (throughout the year vs. seasonal/occasional). Full details on the indicators comprising the domains and the derivation of the factor scores have been previously published [27].

2.3. Statistical analysis

We restricted the analytic sample to children aged 36–59 months with available data on child development. We merged the couples and child files to create a mother-father-child triad, and we therefore refer to women and men as mothers and fathers. Since only 97 (1.11%) fathers were not employed in the last 12 months, we restricted the sample to households where the father was employed. We tested for differences between included and excluded households using a Wald test, considered significant at p<0.05. Excluded households with unemployed fathers were generally similar to included households with employed fathers, except for the former provided less stimulation (S2 Table). We further excluded households with missing data on parental occupation (N = 142, 1.64%). Women in households without data on parental occupation were less empowered and more likely to have no education compared to women in households with data on parental occupation (S2 Table). In addition, fathers in households without data on parental occupation were older and more likely to have no education than fathers in households with data on parental occupation. Excluded households were also more likely to live in an urban area and were somewhat wealthier than included households (S2 Table). The final analytic sample included 8,516 children aged 36–59 months with data on child development and maternal and paternal occupation.

We used generalised linear models to assess the associations of interest. First, given that all fathers in our sample were employed, we examined the associations between maternal employment and child development (the outcome) and maternal employment and child stimulation, child supervision, ECCE attendance, and women’s empowerment (the potential mechanisms). Then, among employed parents, we examined the associations between parental occupation and child development, and parental occupation and child stimulation, child supervision, ECCE attendance, and women’s empowerment. We used log-Poisson models for the binary variables (overall development on-track, child supervision, and ECCE attendance) and calculated unadjusted and adjusted relative risk (RR) and 95% confidence intervals (CIs). For child stimulation and women’s empowerment, we used linear models and calculated unadjusted and adjusted mean differences (MD) and 95% CIs. Adjusted models controlled for child age and sex, maternal age and education, paternal age and education, household size, wealth, and location (urban vs. rural). Standard errors were clustered at the primary sampling unit level. Models were weighted for representativeness using country-specific weights. Missing data on mechanisms (N = 64; 0.75%) were imputed using country-specific mean imputation.

In models including the four-category variable for parental occupation, we considered the category “both parents employed in non-agriculture” as the reference category. We tested for equality across exposure categories using a Wald test. In all models, the Wald tests indicated differences across exposure categories (all p-values <0.001). Given this non-equivalence, we then re-estimated the models changing the reference category to “both parents employed in agriculture” to assess whether associations differed depending on which parent was employed in agriculture. Thus, for a given outcome, the models using the categorical exposure were the same, except for this change in the reference category to assess the different associations.

Given the role of child nutrition (safe and nutritious foods) in supporting child wellbeing [57] and evidence that severe childhood malnutrition can impair child development [58], we conducted a sensitivity analysis accounting for child wasting (defined as weight-for-height Z-score <-2 SD) in the analysis. We first examined biserial correlations between child wasting, parental employment, and child development in the pooled sample and separately by country. We then included child wasting as a control variable in the adjusted models for the associations between maternal employment, parental occupation, and child development. Of note, child wasting was not available for the sample of children with data on child development and maternal and paternal occupation in Cambodia and Congo. Therefore, these sensitivity analyses were conducted on a dataset pooling the remaining seven countries.

To explore whether adjusted associations between parental occupation and child development, child stimulation, child supervision, ECCE attendance, and women’s empowerment differed across household location (urban vs. rural), household wealth, maternal and paternal education, and country income level (lower income vs. lower-middle income), we included an interaction term between the categorical exposure variable and each of these modifiers. Given the exploratory nature of these analyses, interactions were considered significant at p<0.10 based on a Wald test. All analyses were conducted in Stata 17 [59].

2.4. Ethical considerations

Ethical clearance for each DHS is granted by the relative institutions in the respective country. DHS data are publicly available de-identified secondary data and thus exempt from further ethical review. Access and permission to use the data for the present analyses was granted by the DHS Program (http://www.dhsprogram.com).

3.1. Sample descriptives

Children in our sample were 46.6 months of age on average and 49% were girls (Table 1). All fathers and 85% of mothers in our sample were employed in the last 12 months. Among employed parents, 44% of mothers and 53% of fathers were employed in agriculture. Child development was generally poor with 40% of children developmentally off-track. Childcare practices and women’s empowerment were also suboptimal. Child, maternal, paternal, and household characteristics by maternal employment and parental occupation are shown in S3 Table. Both parents were engaged in agriculture in 35% of the poorest households and 3% of the wealthiest households. Both parents were engaged in non-agriculture in 8% of the poorest households and 38% of the wealthiest households (S3 Table). Overall, households where both parents were employed in agriculture appeared poorer and less educated than households where only one or neither parent was employed in agriculture. Further, both agricultural and non-agricultural occupations differed by parental education, household wealth, and household location (S4 Table). The proportion of mothers and fathers employed in agriculture was lower among more educated individuals, wealthier households, and urban households. Respectively, the proportion of parents employed in professional/technical/managerial, sales, or skilled manual occupations was higher among more educated individuals, wealthier households, and urban households (S4 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Household, parental, and child characteristics of the 8,516 children aged 36–59 months in the analytic sample.

https://doi.org/10.1371/journal.pgph.0001116.t001

Overall, 2.4% of children in our sample were wasted, ranging from 0.6% in Rwanda to 10% in Senegal. In the pooled sample, child wasting was significantly correlated with maternal employment (p<0.01); however, the magnitude of this correlation was small (-0.03). Child wasting was not correlated with child development (overall or by domain) in our sample (all p-values >0.3 and all correlations were between -0.01 and 0.01).

3.2. Associations between maternal employment and child outcomes

Maternal employment was not associated with child development in adjusted models (Table 2). Results were consistent in sensitivity analyses controlling for child wasting (RR 1.01 (95% CI 0.95, 1.08)). However, maternal employment was associated with more stimulation provided by fathers and other household members, but with less adequate child supervision (Table 2). Maternal employment was not associated with stimulation provided by the mother or ECCE attendance. Maternal employment was positively associated with women’s total empowerment, access to and control over resources, and decision-making (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Associations between maternal employment and child development, childcare practices, and women’s empowerment¹.

https://doi.org/10.1371/journal.pgph.0001116.t002

3.3. Associations between parental agricultural employment and child development

Among employed parents, children whose parents were employed in agriculture relative to non-agriculture were less likely to be developmentally on-track (Table 3). Results were consistent in sensitivity analyses controlling for child wasting (RR 0.85 (95% CI 0.79, 0.92)). More specifically, overall ECDI score, cognitive, socio-emotional, and literacy-numeracy development were poorer among children whose parents were both employed in agriculture relative to non-agriculture (S5 Table). When assessing whether associations differed by which parent was employed in agriculture, we observed that children were more likely to be developmentally on-track if only one parent was employed in agriculture as compared to both, regardless of whether it was the mother or father (S6 Table). The one exception was that children were more likely to be cognitively on-track if only the father was employed in agriculture relative to both parents (S6 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Associations between parental occupation and development among children aged 36–59 months¹.

https://doi.org/10.1371/journal.pgph.0001116.t003

3.4. Associations between parental agricultural employment and hypothesised mechanisms

With respect to the potential mechanisms we examined, parental agricultural employment relative to non-agricultural employment was not associated with stimulation by mothers or fathers (Table 4). However, stimulation by other household members was higher (15% additional activities) when both parents were employed in agriculture compared to when both parents were employed in non-agriculture (Table 4). Stimulation by other household members was significantly lower when only one parent was employed in agriculture relative to both parents (S7 Table). In addition, stimulation by mothers was higher if only the mother was employed in agriculture relative to both parents (S7 Table). Further, parental agricultural employment relative to non-agricultural employment was associated with inadequate child supervision, particularly leaving the child with an older sibling (Table 4). Parental agricultural employment was also associated with less ECCE attendance than parental non-agricultural employment. These associations with child supervision and ECCE attendance were largely driven by paternal employment in agriculture (S7 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Associations between parental occupation and childcare practices and women’s empowerment among children aged 36–59 months¹.

https://doi.org/10.1371/journal.pgph.0001116.t004

In addition, we found that parental agricultural employment was negatively associated with women’s empowerment relative to parental non-agricultural employment: in households where both parents were employed in agriculture, women had lower scores on total empowerment, access to and control over resources, and attitudes towards wife beating (Table 4). These associations appeared to be largely driven by maternal agricultural employment (S9 Table).

3.5. Heterogeneity of associations

We found that the magnitude of the associations between parental occupation (both in agriculture vs. both in non-agriculture) and child development was larger among less educated parents and in low income countries (p-values for interaction <0.10) (Fig 1, S9 Table). There was also evidence that household location (urban vs. rural), parental education, and country income level modified the associations between occupation and stimulation (p-values for interaction <0.10). We observed less beneficial associations for stimulation by the mother among rural households and more educated parents, a less beneficial association for stimulation by the father in lower-middle income countries, and a more beneficial association for stimulation by other household members among rural households and in low income countries (Fig 2, S9 Table). Further, household wealth and country income level modified the association between occupation and adequate supervision, whereas maternal education and country income level modified the association between occupation and ECCE attendance (p-values for interaction <0.10). However, no clear patterns appeared by household wealth quintile or maternal education level. With respect to country income level, associations generally appeared less beneficial among lower-middle income countries. Lastly, household location, household wealth, and country income level modified the associations between parental occupation and women’s empowerment (p-value for interaction <0.10) with larger, more negative associations among rural household, wealthier households, and lower-middle income countries (Fig 2, S9 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Heterogeneity of the associations between parental occupation and child development by household and parental characteristics, comparing households where both parents were employed in agriculture and households where both parents were employed in non-agriculture.

Models adjusted for child age and sex, maternal age and education, paternal age and education, household size, wealth, and location (urban vs. rural), and accounted for representativeness. SEs were clustered at the primary sapling unit level.

https://doi.org/10.1371/journal.pgph.0001116.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Heterogeneity of the associations between parental occupation, childcare practices, and women’s empowerment by household and parental characteristics, comparing households where both parents were employed in agriculture and households where both parents were employed in non-agriculture.

Models adjusted for child age and sex, maternal age and education, paternal age and education, household size, wealth, and location (urban vs. rural), and accounted for representativeness. SEs were clustered at the primary sapling unit level.

https://doi.org/10.1371/journal.pgph.0001116.g002