Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

Nigerian family models

As in the Javanese model, separate versions of the Yoruba model were tested in separate steps to examine the size and significance of the linkages in the separate models before the outcome variables were combined. The outcome variables are Growth (WAZ and HAZ), Mental Development Index (MDI), and Physical Development Index (PDI). In the Nigerian models, two-year-old children were examined (figs. 8.8-8.10); SQ was not measured in these children.

The general measurements of family-level functioning are comparable to the Javanese model. The family dynamic variables (family management and caring behaviour variables/parental care) consist of Academic Stimulation (measured by "teach" and "toys" observed variables); Affection and Attention (measured by "warm," "verbal responsiveness," and "index of child care support"); Feeding Practices (measured by "index of food investment" and "food belief"); and Hygienic Practices (measured by "index of sanitary practices," "water source," and "cleanliness of the environment"). All of these latent variables are hypothesized to influence outcome variables such as Growth and Mental Development directly.

Resource variables include Material (measured by "housing quality" and "food budget") and Social Resources (measured by "maternal literacy," "exposure to media," and "rural-urban location"). The way that these two latent variables affect parental care variables is postulated to be the same as those in the Javanese models. The assumptions made in the Javanese model also apply to the Nigerian model. For example, if Material Resources affects Academic Stimulation and Affection, it is postulated to be through a covariance pathway to Social Resources. Another resource variable is a latent variable, Modernization within the Family (measured by "modern behaviours" and "the intensity of the father's living in the house"). This variable is assumed to affect Feeding Practice and Affection. A covariance pathway was made between Material Resources and Modernization within the Family to allow a certain degree of association between these two latent variables.

Parental Care variables affecting child developmental outcomes are influenced by the characteristics of the child. The latent variable Child Personality is included in the model to control for this effect. Note that only covariance pathways link Child Personality and Parental Care variables (i.e. Academic Stimulation, Affection, and Feeding Practice); this means that Child Personality and Parental Care variables can be thought of as mutually related. Drawing pathways would produce too many paths to estimate precisely with the data available.

Almost all of our measured variables are strongly identified with the latent variables they represent. Weak measures include the measure of Hygienic Practices and Material Resources in all models. These are indicated by the high coefficients (from "housing" and "water source") that are not significant, indicating that their standard errors are high and the measurements unreliable. The latent variables, Material Resources and Hygienic Practices, are complex variables unlikely to be characterized easily by such relatively simple indicators.

Growth model

In the growth model (fig. 8.8), Social Resources are linked strongly to Parental Care variables, Affection, and Academic Stimulation. The significant effect of Social Resources on Affection contradicts our findings in the Javanese setting. This suggests that the expression of maternal affection and attention behaviours towards infants in Yoruba society shifts with formal education, whereas, in Java, affectionate behaviour is internalized early during the mothers' childhood; therefore, education may have little impact on this behaviour in the Javanese society. Our finding is also consistent with the findings of LeVine, Klein, and Owen (1967) in Nigeria, that more-educated parents are more friendly and demonstrate more affection towards their children. The expression of Affection and Attention is strongly ( #) linked to Growth. Academic Stimulation, however, only marginally affects Growth. This is consistent with our findings in Java, even though the age of the children differs between the two data sets.

Fig. 8.8 Path diagram of growth, Nigeria. N=170 2-year-old and their families. GFI (adjusted) = 0.754;BB Index = 0.641. Ovals: latent variables; rectangles: measured variables; statistical significance of path coefficients: *0.025 < p < 0.050; D 0.010 < p < 0.025;# p < 0.010

The effects of Material Resources on Feeding Practices were not significant at this sample size. Feeding Practices showed a strong effect (D ) on Growth. This is not surprising, given the supposition of a direct linkage between Feeding Practices and Growth.

As expected, Modernization within the Family is linked to Maternal Affection. The extent to which fathers live in the house, and modern child-rearing that extends the relative indulgence of infancy, affect the way the mother expresses her affection and attention. Even though Social Resources has been controlled for, this does not alter the positive effect of Modernization variables on Maternal Affection. Modernization does not have a significant effect on Feeding Practices.

The link between the Hygienic Practices variable and Growth does not have the expected sign and is statistically insignificant. As with the Health Practices variable in the Javanese model (measured by immunization and vitamin A supplementation), conceptually there is a potential simultaneity problem between Hygienic Practices and Growth. Also, as in the Javanese case, the Health Practices variable is not a good indicator in this model, making the underlying relationship unreliable. Another possibility is that the Health Practices variable cannot be thought of as directly affecting Growth. There may be overlooked factors mediating this linkage that cannot be included in this model.6

Development model

In figure 8.9, two observed variables, the PDI and the MDI, were used to measure the Mental Development construct. Before we combined the two variables into a single index, we estimated separate models with each variable. All the coefficient estimates are similar for these two models; therefore, these two variables represent a single concept called Mental Development.

The strong relationships between Social Resources and Academic Stimulation and Affection are again demonstrated in this model (fig. 8.9). As in Javanese society, more-educated parents provide more academic encouragement for their children. The effect of Academic Stimulation on Mental Development is significant. Even though the age category is different between the Javanese and the Nigerian samples, Academic Stimulation in both data sets appears to be important. The latent variables Affection and Attention and Feeding Practice turned out to be the most important factor in determining the child's Mental Development (#).

Fig. 8.9 Path diagram of mental development, Nigeria. N=170 2-year-old and their families. GFI (adjusted) = 0.736;BB Index = 0.569. Ovals: latent variables; rectangles: measured variables; statistical significance of path coefficients: *0.025 < p < 0.050; D 0.010 < p < 0.025;# p < 0.010

We found in the Javanese model with older children that the effects of Maternal Affection and Feeding Practices on Mental Development were not direct, but were mediated by Growth. Perhaps learning that occurs rapidly in early life is characterized primarily by strong maternal and child bonds. The physical environment, such as the provision of toys and attempts to teach the alphabet, has a less important role than Maternal Affection in shaping Mental Development in the early first or second year of a child's life. After the critical period of emotional bonding in early childhood, Academic Stimulation will have a stronger impact on Mental Development than Maternal Affection, as was found in older children in the Javanese model. This is the period when a child begins to be an independent person, and is less attached to the mother.

Other path coefficients and significance levels were quite similar to those in the growth model. We take this stability as a sign of the validity of the model. When the model is run using different outcome indicators, the coefficients shift very little (except for Academic Stimulation, which shows a stronger effect in the development model).

Overall child development model

In previous steps, the nature of the relationships between family-level variables and each outcome indicator was estimated separately. Those models reveal that the coefficient for each pathway and the relative ranking of coefficients are similar. This leads us to believe that Mental Development and Growth should be combined to represent overall Child Development.

Before making such a decision, we attempted to construct a model with two outcome indicators, Mental Development and Growth, in the same model. This produced so many pathways linking Parental Care variables and outcome indicators that we failed to obtain a SAS program that converged into a stable solution. Another similar model was tested linking Academic Stimulation to Mental Development only, and the rest of the pathways stayed the same: again, the convergence failed. We cannot design a model that is the same as the Javanese overall child development model because the nature of the linkages differs between the two settings.

Examination of the Pearson correlation coefficients between all outcome indicators (WAZ, HAZ, MDI, and PDI) reveals that they are highly correlated (more than .33). The correlations among outcome indicators in older children (in Java) are weaker (HAZ and SQ and HAZ and IQ are .24 and .26, respectively), and are not significant between WAZ and SQ or IQ. Perhaps, in a rapid growth period, Growth and Mental Development are strongly mutually related: the development of each will reinforce the development of the other. This pattern seems to be less clear in older children, in which it appears that each outcome indicator becomes a more separate entity, although they may be mutually related. In addition, based on our analysis in the Javanese model, some Parental Care variables behave differently in each child development model. Therefore, we believe that Growth and Mental Development represent a single concept in younger children, which we call Overall Child Development. This construct was used in subsequent work.

This model (fig. 8.10) again shows the strong effect of the latent variable Affection and Attention on Overall Child Development. It indicates that a warm and attentive mother is clearly beneficial in producing positive child development. The latent variable, Academic Stimulation, had significant effect but was weaker than the effect of Affection and Attention and Feeding Practices in this model.

The effect of the latent variable Feeding Practices is very strong ( # ), and stronger than in the older children's model in Java (*). This suggests that good feeding practices are important in rapidly growing children.

The latent variable Modernization within the Family behaves in relatively the same way as in the previous model, but has a weaker effect in the overall model (*).

There is a strong reciprocal influence between Child Personality and Maternal Affection (and also Feeding Practices). This association appears consistent in all models. Affectionate mothers are associated with good Child Personality (measured by such factors as smiling and ease of handling), and vice versa. A smiling baby may be attractive to the mother, encouraging her to stay with the infant and play with it. This interlude may help to cement the bond between the child and its mother, producing a powerful reciprocal attachment that further affects positive child development outcomes.

Besides the strong effect of maternal affection on child development, its strong effect on maternal bonding also can influence the child's personality later in life. According to Morris (1969), this bond is extremely important. A bond developed in the first year of life will imprint a large capacity for making strong bonds during adult life.

Fig. 8.10 Path diagram of overall development, Nigeria. N=170 2-year-old and their families. GFI (adjusted) = 0.754; BB Index = 0.641. Ovals: latent variables; rectangles: measured variables; statistical significance of path coefficients: *0.025 < p < 0.050; D 0.010 < p < 0.025;# p < 0.010

As in the Javanese model, we failed to find an association between the Hygienic Practices variable and all outcome indicators. Future efforts to improve methods of gathering Health and Hygienic Practices information are needed.

We conclude that after controlling for SES, the quality of care including Maternal Affection (which includes verbal responsiveness), Feeding Practices and Academic Stimulation positively affect all outcome indicators. Even though Academic Stimulation shows a relatively strong effect on all outcome indicators in younger children, its effect is weaker than the effect of Maternal Affection and Feeding Practices.

Total effects of indicator variables

A summary of the findings from the models is presented in tables 8.5-8.8. The coefficients of most variables are relatively stable from one model to the next.

Note that three factors emerged in all models as having the greatest effect: they are Maternal Affection, Feeding Practices (and Academic Stimulation in the mental development model), and Social Resources. The negative effect of Material Resources is unreliable in this model because we do not have good measures of this factor. Also, Material Resources are hypothesized to influence outcome indicators through Hygienic Practices. Since we could not obtain a good indicator of Hygienic Practices, and its effect is negative on child developmental outcomes, the total effect of Material Resources through Hygienic Practices became negative.

Table 8.5 Total effects in Nigerian growth model

  Stimulation Affection Feeding Hygiene Growth
Social .286 235 - - .268
Material - - .201 .606 - 2.630
Stimulation - - - - .320
Affection - - - - .751
Feeding practices - - - - .928
Hygiene practices - - - - - 7.449
Modern family - .585 .298 - - .716

Table 8.6 Total effects in Nigerian mental development model

  Stimulation Affection Feeding Hygiene MDa
Social .331 .223 - - .306
Material - - - .325 .284 - .481
Stimulation - - - - .384
Affection - - - - .835
Feeding practices - - - - .549
Hygiene practices - - - - - 20.565
Modern family - .004 2.845 - 2.184

a. Mental development (MDI and PDI).

Table 8.7 Total effects in Nigerian overall child development model

  Stimulation Affection Feeding Hygiene All deva
Social .288 .240 - - .362
Material - - 1.880 .604 - 3.169
Stimulation - - - - .406
Affection - - - - 1.023
Feeding - - - - 1.136
Hygiene practices - - - - - 8.792
Modern family - -.136 .291 - .900

a. Overall development.

This study again demonstrates the importance of the dynamics within the family in promoting positive development. Improving family resources alone, a point that is often emphasized by the economist and by those involved in development, cannot guarantee improvements in child welfare unless Family Management, Beliefs, and Caring Activities are favourable for this purpose.

Compared with the Javanese model, in which Social Resources has a small effect on the expression of Maternal Affection, this factor also is dominant in affecting positive child developmental outcomes. This suggests that, in the Javanese society, improvements in maternal education cannot be relied upon to change maternal expression of affection because of maternal childhood experience in adopting this attitude (ch. 6); this is not true in Yoruba society. Improving maternal attitudes and behaviours should go hand in hand with the improvement of the educational level of the family; however, improving social resources is very important in order to increase the awareness of parents to stimulate their children academically, in both cultures.

Table 8.8 Legend for Nigerian models

Variable code Description
LITERACY Mother's last year of schooling and her literacy (reads Yoruba and reads English)
MEDIA Mother's exposure to media
LOCATION Location (rural, semi-rural, and urban)
HOUSING Housing quality
FDBUDGT Total food budget
FOODINVEST Food investment score for the child
FOODBELIEF Food belief
MODBEHAV Modern behaviour score
PALIVE Father lives with mother
TEACH Teach child
TOYS The amount of toys score
WARM Maternal affection
VERBAL Maternal verbal responsiveness
CHCARE Child-care arrangements
CLEAN Cleanliness of the environment
SANITA Hygiene practices
WATERSRC Water source
ELICBABY Eliciting baby
COOPBABY Easiness to be handled
MDI Bayley mental development index
PDI Bayley physical development index
WAZ Weight-for-age Z-score
HAZ Height-for-age Z-score


1. Error terms commonly represented by Es and Ds have been removed from our diagrams for simplicity of visual representation, but are taken into account in the analysis.

2. The RAM specification of the PROC CALIS was used from SAS Version 6.01. The doc umentation for this procedure is available on pp. 292-365 of the SAS/STAT User's Guide
(SAS Institute Inc. 1989).

3. Polychoric correlation would be used to estimate the correlation based on a pair of catego rical variables, and polyserial correlation would be used when one variable was continuous and the other was categorical.

4. Some of our original analysis attempted to link the Material Resources and Social Resources variables with all aspects of parental care. Since there is some correlation between these variables, this approach produced too many pathways to estimate precisely with the data available.

5. A pathway between two variables with arrows at each end represents correlation between the two variables.

6. A trial was made to include morbidity variable as a mediator variable. This produced too many factors to estimate precisely with the data available. Also, the simultaneity problem between growth and morbidity needed to be solved, making the model too complicated, given the limited sample size.

Contents - Previous - Next