Mother’s formal schooling—even at the primary level—is associated with lower risk of child mortality, although the reasons why remain unclear. This study examines whether mother’s reading skills help to explain the association in Nigeria. Using data from the Demographic and Health Survey, the analysis demonstrates that women’s reading skills increase linearly with years of primary school; however, many women with several years of formal school are unable to read at all. The results further show that mother’s reading skills help to explain the relationship between mother’s formal schooling and child mortality, and that mother’s reading skills are highly associated with child mortality. The study highlights the need for more data on literacy and for more research on whether and how mother’s reading skills lower child mortality in other contexts.
Demographers and policy makers agree that mother’s formal schooling—even at the primary level—is associated with lower risk of child mortality. There is less agreement about why this is the case. The fact that even a few years of primary school significantly increase the survival prospects of women’s children is puzzling. This study focuses on one mechanism that has received limited attention in demography: reading skills. The study explores the associations between mother’s formal schooling, reading skills, and child mortality in Nigeria, where approximately 10 % of the world’s child deaths occur each year (You et al. 2010).
Demographers rarely study literacy skills in their own right (notable exceptions include Glewwe (1999) and Thomas (1999)), but instead commonly assume individuals to be “literate” if they are “formally educated” (LeVine and LeVine 2001). In countries like Nigeria, where educational attainment is low and school quality is highly variable (Adekola 2007), such assumptions may be especially problematic.
Education research, however, suggests that women’s reading skills may uniquely benefit children’s health and survival by improving women’s cognitive comprehension of health information, and health communication skills (LeVine et al. 2012). Reading skills influence cognition broadly (Nicolopoulou and Cole 1999; Uhry and Ehri 1999); literate individuals perform better on tests assessing deductive reasoning, categorization, visuospatial discrimination, numeracy, and abstract speech than those who cannot read (Abadzi et al. 2005); and researchers have argued that these cognitive skills act as a “social vaccine” by encouraging individuals to adopt healthier behaviors and avoid health-related risks (Baker et al. 2011).
Beyond shaping the way individuals think, being able to read may operate at a more rudimentary level. Reading enables individuals to understand written health information, including banners and brochures, which are commonly used in public health campaigns. Reading skills are also linked to greater comprehension of auditory health messages, such as radio broadcasts (LeVine et al. 2004), which may be particularly relevant in developing contexts, where radio is a primary source of health information (National Population Commission NPC [Nigeria] and ORC Macro 2004). Being able to read health material and better comprehend radio messages may encourage women to adopt healthy reproductive behaviors that improve their children’s likelihood of survival (LeVine et al. 2004).
In addition to improving comprehension of health information, reading skills have been shown to increase individuals’ ability to communicate health concerns effectively. In comparison with those who cannot read, for example, literate individuals provide more organized and detailed accounts of child illness (Schnell-Anzola et al. 2005). Superior health communication skills may increase a mother’s level of comfort with professionals in institutional settings, such as family planning clinics and hospitals, thus increasing the likelihood that she will use such institutions for herself and her children (LeVine et al. 2012).
Researchers have extended this evidence to hypothesize that these literacy- and language-based skills improve women’s reproductive behaviors, thereby improving their children’s health and likelihood of survival (LeVine et al. 2012). However, the hypothesized associations among mother’s reading skills, reproductive behaviors, and child mortality have not been confirmed empirically. This study provides the first investigation of these relationships.
Data and Sample
I use data from the 2003 Nigerian Demographic and Health Survey (NDHS), which includes a sample of 12,076 children born to 4,579 women between 1993 and 2003. For two reasons, I restrict the analytic sample to children whose mothers never attended secondary school. First, the NDHS assumed that women who attended secondary school can read and thus did not administer the reading evaluation to them. Second, this sample choice allows me to assess the importance of reading skills among children who continue to have the highest rates of child mortality (i.e., those whose mothers have little or no formal school). Table 3 in the appendix demonstrates that this restriction results in socioeconomically disadvantaged children, including those with Muslim mothers and those living in rural communities in the northern region of Nigeria, being overrepresented in the final sample. It is important to keep this in mind when discussing the generalizability of these findings. I use the Stata 12 ICE program (Royston 2004) to multiply impute data for the 691 children (7 %) whose mothers had missing values, resulting in a final sample of 9,385 children born between 1993 and 2003.1
To calculate the hazard of child mortality, I use women’s retrospective birth histories of children born within 10 years preceding the survey. Following convention, the hazards represent the risk of death between birth and 59 months of age or between birth and the survey date in the case of children who had not yet reached age 5 in 2003.2
I measure mother’s reading skills using the NDHS’s interviewer assessment of whether respondents can read a sentence aloud in their preferred language. Interviewers categorized respondents as having no reading skills (did not read any of the words), some reading skills (read some of the words), or full reading skills (read every word). To validate the NDHS reading assessment, I compared the NDHS estimates with other published estimates of women’s literacy in Nigeria (see Table 4 in the appendix) and found that they are comparable in magnitude and follow the expected patterns by age and region.
I also include a set of classical reproductive factors (i.e., birth spacing, mother’s age, parity, and antenatal care), which are known to be associated with children’s risk of mortality (Mosley and Chen 1984). In the multivariate models, I assess whether inclusion of these reproductive behaviors attenuate the estimated association between mother’s reading skills and child mortality to evaluate whether they help explain the relationship. Table 1 provides descriptive statistics for all covariates used in these analyses by level of mother’s reading skills.
Figure 1 shows that in Nigeria, reading skills cannot be reasonably inferred from formal school background. Although there is a linear increase in the percentage of women who can read by the primary school level, many “formally educated” women have no reading skills. In fact, even among women who attended five years of primary school—a common international standard for classifying individuals as “literate”—fewer than one-half (45 %) are able to read at all.
Table 2 presents the results from the hazard models assessing the associations among mother’s formal schooling, reading skills, reproductive behaviors, and child mortality.3 Model 1 shows the zero-order relationship between mother’s formal schooling and child mortality risk. Corroborating prior findings, each year of mother’s primary school corresponds to an estimated 5 % reduction in the risk of child mortality.
Do mother’s reading skills help to explain the well-known protective effect of her formal schooling? Upon inclusion of the measure of reading skills in Model 2, the association between mother’s formal schooling and risk of child mortality is reduced by one-half and is no longer statistically significant. Compared with children whose mothers cannot read at all, those with mothers who have full reading skills have a 30 % lower risk of child mortality (p < .001). There is no evidence that children of mothers with some reading skills experience lower risk of mortality.
Model 3 demonstrates that the association between mother’s reading skills and child mortality is robust to a host of child, maternal, household, and contextual confounders. After more than a dozen such characteristics are controlled for, having a mother who can read is still associated with a 28 % lower risk of child mortality.
Based on the theorized links between mother’s reading skills and child mortality, Model 4 includes a set of classical reproductive behaviors (i.e., birth spacing, mother’s age, parity, and antenatal care) to determine whether these known proximate determinants of child mortality help to explain the association. The estimated association between mother’s full reading skills and child mortality is not meaningfully attenuated but instead remains significantly and largely associated with child mortality risk.
This study makes two contributions. First, the study demonstrates that researchers must abandon the common practice of assuming that “formally educated” individuals can read. The inclusion of a direct reading assessment in the DHS represents a shift away from this convention; however, the DHS continues to assume that individuals who ever attended secondary school can read. Reading skills do increase linearly with years of primary school; however, it is highly unlikely that the percentage of women with full reading skills increases from 30 % for those with six years of schooling to 100 % for those with one additional year of formal school. To accurately assess reading skills, demographic studies must directly assess reading skills regardless of school background.
Second, consistent with hypotheses generated by educational researchers studying the potential health implications of literacy, mother’s reading skills strongly and independently lower their children’s risk of mortality before age 5 in Nigeria. Two aspects of these results are particularly noteworthy: mother’s reading skills fully explain the survival advantage of mother’s formal schooling, and mother’s reading skills have a remarkably large, independent influence on child mortality. These findings signal literacy as a new area of research that has the potential to significantly advance understanding on education and health more broadly.
The analyses presented here reflect the situation of mothers with limited or no formal school in Nigeria—a population with exceptionally high rates of child mortality. Further research is needed to confirm the generalizability of these findings. This study also raises additional questions about exactly how mother’s reading skills reduce child mortality risk. Prior theories suggest that reading skills enhance women’s cognitive, health comprehension, and health communication skills, which encourage healthier reproductive behaviors that ultimately lead to lower child mortality. It is not possible to fully estimate the links of this hypothesized causal chain using the NDHS or any other available data source. However, the NDHS’s strong measures of classical reproductive behaviors (i.e., birth spacing, mother’s age, parity, and antenatal care) do not explain the association between mother’s reading skills and child mortality, underscoring the need to develop and test new conceptual models.
Although this single-country study provides only a first glance at the relationships among mother’s formal schooling, reading skills, and child mortality, it clearly establishes literacy as a new, fruitful line of inquiry in research on children’s health in low-income countries. Future research that replicates the protective value of mother’s reading skills for child health will confirm the need to fully incorporate direct reading assessments into demographic surveys. Sophisticated empirical efforts to identify the processes linking mother’s reading skills to child mortality will provide valuable insights on the relationships shown here. While leaving much to be done in the future, this study shows that research on health in low-income countries must not overlook the unique importance of literacy and calls for new approaches to conceptualizing and measuring literacy.
A previous version of this article was presented at the 2010 annual meeting of the Population Association of America in Dallas, Texas. I thank Jenny Trinitapoli and Michelle Frisco for their mentoring as I prepared this manuscript; and I thank three anonymous reviewers, Lauren Bachan, David Baker, Adam Lippert, and Molly Martin for their comments on earlier versions. I also acknowledge the support of the Predoctoral Traineeship in Family Demography (No. T-32HD 007514) by the Eunice Kennedy Shriver National Institute of Child Health and Human Development to the Pennsylvania State University Population Research Institute and assistance provided by the Population Research Center at Penn State University, which is supported by an infrastructure grant by the National Institutes of Health (R-24HD041025).
The results using imputed data are consistent with results using listwise deletion.
In supplementary analyses, I exclude children who died before 1 month of age because these deaths may be the result of complications that are less closely associated with maternal behaviors. These results (N = 8,843) show that the association between mother’s reading skills and child mortality is larger and stronger when early child deaths are omitted, suggesting that the results shown here downwardly bias the influence of mother’s reading skills on mortality risk in later childhood.
Briefly, the proportional hazards models assume that for an individual with a vector of covariates x, the hazard rate (death rate) at time t is , where is the hazard function at time t, and is a vector of unknown coefficients. The hazards model assumes the proportionality of the risk of child mortality. In other words, the outcome measure is time to death in the specified interval (0–59 months).