Background
In the past decade, child protection indicators such as birth registration (BR)(
1 BR is defined as “the continuous, permanent, and universal recording within the civil registry, of the occurrence and characteristics of births in accordance with the legal requirements of a country” [
1].) have been incorporated into the development goals agenda [
2]. For example, target 16.9 of the Sustainable Development Goals (SDG) suggests a
legal identity for all by 2030. The United Nations Convention on the Rights of the Child (UNCRC) notes that BR is a child’s ‘first right’ as it provides the legal ‘membership card’ to participate in society [
3,
4]. In addition, it is considered critical to improvements in human capital outcomes - utilization of healthcare services and educational attainment (school enrolment, progression and completion [
5]. Without being registered, a child will be born and die without any trace of their existence documented officially. This term is known as the
scandal of invisibility – being unseen, uncountable, and hence uncounted [
6,
7].
Universal BR is a public good that is used to monitor the progress of development goals and the functioning of the rule of law in modern societies [
1,
8]. As a proof, registered children are issued a
birth certificate, which documents their right to an identity and ensures that their human rights are not overlooked when violated or abused [
5,
9,
10]. In some countries, birth certificates are required for enrolment, taking tests and examinations, scholarship applications and graduation for primary and secondary schools [
11]. This valid proof of age can thus mitigate long-term risks by making it easier to claim rights and/or privileges in the future, e.g., formal labour market participation, civil marriage registration, bank account ownership, owning/inheriting land and housing, access to social welfare programs (e.g. scholarships, food assistance, insurance and pensions), voting and obtaining a passport for international travel [
12‐
15].
In Nigeria, 70% of children under five years of age are unregistered, and 83% of them do not have birth certificates, according to a recent household survey [
16]. The rates of non-documentation of births in Nigeria are among the world’s highest, and rank higher than those in Sub-Saharan Africa (59%) and West and Central Africa (55%) [
17]. To address the ‘scandal of invisibility’, it is important to identify the protective and risk factors of birth registration to facilitate well-targeted policies. For example, some children may not be registered due to the low number of BR centres in their community (low supply), while others may be due to being born into low-income families or poorly educated parents (low demand).
Extant studies suggest that the proximate determinants of child survival are most likely related to the underlying factors that affect a child’s probability of having their births registered or certified [
18]. Some of the identified socioeconomic and demographic (SED) factors deemed important for BR include
the gender of the child (Nigeria) [
19];
utilization of perinatal health services (Edo, Nigeria; Nusa Tenggara, Indonesia) [
20,
21];
mother’s age at child’s birth (Ghana; Western Australia; Bolivia, Brazil, Colombia, the Dominican Republic, Peru, and Nicaragua) [
22‐
24];
maternal health-seeking behavior (Bauchi and Cross-River, Nigeria) [
25];
parent’s literacy and education (Lombok, Indonesia; Ghana) [
26,
27];
wealth (Indonesia; Dominican Republic) [
6,
28]; and
geographical location (Bolivia, the Dominican Republic, and Peru) [
29]. It is important to note that most of the existing studies on the determinants of birth certification are beyond the Nigerian context. Analysing how the identified determinants or factors play out in the Nigerian setting therefore makes an important contribution.
The objective of the current study is two-fold and aims to provide answers to two questions: (1) What is the pattern and coverage of birth certification and (2) what individual and community factors are associated with birth certification for children under-five years of age? To achieve the stated objectives, the study employs descriptive statistics and multilevel regression models to analyse a pooled sample of data of children under-five years from the last three rounds of the nationally representative NDHS data. The adoption of the multilevel regression analysis provides insight into what level of factors are most critical in driving BC in Nigeria. The study makes two key contributions to the literature. First, it adopts birth certification (BC) instead of BR (
2 Birth certification and registration rates may differ for several reasons. First, parents may begin but not complete the registration process. This may be due to issues with payment of fees. While the Nigerian law states that the birth registration should be free at least for newborn children, some registration offices charge fees for this. Another reason could be linked to the fact that some mothers mistake the attestation of a live birth for the birth certification. Finally, some parents could have lost or have not picked up the birth certificate and so could not provide it at the time of the interview. These reasons are currently speculative and would need further research as evidence) because the latter is based on the parent/caregiver report and the outcome is open to bias. Second, the study captures supply-side factors e.g., the availability of civil registration centres, which have been missing in earlier studies. The lack of an enabling environment for birth registration within or close to a community can increase financial and opportunity costs, and hinder the efforts of parents/caregivers to register and certify their children’s births [
29,
30]. Third, among the existing studies on Nigeria, none analysed the SED characteristics using different waves of the Nigerian Demographic & Health Survey (NDHS) while simultaneously accounting for the hierarchical structure of the data. Pooling data helps increase the sample size, and the enhanced statistical power allows for analyzing the factors associated with BC for comparable children across communities. In the NDHS data, individuals are nested within a cluster/community, and their characteristics may be similar to those living in the same cluster/community compared to the rest of Nigeria. Adjusting for the hierarchical nature of the data provides more realistic standard errors and estimates, as multilevel analysis explicitly models the correlation of responses at the group-level. In addition, it avoids relying on the assumption of independence of observations within the same survey cluster and equal variance across clusters [
31]. The findings from the study will prove important for programs and policies targeted at improving BR and BC in Nigeria.
Statistical analysis
First, univariate analysis using proportions was employed to describe the characteristics of the sample. Second, a bivariate cross-tabulation chi-squared test was used to check differences in the characteristics among the certified and non-certified children. Third, a multilevel logistic regression model (MLRM) which accounts for the binary nature of the outcome variable and the hierarchical structure of the data, was employed to analyse the relationship between the SED factors and birth certification. The MLRM which accomodates both fixed and random effects helps prevent misleading conclusions about the relative importance of the different sampling levels on the child’s BC [
38]. Further, a multilevel poisson model [MPM] was used to test the sensitivity of the MLRM estimates. Studies suggest that the poisson regression can be adapted for dichotomous variable when the prevalence is no longer rare e.g., larger than 10%, as in such cases the differences between the odds ratio and the relative risk may not be negligible [
39].
In the MLRM analysis, the children
i are nested within communities
j.. The two-level MLRM which follows the Bernoulli (
πij) distribution with a logit link function is defined as follows:
$${BC}_{ij}={\beta}_{0,0}^{\ast }+{\sum}_{a=1}^m{\beta}_{a,0}^{\ast }{W}_{a, ij}+{\sum}_{b=1}^p{\beta}_{b,0}^{\ast }{X}_j+{\varepsilon}_{0,j}$$
(1)
where
BCij is the predicted log odds of individual child
i (Level-1) living in community
j, (Level-2) having their birth certified.
\({\beta}_{0,0}^{\ast }\) denotes the overall intercept (the grand mean of Level-2),
\({\beta}_{a,0}^{\ast }\) and
\({\beta}_{b,0}^{\ast }\) are respectively vectors containing the
mth and
nth coefficients associated with W (Level-1) and X (Level-2) predictors of BC. Further,
ε0, j represents the random effect at the community-level - variance between communities - based on an assumption that
ε0, j ~
N(0,
\({\sigma}_{com}^2\)) follows a normal distribution with mean zero and the covariance matrix for a two-level model. The logistic transformation for Eq. (
1) is as follows:
$${BC}_{ij}=\ln \left(\frac{\pi_{ij}}{1-{\pi}_{ij}}\right)$$
(2)
and represents the probability that an
ith child in the
jth community (com) will have their births certified. A positive regression estimate indicates a positive association between the factors and BC, and vice versa. Four separate models were fitted for both the MLRM and the MPM, respectively:
Model 1 (null model) was fitted without any explanatory variables to measure the variation across communities and estimate the intraclass correlation coefficient (ICC) at the community level. The ICC measures the correlation between two observations within the same community i.e., the proportion of total observed individual variation in BC attributed to variations between communities. The larger the ICC (i.e., the higher the correlation
within the clusters), the lower the variability is
within the clusters, and consequently, the higher the variability is
between the clusters [
40]. The second model was adjusted with the child-, parent- and household-level characteristics; the third model was adjusted for the community-level characteristics only, while the fourth model adjusted for all the explanatory variables. Year-and-month dummies for the interview date and the child’s birth were also included in all models, except the null model, to account for the differences in lag lengths, and potential observed and unobserved heterogeneity in trends across space and time [
41].
The beta (β) estimates are the FEs of the MLRM, and capture the overall average relationship (
measures of association) between BC and the predictor variables. They are expressed as adjusted odds ratios (AORs) with their corresponding 95% confidence intervals, and a
P-value of <0.05 was used to declare a statistically significant association between the variable and BC. Further, the intercepts and residual errors represent the RE and imply how the relationships between the communities differ from the overall average relationship (
measures of variation). They are also assumed to be independent of the individual- and community-level characteristics, as they pertain to the random part of the model. The RE is interpreted using the ICC, the proportional change in variance (PCV) and the median odds ratio (MOR). In an MLRM, the individual level (Level-1) has a standard logistic distribution with variance
\(\frac{\pi^2}{3}\) ≈ 3.29.(
12 π denotes the mathematical constant 3.1416 and not the probability.) The ICC is the correlation between two children within the same community (explains cluster variation), and is calculated as
\({ICC}_{com}=\frac{\tau^2}{\tau^2+\frac{\pi^2}{3}}\);
τ2 is the estimated variance of communities/clusters and
\(\frac{\pi^2}{3}\) is the children/individuals’ variance.(
13 The ICC is calculated based on the widely adopted latent response formulation that assumes a latent continuous response underlies the observed binary response.)
ICCcom is used to evaluate whether the community factors are more relevant in explaining the variation in the outcome variable and establish the need for the multilevel analysis [
38,
40]. The higher the ICC, the larger the contribution of the community factors in explaining the variation, and an ICC of less than 5% at the null model suggests that hierarchical modelling may not be necessary [
42]. The PCV, which indicates the additional effects of the included variables (individual- and community-level factors), is calculated as [PCV =
\(\frac{V_e-{V}_m}{V_e}\)], where
Ve is the variance in the null model and
Vm is the variance in the successive models. Finally, the MOR is defined as the median value of the odds ratio between the highest risk area and the lowest risk area when two areas are randomly sampled and is calculated as [MOR = exp.(
\(\sqrt{2\ast {\sigma}_{\mu}^2\ast {\phi}^{-1}(0.75)}\)], where
\({\sigma}_{\mu}^2\) is the cluster variance and
ϕ−1 is the inverse cumulative standard normal distribution function which is approximately 0.6745. In simple terms, the MOR is a measure of unexplained cluster heterogeneity and can be interpreted as the median increased odds of being certified when a child moves from an area with a low to an area with a higher probability of certification. The higher the MOR is, the larger the general contextual effect. If the MOR is equal to 1 (i.e.,
\({\sigma}_{\mu}^2\) = 0), it implies the absence of a neighbourhood variation.
It is important to note at this point that the non-numerical-based categorical variables - gender, size-at-birth, religion, ethnicity, and place of residence - are
effect-coded for the regressions, which is coded to yield a sum to zero constraint [
43]: in the case of two distinct values (e.g. gender) this will have values of 1 and − 1. This is especially important for the MLRM as effect-coded variables have greater speed of convergence than dummy-coded variables [
43]. All regressions were conducted in Stata 16 MP [
44]. To further accommodate the complex nature of the NDHS, the data was adjusted for under-reporting using the “
svyset” command prior to statistical analysis and the
svy command was used to run the commands. The data were weighted using sampling weights adjusted for the unequal probability of selection between the geographically defined strata as well as the non-responses to ensure that the findings can be generalized [
45]. Given that the final sample is based on a pooling three rounds of the NDHS, the sampling weight provided in the data have been de-normalized and scaled. A detailed explaination of the procedure for de-normalization and re-scaling can be found in the DHS methodology reports [
45].
Discussion
This study aimed to assess the pattern of and factors associated with birth certification in Nigeria using pooled data from the 2008, 2013 and 2018 rounds of the Nigerian Demographic and Health Survey, and multiple statistical approaches. The data of 66,630 children under-five years were included in the study, and 17.1% of them had their births certified. This suggests that the nation is miles away from ensuring that children’s births are registered and certified, and this poses an obstacle to achieving SDG target 16.9 “to provide legal identity for all, including birth registration, by 2030”. The setting of Nigeria is particularly important for a study of this nature, as the country ranks high in child population, coupled with the high prevalence of poverty and low parental education (key measures of socioeconomic status). In addition, the inequality in land size across the nation also has interesting effects on access to public facilities, which in turn has significant implications for the utilization of public services. Further, the results suggest that the prevalence of birth certification was above the national average in 19 out the 37 administrative states of Nigeria, and low birth registration coverage is clustered in the northern part of the country. The finding of north-south gaps are consistent with the findings of other studies in Nigeria [
46‐
48], and the underprovision of public facilities in northern Nigeria may have historical roots [
46].
The results of the multilevel analysis suggest that individual-level characteristics may matter more than the community-level factors in explaining the patterns of BC, and several individual-level (level one) variables have strong and significant associations with BC. At level-one, access to healthcare systems (measured by the having an SBA and being vaccinated (child-level) and prenatal visits (mother-level)), maternal age at birth, ownership of bank account, and socioeconomic status (measured using parental education and wealth) are significant protective factors of BC. The positive effects of health access on BC confirm the findings of [
20,
21] as well as those reported in Ghana [
27] and selected Latin American countries [
28], respectively. This finding reflects positive spillovers of healthcare utilization, as being closer to health services and skilled personnel increases the chances of receiving information about the need and the process of BC. The information that may be provided by healthcare officials (e.g., midwives, nurses, doctors) could help parents/caregivers decide to start the birth registration process, and complete it by collecting the child’s birth certificate from the local authorities. Maternal age at birth also increases the likelihood of BC which confirms the findings of [
22‐
24]. However, the positive effects of maternal age at birth on BC peaks for children born to mothers aged 30–34 years and declines at advanced gestational age i.e., 35 years and above. This non-linear effect has not been noted elsewhere for BC. Other studies, however, confirm that maternal age can have a non-linear effect on child development [
51,
52]. One potential explanation is that mothers at that age may have greater levels of responsibilities and time demands in-and-out of the home [
53], which can lead them to postpone the certification of their child’s birth.
Evidence of a strong statistical association also exists between parental education and childhood BC: children whose parents had more years of education were more likely to be certified. This finding is in line with previous studies conducted on the determinants of birth registration in Nigeria and elsewhere [
19,
28]. The arguments rest on the premise that the knowledge parents acquire from the formal education system could help them better process the information regarding step-by-step procedure for BC: receive a birth notification slip from the health centre at the time of birth and proceed to the nearest accredited centre for registration and certification of the birth. In addition, human capital theorists hypothesize that more educated parents are more likely to choose better options for their children to enhance future economic and social mobility [
54]. Also, educated parents may engaged in other behaviour such as better health-seeking behaviours for their children as compared to uneducated parents which can help improve the likelihood of BC. The positive relationship between parental education and child health-seeking behaviour has been noted in Nigeria and elsewhere [
55]. In addition, being from a middle-income or rich HH increases the likelihood of BC for the children, which is in line with the findings of [
6,
28,
53]. Moreover, it confirms the argument that richer parents may be more aware of the importance of BC on the child’s future mobility (for example, university education, participation in the formal job market and legal migration abroad) [
9]. Higher education and wealth have also been linked to better decision-making as the more educated and wealthier individuals/families are in a better position to earn money and afford the cost of registration services [
56,
57]. Another interesting finding of this paper is the significant influence of bank account ownership on BC, which has not been noted anywhere else. One potential explanation is that parents/caregivers who have made contact and interact with the formal financial system may be knowledgeable about the importance of BC. For instance, one of the required documentation for opening ain bank account in Nigeria is a valid proof of age, and a birth certificate can provide such information.
Mixed results were also noted for the effect of ethnicity on the likelihood of BC: Children of Yoruba ethnic group had higher chances of being certified compared to Hausa/Fulani children. This finding is in line with the study on BC by [
58]. This could be linked to the differences in cultural practices which shape the reproductive health decision-making around pregnancy, child birth and the postnatal period [
59]. Notable deterrents to BC include child’s age (48–59 months), 3rd or 4th or higher-order birth (child), having two or more dead children (mother), working in a low-skilled job (father), being of minority religion (HH-level), living far away from the registration center (>5+ kms), living in a poor community and being from the SE (community-level). This finding of the negative effect of distance to registration centers on BC has not been recorded for Nigeria; however, [
29] reports similar findings for children living in selected countries in Latin America and the Caribbean. The potential explanation for the negative association is that the greater distance to the registration center increases the financial and opportunity costs for the family (especially for the poor) and thus lowers the likelihood of birth registration and certification. The findings of higher birth order and father’s work status as significant obstacles to BC are also in line with other studies of birth registration [
58,
60,
61]. The effect of birth order remains mixed in the literature on child development; however, a typical suggestion for the negative effect on BC is related to the delay in the benefits of birth registration. After the first child is born and is successfully registered and certified, without immediate returns to the certification, it becomes less likely that later-born children will have their births certified. Another explanation for the negative effect of birth order lies in the resource dilution hypothesis. Older children are proxies for larger family size, and in larger families it is hypothesized that the resources spent on caring for children are diluted and it becomes more costly for parents to make the commitment towards registration and certification [
62]. Despite this, to draw any conclusions on the effects of birth order on birth certification, more empirical work needs to be done to account for unobserved heterogeneities that may occur within-families to draw causal inference. The findings from the random parts of the MLRM and the two-level poisson showed significant variance between communities even after adjusting for the characteristics at the child, parent, HH and community levels. This confirms the need for multilevel modelling techniques. This finding was consistent with the study conducted by [
58] and justified by the existence of differences in coverage, social norms, cultural beliefs, geography, quality of health services and distribution of registration centers. One can thus conclude that about 31.6% of the variation in BC is attributed to differences across children nested within communities.
Strengths and limitations
To the best of my knowledge, this paper counts among the first studies highlighting the determinants of birth certification for Nigerian children using a robust range of socioeconomic and demographic factors potentially associated with the registration and certification process. The study has numerous strengths. First, the study uses the NDHS data, the largest nationally representative and mutually comparable repeated cross-sectional data sample available for 66,630 children born to 46,672 mothers in 3127 communities in Nigeria for the period 2008–2018. The pooled NDHS sample provides substantial heterogeneity within and between communities to analyse the association between SED factors and BC. The NDHS surveys are comparable across settings due to the standardized nature of the variables within the dataset. Hence, these findings could be tested in and generalized to other developing countries for whom data are available. Second, the multilevel regression modelling corrects for bias on the parameter estimates as it explicitly models and uncovers heterogeneity in covariate effects [
41]. The wide range of variables employed in the analyses allows a more realistic depiction of the socioeconomic and demographic determinants of BC for children under five years, and go beyond the small-scale studies of [
19,
25,
63] and capture more variables with the potential to influence BC than [
58]. This study may thus provide a guide for future empirical studies investigating the predictors of BC in sub-Saharan African context.
Despite the strengths mentioned above, it is important to note the limitations of this study. First, the cross-sectional nature of NDHS data makes it difficult to track the children over time to confirm their certification status or draw causal conclusions on the derived estimates. To verify the validity of the observed associations and make causal claims, the individual- and community-level effects need to be unpacked by using longitudinal data or instrumental variable techniques. Second, the data do not precisely state whether the children had their births certified in their current location which may affect the patterns reported at the state level. Further studies can incorporate the effect of internal migration on BC outcomes. Third, the data does not provide information on the exact timing of the certification. Hence, discussion on whether the child was registered and certified within the recommended time (i.e., within the first 60 days of birth) is outside the scope of this study. Furthermore, the spatial data used to construct the accessibility to registration centers are de-identified to ensure confidentiality of the respondents; however, this may introduce some measurement error in calculating the supply side variables. Future studies should employ longitudinal or experimental analysis to allow for a stronger generalization of the findings.
Conclusion
Overall, this paper investigated the determinants of BC in Nigeria and contributes to the sub-Saharan and low-and-middle income context. This study has provided significant insights into the role of individual and community-level factors on the birth certification status of children under five years of age in a large sub-Saharan and middle-income country, Nigeria. The results highlight the crucial roles played by health service utilization, socioeconomic status and accessibility to registration services. The knowledge of these factors as key influencers for birth certification can help drive well-targeted policies by the government (e.g., the NPC) and local and multinational organizations interested in improving birth registration and certification rates in Nigeria. For instance, the government through the NPC can work towards improving birth certification rates by addressing geographical accessibility to registration centres. One way can be to increase the number of registration centers in the country and reduce the distance parents must travel to register and certify their children’s births (e.g., within a 5 km radius). This would need strong political will and concerted efforts from the Nigerian government and other stakeholders in the form of administrative and financial support for CRVS systems. Also, the stakeholders can design social protection programs in the spirit of the child development grant program (CDGP), with literacy components and cash transfers which are conditioned on parents registering and certifying the births of their children. Furthermore, free registration can be enforced to ensure that parents begin and complete the birth registration for the children on time. This can help improve birth certification among children born to uneducated parents and living in the poor households. The findings that children near-school going age (48–59 months) are less likely to have their births certified suggest significant delay in birth certification. This goes against the mandate of the 2003 Child Rights Act that parents should act in the best interest of the child. However, this can be addressed by large-scale campaigns on awareness on civil registration (especially child births) and its intended benefits. These can be addressed at the community level by engaging private sector and non-formal institutions (religious and community leaders, and civil society organisations) in the birth certification agenda. I suggest the design and implementation of well-targeted birth registration and health programs to ensure children are registered and certified on time. Ensuring this will be valuable to achieving the target of universal birth registration by 2030. Another suggestion is to ensure a holistic national child policy, which comprises all child development factors - health, education and protection that are necessary to ensure that children survive and thrive in adulthood. Finally, for a comprehensive policy to be enacted, it is important that future studies conduct a causal investigation for each of the significant factors found in this study.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.