3.1 Household level

The birth registration decision is made at the household level, generally by the parents or caregivers of the child involved [14, 19]. Their decision might be influenced by socio-economic and demographic characteristics of the household. One of the most important determinants at this level is household wealth [13, 19]. In most SSA countries, a fee is involved in the birth registration procedure that may prevent poor families from registering their children [3, 13, 14, 19]. Even if parents do not have to pay for registration initially, there might be a fee for late registration [20]. Not only direct costs, but also indirect costs, like travelling and time that cannot be spend on working, may prevent births from being registered [3, 13, 14, 19]. These indirect costs are often higher for poorer families as they tend to live in more disadvantaged neighbourhoods that often are located further away from registration offices and with worse infrastructures [3, 7]. Besides wealth, access to a mobile phone might be beneficial for registration rates as it may supplement traditional media, like radio and TV, as a source of information. Mobile phones may grant access to the internet, which can improve the access to information about birth registration [28]. Moreover, technological advancement may allow for the development of systems in which a birth can be registered by using a mobile phone [22, 29]. In West Africa, these systems are already starting to work and reduce the costs of birth registration [29].

Besides the proper resources, knowledge about the importance and procedures of birth registration is needed [7, 14, 19]. An important factor influencing knowledge is education [19, 30, 31]. The higher the level of education, the higher the chance that parents recognize the value of having a birth registration and know how to register a birth [14, 32]. Indirect effects can also be seen as educated parents often have better educated social networks, which is one of the most common ways to acquire knowledge and advice about birth registration [20, 33].

Demographic factors like the age of the child or family structures may be important too. Identity documents become more important if children grow older, as they may be needed for entering or graduating from school or to access health services [10, 14, 18, 31, 34]. Regarding family structures, missing one or both parents can make birth registration problematic, for example because in some countries legislation requires the father to register a child [13, 17, 19, 35]. A missing mother may lead to disregarding the birth registration, as mothers are primary caregivers regularly [17]. If the mother is present, also her autonomy and bargaining power within the household are important. According to Mohanty & Gebremedhin [20], bargaining power is important, since women with more bargaining power are more likely to spend resources on their children. These authors also consider the mother’s ability to move around as crucial for activities that enhance the welfare of children, such as immunization, health check-ups, and birth registration. Bloom, Wypij, & Das Gupta [36] provide evidence that women with more autonomy seek more antenatal or prenatal care, which leads to better child health outcomes.

The ethnic or religious group to which a household belongs can also influence the birth registration decision [13, 17, 19, 37]. Such groups can have multiple reasons for not registering births, including name giving traditions that may make early birth registration problematic [14, 17, 38] and misgivings about the way birth registrations are handled [20, 26], like with the Rwandan genocide in 1994 or the apartheid regime in South Africa [7, 13, 17, 24]. Language issues may play a role as well, as minorities in the SSA context often have their own language, which can cause registration problems due to language barriers and illiteracy [12, 13, 37].

Besides socio-economic factors, care-related factors are important. According to Fagernäs & Odame [3], “registration offices are often located within health facilities or close to them, which creates a direct connection between health care and registration.” Health care related to pregnancy and early life can increase awareness of the importance of birth registration. Skilled health personnel, seen as a credible source of information, can provide women with information about birth registration, recommend to register the child, help with the paperwork needed and propose a registration office [20, 31]. This is especially important as traditional birth attendants were found to motivate parents less to register births than skilled birth attendants [14, 17, 28, 37]. Moreover, giving birth at home poses more restrictions on birth registration, for example due to travelling costs and not having help with filling in forms. Receiving care or giving birth in an institutional facility is therefore of great importance.

Availability of primary care, provided by health institutions, in the first years of a child’s life is important as well [4, 28]. According to Pelowski et al. [23], “using vaccine delivery (particularly Diphtheria-tetanus-pertussis, DTP) as an occasion to register births may also provide a means of reaching children born outside health facilities.” The same reasoning holds for receiving vitamin A [3, 19]. Health care personnel can discuss the missing birth registration of a child during the administration of vaccination or vitamin supplements [3, 19].

3.2 Context factors

Although the birth registration decision is made at the household level, the situation in the regional and national context can influence the outcome of this decision, as the availability and efficacy of services, policies, and infrastructures vary considerably within and across countries [9, 20, 34]. An important characteristic of a region is the degree of urbanization, reflecting the travelling distance to the nearest registry office–which often do not stretch out to underdeveloped and remote areas–and infrastructure within the region, both affecting the costs and the information flow of birth registration [9, 19, 20, 25, 37].

The availability of services, such as health care and education, is also important [20, 39]. Following Corbacho & Osorio Rivas [25], the further the travel distance, the higher the probability of not making use of available services. Related to this, the mother’s inability to move around within the region can also decrease the use of services [20, 36]. This means that less information on and help with birth registration will be obtained and the chance to receive and spread information about registration (spill over effects) is lower too [20, 39, 40]. A better information infrastructure counters these effects.

National context factors can be expected to be more important than local factors, as birth registration legislation is made at the national level. In low income countries, civil registration systems are often underdeveloped due to the lack of economic resources while these resources are needed for the creation and maintenance of good systems [12–14, 17]. This may lead to legislative barriers too: no legislation at all, outdated legislation, or weak enforcement of existing laws as laws are not harmonized [7, 11, 13, 41]. Besides that, the content of the law might be harmful for registration [12, 31, 42]. A registration fee can be detrimental if poverty is a major issue and specified legal time periods can lead to procrastination–when given a lot of time–or can make registration unfeasible for people living under difficult circumstances–when given little time [11, 14, 18, 19, 23].

Another consequence of lack of resources is that the formation of appropriate institutions for birth registration is problematic [7]. Due to the complex and expensive nature of decentralized systems–as different institutions and actors must interact–centralized systems often are preferred, even though these systems are less flexible and less accessible for people living in rural areas [12, 31, 38]. Countries like Bangladesh, Kenya and Tanzania have seen their birth registration rates go up after moving from a centralized to a decentralized system, which has more local registration points and in which less steps are needed to complete the birth registration process [8, 23, 43].

With regard to demographic factors at the national level, both fertility and child mortality may be important. If fertility rates are high, registering every child may be problematic because of the substantial direct and indirect costs of registration [3, 14, 19, 29]. High child mortality rates, on the other hand, may reduce the motivation of parents to bear the registration costs [29, 41].

A final relevant factor might be the history of the country. The colonial era and periods of war and conflict may have long-lasting consequences for institutional arrangements [13, 17, 44]. With regards to colonialism, path dependency may hamper the development of birth registration processes, for example as colonizers introduced birth registration only in specific regions or only for non-Africans [7, 17, 44]. War and conflicts may have devastating effects on existing registration systems as well [13, 29, 41]. Misuse of these systems during (civil) war, may lead to mistrust and breaking down of them [8, 9, 24, 26]. Episodes of genocide, like in Rwanda and the Demographic Republic of the Congo may have particularly long lasting consequences for registration rates [8, 13, 26].

4.1 Data

For this study, combined datasets from the Demographic and Health Surveys (DHS; www.dhsprogram.com) and Multiple Indicator Cluster Surveys (MICS; https://mics.unicef.org/surveys) have been used, which were derived from the Global data Lab (www.globaldatalab.org) [45]. DHS and MICS are large, nationally representative household surveys. For each survey, non-overlapping areas (often enumeration areas) are randomly selected. These areas (called “clusters” henceforth) are usually communities, villages, or city quarters. In the selected clusters, all households are listed and a random sample of 25–30 households is selected for the interviews. To get a maximum discriminatory power, the data of the most recent standard DHS and MICS surveys for SSA countries, that contain the relevant variables, have been pooled.

Our combined dataset contained information on 442,433 children aged 0 to 4 years old who were living in 809 sub-national regions within 40 sub-Saharan African countries. For an overview of the countries and the numbers of subnational regions, see S1 Table. As a consequence of missing cases on the dependent variable, 81,951 observations could not be included in the analysis. Missing cases on independent variables with less than 500 missing were handled by listwise deletion. For this reason 1640 observations were removed. Accordingly, 358,842 observations remained for the analysis. For missing characteristics of parents, birth registration legislation, or conflicts, dummy variable adjustment was used [40, 46]. In accordance with ICF International [47], the standard weights included in the DHS surveys were de-normalized and subsequently normalized according to the national population sizes. In this way a representative sample of the population in the 40 countries was created.

Context factors at the sub-national regional level were constructed by aggregating data from the household surveys to the local area level, using the regional codes present in the datasets [48, 49]. For urban and rural areas of sub-national regions, separate indicators are used. Data on birth registration systems and legislation were derived from UNICEF [50]. Data for national income, rule of law, government effectiveness, the fertility rate, and under-five mortality were derived from the World Bank [51–53]. For South Sudan, no data for 2010 was available for the variable rule of law. Therefore, the year of 2011 was used instead. Information about the history of the country was retrieved from the Uppsala Conflict Data Program [54] Version 19.1 (UCDP) [55, 56]. In the few cases that no information on the colonial period was available at the UCDP, the Encyclopaedia Britannica was used instead [57].

4.2 Methods

A three-level multilevel logistic analysis is used to address the clustering of households within sub-national regions and countries [20, 27]. Logistic regression is used as the dependent variable is a binary outcome with value 1 if the child was reported to have its birth registered and value 0 if it was reported to have no registration. The variable is based on the question: “Does (NAME) have a birth certificate? (If no, probe): Has (NAME)’s birth ever been registered with the civil authority?” Children that were reported to have a birth certificate or to be registered obtained a value of 1 and children that were reported to have neither a birth certificate nor a registration obtained a value of 0. The categories ‘don’t know’ and ‘missing’ were marked as missing values.

Following Mohanty & Gebremedhin [20], first two empty models with random effects at the national and local level are estimated. For these models the intraclass correlation will be estimated in order to determine the variation within and across different levels [20, 58]. Following these models, household and context variables are added [27, 40]. For studying the role of the context factors, given the restricted number of regions and countries, an explorative approach is used in which only significant variables are included in the models. At the household level more observations are available, so that statistical significance is less an issue there and effect sizes are more important. We checked for multicollinearity of the interval variables and found all of them to have a VIF value below the critical value of 5. All models were estimated with MLWIN version 3.04.

4.3 Independent variables

Household wealth was measured by the International Wealth Index (IWI), which indicates the standard of living of households based on their possession of durable goods, the quality of their housing and access to basic services [59]. Mobile phone ownership is measured by a dummy variable with value 1 if the household owns a mobile phone and value 0 if not. Education of the parents was measured by the years of education they completed.

Regarding demographics, age was measured in years. Missing of one or both parent(s) was measured by two dummy variables, indicating whether (1) or not (0) the father or mother was missing from the household. The position of women in the households was indicated by two variables, a dummy indicating whether (1) or not (0) the mother has given birth before the age of 18 [40] and whether the mother can decide on contraception [20] with three categories indicating whether the decision on contraception is taken by the mother herself (1), the partner (2), or whether it is a joint decision (3).

The variables ethnicity and religion were based on pre-coded questions with the option to add additional categories. Religion consists of 7 categories, namely: (1) Catholic, (2) Protestant, (3) Christian, not specified, (4) Muslim, (5) no religion, (6) other, and (7) Traditional. For ethnicity a three-category variable was constructed indicating whether the ethnicity of the household is a (1) majority group, a (2) normal sized group, or a (3) minority group. These values were given based on the percentage distribution of the existing groups within the country. If a group concerned 0–10% it was seen as a minority, if it concerned 10–30% it was seen as a regular group, and if it concerned 30+% it was seen as a majority.

Prenatal care personnel indicates who has performed the prenatal check: (1) no prenatal check was performed, (2) a traditional health care worker, (3) another person, (4) or skilled health personnel. The place of delivery is indicated by a dummy variable with value 1 when the birth took place at (someone’s) home while value 0 when at an institution. Assistance during delivery was computed to show whether the birth was assisted by: (1) no one, (2) a traditional birth attendant, (3) another person, (4) or skilled health personnel. The variable postnatal check reflects whether (0) or not (1) a postnatal check has been performed within 2 months after the birth. The variable vaccination shows whether children have ever received a vaccination (0) or not (1) while the variable vitamin A reflects whether a child received vitamin A in the last 6 months (0) or not (1).

Context factors at the regional level were aggregated from the household data, following Smits & Huisman [27]. Availability of health facilities was measured by the mean number of prenatal check-up visits of women in the region. The information infrastructure was indicated by the average number of years of education of adult males and the percentage of households with a phone in the region. For the position of women, the mean age at first birth in the region was used. The variable urbanization reflects whether (1) or not (0) the household lives in an urban area according to the definition used in the surveys.

The economic situation of the country was measured by the Gross Domestic Product per capita (in current US dollars). The variable birth registration legislation indicates whether a country has (1) a legislation for birth registration or not (0). In turn, the variable no update in legislation represents whether there has been an update in birth registration legislation over the years (0) or not (1). Time allowed for registration is a dummy variable that indicates that a birth must be registered within a month (1) or not (0). Registration fee is a dummy variable indicating whether (1) or not (0) a fee was involved in birth registration. The organizational structure indicates whether the procedure is decentralized (0) or centralized (1). The level of governance of the country is proxied by the variable rule of law, in which -2.5 is the lowest score and 2.5 the highest [53]. The demographic situation of the country was indicated by the fertility rate and the mortality rate of children under 5 years old. Finally, the history of the country is represented by the number of conflicts between 1990 and the year of the survey, computed by adding non-state conflicts and state conflicts, and whether a country has been colonized (1) or not (0) [55, 60–62].

4.4 Descriptive statistics

S2 and S3 Tables show descriptive statistics for the variables used in our analyses. We observe that 52.1% of the children aged 0–4 in the sample did not have their birth registered. The average IWI of the households in this study is 31, showing that asset ownership and quality of housing and services is low in these countries. However, in 67% of the households a phone is present. Of the children, 28% lived in a household where the father is not present and 5% in a household where the mothers is not present. The average years of education is with 4.9 about one year higher for men than for women. Also, 36% of the women, of whom age at first birth is known, have given birth before the age of 18. Of the children with valid information on the care variables, 60% is born in a health institution and 58% is born with help of skilled health personnel while 40% is still born at home and 18% with help of a traditional birth attendant. In other cases, no care is received at all. For 12% of children, the mother received no prenatal care. Regarding the children, 54% has not received a postnatal check-up within 2 months, 30% has not received a vaccination, and 42% has not received a vitamin A supplement in the last 6 months. The households lived predominantly in rural areas (72%). As most countries have birth registration legislation (96%), the content of the legislation is expected to explain most of the variation in registration at the national level.

5.1 Regression analysis

Table 1 presents the results of the multilevel logistic regression analysis. The table shows that most of the household level factors have the expected effect: having more wealth, a higher age, and more years of education as a parent all significantly increase the odds of having a birth registration. The coefficients of the variables phone, father missing, and mother missing show no significant effects. Also the variables regarding decision making about contraceptive use are not significant. Nevertheless, the position of women in the household seems important for birth registration, as children of a mother who has had her first birth under the age of 18 have significantly lower odds of being registered. Whether you belong to a major, regular, or minor ethnic group within the country is not important. However, belonging to a religious group seems to play a role. Children that belong to Catholic, Protestant, Christian, or Islamic families have higher odds of having a birth registration than children from families with a traditional religion. Children with a Catholic background have the highest odds of being registered.

Download:

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

Table 1. Logistic fixed effects models with random intercepts ^a.

https://doi.org/10.1371/journal.pone.0265882.t001

Among the care variables at the household level, we see that children born at a health institution, with help of skilled health personnel, who were checked within 2 months after birth, who have a vaccination, and who have received vitamin A in the last 6 months all have significantly higher odds of being registered. Receiving prenatal care is also important, but whether it is given by a professional or a traditional health care worker does not make much difference.

Regarding context factors at the sub-national regional level, both the average number of prenatal care visits in the region and urbanization show significant effects. As expected, registration rates are higher among children living in urban areas and in areas in which the average number of prenatal care visits is higher. If the average number of prenatal care visits increases by 1, or one lives in an urban area, the odds of having a birth registration increase by about 10%. No significant effects are found for the age at first birth of women in the region, the educational level and the percentage of households with a phone in the region.

With respect to the variables at the national level, GDP per capita shows a significantly positive relationship with birth registration. Moreover, the organization of the birth registration system is important. In countries with a centralized system, the odds of having a birth registered are 72% lower, and in countries where a fee has to be paid they are 67% lower. Next to the content of birth registration legislation, the fertility rate in the country and the number of conflicts experienced in recent decades are significantly associated with registration. Birth registration rates are higher in countries with higher fertility rates and in countries that experienced less conflicts.