Part One Background and Conceptual Clarifications for Gender Analysis of Census Data
Source: Elaborated on the basis of data provided by UN Statistics Division (see footnote 1)
8. A census is one of the most important tools for policymakers. It takes stock of the most important asset of a country – its human capital, women and men, girls and boys. Population data gained from censuses, together with vital registration data and various kinds of administrative records are critical for ensuring that appropriate policies and programmes are prioritized at national and local levels.
Text Box 2: Uses of Census Data
Census data can serve
Planning at the national, regional and local levels
Decision-making for strategies or policies
Policy and programme development, (monitoring) and evaluation
9. Therefore, censuses are a rich source of information about the differences between men and women, girls and boys, or about the needs and requirements of population subgroups such as elderly men in rural areas or adolescent girls. Their greatest advantage for the purpose of gender analysis is that censuses allow for disaggregation down to the smallest geographical unit. Regional NGOs and policy-makers in city councils, for instance, will be able to extract data specifically on their region/city of interest and understand the population composition within that restricted area.
10. Censuses can also provide basic national-level development indicators, for instance on fertility and spatial distribution of men and women. For more complex indicators, census data often serve as a denominator. They can for instance help uncover gender disparities in employment, literacy and age of marriage. Where international definitions and classifications are used, indicators derived from censuses are comparable among countries. Such indicators can then be used to benchmark progress in achieving the Millennium Development Goals (MDGs) or the ICPD Programme of Action and to monitor compliance with human rights obligations such as CEDAW.
Methodology Box 1: Combining Data from Different Sources
The main advantage of census data is their universal coverage. The main drawback is the generality of the information provided, which is usually lacking in detail for the purposes of an in-depth gender analysis. However, census data may be combined with other sources to examine many of the topics discussed in Part Two of this manual. Drawing on multiple data sources enables one to carry out analyses that cannot be supported by census data alone. The simplest strategy for doing this is to compute aggregate values for the relevant variables from both sources separately at the level of relevant population groups. For example, one may be interested in fertility preferences and income levels for women of different educational levels. Since education is a variable included in both the census and the DHS, these indicators can be computed separately using either data source, and the results can be compared. The main limitations of this approach are that it only works for groups that can be defined in terms of both data sources and that the number of such groups cannot be too large, as the DHS does not allow much disaggregation.
In order to go beyond such simple comparisons of groups, one needs to integrate the two data bases. To this end, there are two main strategies: construction of proxy variables and statistical matching. The construction of proxy variables consists in developing a regression model or other multivariate model based on the survey data and using explanatory variables that are common to the survey and the census, to predict the value of the variable that one would like to include in the census data base. The census value of the variable is then constructed by using the same equation on the explanatory variables, as found in the census. Typically, this approach has been used for the construction of household income data for censuses that do not have this information, by regressing household characteristics such as ownership of consumer durables or the quality of construction of the home on income data from a Living Standards Measurement Survey or other kind of household survey that provides income data (Elbers, Lanjouw and Lanjouw, 2002). The primary objective, in this case, is to construct poverty estimates for smaller geographic areas than is feasible with the income survey itself. But the approach is not necessarily limited to this application. In the particular case mentioned above, one might predict desired family sizes based on, for example, the age and number of living children, level of education and urban/rural residence of the woman and then apply the same equation in the census, in order to relate the desired number of children to typical census variables.
In the statistical matching or “data borrowing” approach, one uses the variables that are common to the census and the survey to construct a measure of similarity or distance between individual cases of the census and survey files. Each individual case found in the census is then matched to its closest neighbour in the survey file. In some cases one may want to divide the data into different subsets, in order to avoid, for example, the matching of men to women or persons from very different parts of the country. The survey data of the closest neighbour are then simply imputed to the individual census records.
When a survey is done shortly after a census it may be possible to establish a match between census records and survey records on the basis of common geographical identifiers. Since surveys typically use a census-based master sample frame such a match is technically quite feasible, as long as the time interval between census and survey is not too long (say, less than 2-3 years). After appending the two data sets the desired survey variables can be estimated for households or persons that were not covered by the survey on the basis of the relationships found amongst those records where both census and survey data is available.
Both methods are not without their pitfalls and complications. Both the construction of proxies and the statistical matching approach assume that once the common variables have been controlled, the remaining variables from the survey are statistically independent from those in the census. The fact that this is often not the case may introduce systematic biases. A number of procedures have been proposed in the literature to deal with this problem (e.g. Rubin, 1986; Moriarty and Scheuren, 2001). Because of these complications, either of the two main strategies should not be applied without calling in appropriate technical support.
B. Gender Analysis of Census Data
11. Gender analysis as a way of interpreting census data has emerged in response to growing need of gender information of countries, e.g. to report on progress made in terms of gender equality and the empowerment of women in line with international obligations. As a way of working with data, gender analysis is more than simply analysing quantitative data by ‘sex’ using standard descriptive statistical techniques. Gender analysis includes a gender-responsive selection of questions to be posed to the data and in the interpretation of sex-disaggregated data in the context of power relations between the sexes, i.e. in a way that includes other sources of knowledge such as qualitative data, knowledge of cultural factors, or further multivariate analyses shedding light on socio-economic realities.
Text Box 3: What is Gender Analysis?
Gender analysis is a critical examination of how differences in gender roles, activities, needs, opportunities and rights/entitlements affect men, women, girls and boys in certain situation or contexts. Gender analysis examines the relationships between females and males and their access to and control of resources and the constraints they face relative to each other.
Source: “Gender Equality, UN Coherence & You – Glossary: Definitions A-Z)
12. Gender analysis does the following:
Critically examines the differences in women's and men's lives;
Searches for the underlying causes of inequality between women and men and boys and girls;
Highlights gender-specific variables and is generally (though not exclusively) used to achieve positive change for women and girls.
13. Gender analysis illuminates the extent to which:
Women's and men's lives and therefore experiences, needs, interests, priorities, and capacities are different.
Women's lives are not all the same – each woman’s life is also shaped by a host of other social characteristics such as ethnicity, religion, income level, immigration status, sexual orientation, age, etc. The same holds for men.
Women's life experiences, needs, issues and priorities are different for different groups of women.
Men and women have triple roles with regards to work:
Reproductive work: including household maintenance and childrearing;
Productive work: generating income or goods;
Community work: activities in the public sphere undertaken for the good of the community.
To a much larger extent than is the case for men, the work of women is often unpaid.
14. Gender analysis goes beyond interpreting data. As part of gender-mainstreaming (see Par. 30), it is also a practical, programmatic tool that seeks to be participatory and holistic. Gender analysis should place great importance on empowerment, consultation and participation of those concerned. In addition, a comprehensive gender analysis of census data may require multivariate techniques that go beyond the usual practices of NSOs and that require the involvement of academic or research institutions. Therefore, gender analysis should not be carried out by National Statistics Offices (NSOs) in isolation. NSOs can identify, plan, implement, monitor and evaluate gender equality with data analysis projects – for example a publication on the status of Women and Men in their countries – with a) representatives of women’s machineries (Ministry for Family/Women; national CEDAW committee ...) and b) with representatives of Civil Society (women’s movement, NGOs), ideally including community members themselves who can testify as to the lived experiences of women and men in the county. This type of publication can then inform national planning and policy development initiatives.
15. The following sums up some of the strengths of census data with regards to gender analysis (Meena and Chaudhury, 2010; Schkolnik, 2011):
a) Censuses provide a basic set of sex-disaggregated data at the smallest geographical level. Thanks to their universal geographical coverage and routine inclusion of demographic categories such as age, sex and marital status, censuses bear great potential for disaggregation. Censuses provide data on the entire resident population in a country at a given reference period by administrative area. They can thus inform local planners about population composition and characteristics for instance, how many widows, adolescent girls or polygamous families are living in a given administrative area? Thus, census data are key to identifying “vulnerable groups” for targeted interventions (e.g. rural vs. urban adolescent girls). Moreover, the fact that census mapping culminates in the delimitation of the entire national territory into small enumeration areas also means that censuses are used to demarcate constituencies – a basic requirement for election processes, ensuring representation based on accurate numbers and social participation. In the event of natural disaster or human-made crisis, more realistic estimates of the women and men, girls and boys affected can be reconstituted, using information on the enumeration areas affected.
b) Censuses provide insights into the private and community spheres and (indirectly) into time-use of women and girls, men and boys. Feminists have long criticized that the public (=male) vs. private (=female) dichotomy allows government to clean its hands of responsibility for the state of the ‘private’ world. The so-called private sphere (sexuality, reproduction, gender relations including gender-based violence, women’s unpaid care- and housework, etc.) is a notoriously sensitive issue and often under-studied. By entering into households and providing details on household and housing characteristics as well as on social infrastructure, the real-life living conditions of girls and boys, men and women including elements of vulnerability are exposed. What does it mean for a widow to head a household composed of herself and her orphaned grandchildren? What impact does the lack of a water source and access to telecommunications have on girls’ education? What does it mean for women of reproductive age to live in a locality with limited vital social infrastructure, such as health facilities, schools, churches, community halls, markets and roads? If analysed with a gender lens, one can learn a lot from censuses about gendered differentials in access to resources and services.
c) Census data for advocacy: A local-level “early warning system” on gender inequalities? Analyses of census data may uncover sex ratio imbalances or unconventional population structures that are symptomatic of growing inequalities in a country, region or municipality. In Viet Nam, for instance, gender advocates are alarmed about the decreasing number of girls in some provinces. As census results are usually published and widely disseminated for development planning, they can – if properly analysed – serve media, civil society actors, NGOs, researchers and individuals in their advocacy efforts, as a recognized and official source of information. Note however, that although the published reports from censuses may be easily available, they often provide only aggregate information. The raw data files, which should enable breakdown of data to smaller units, are often not easily accessible, especially to non-State actors (more on this issue below). At worst, they may be so poorly conserved that they are no longer retrievable or they may be rendered inaccessible for political reasons, for instance in multi-ethnic countries.
d) Censuses provide essential background information allowing for further research on women and men, girls and boys. Most importantly, the census lays the groundwork for population projections (e.g. how many boys and girls will need schooling in 2025?) by providing details on the key elements of population dynamics, i.e. fertility, mortality and migration. Second, most surveys (e.g. labour force surveys or studies on maternal mortality) draw their samples from master sampling frames provided by the most recent census.
16. Among the weaknesses of census data for the purpose of gender analysis, the following can be highlighted (Meena and Chaudhury, 2010; Schkolnik, 2011):
a) Census data may not have been produced in a gender-responsive way: In most countries, statisticians without specific training in gender are responsible for producing census questionnaires, defining concepts, variables and classifications and for managing the field operations including enumerator training. As a consequence, the data collected in censuses may not lend themselves easily to gender analysis but may in fact already be gender-biased (ECLAC, 2006 a). For instance, the concept of “head of household” is problematic in several ways, but in particular where question wording (or indeed, an enumerator) refers to the head of household as “he”, respondents are likely to underreport on female-headed households.
b) Census data are of very limited scope and depth: Census data do not provide all the information needed for gender analysis. For instance, census questionnaires do not generally ask (and indeed, given their objectives and constraints, cannot be expected to ask) about issues such as women’s unpaid domestic work or gender-based discrimination in public decision-making. Nor do they ask questions about fertility preferences, time-use, sexual behaviour and many other gender-relevant issues. Questions on gender-based violence (GBV) require specific ethnical and safety standards in data collection to protect the victims, making their inclusion in censuses unrealistic. The mode of census enumeration, calling upon a very high number of interviewers, allows neither a rigorous selection of high-level personnel nor a thorough training that would adequately protect the respondents. In addition, the population may be reluctant to answer a long questionnaire and might feel that asking detailed questions on sensitive issues (income, ethnicity, etc.) is an intrusion into privacy. Census offices are generally reluctant to increase the number of questions on census forms, especially if the issues may elicit controversy. Not only does each additional question imply a substantial cost increase, but there is also a risk that it will deteriorate the quality of the core information. Where information is collected on maternal mortality, time-use or violence, it is therefore often neither sufficiently detailed nor accurate.
c) Gender-related discrimination is not explicitly measured by censuses: Census data do not reveal information related to, for example, behavioural consequences of laws or policy. Thus, one of the key tenets of gender analysis – that women, men, girls and boys have different needs and are affected differently by policies and programmes – cannot be examined on the basis of census data. Linked to this, it must be acknowledged that census data can also be misleading about gender relations depending on how questionnaires are designed and administered because respondents may be influenced by gender-related power dynamics. There are a number of good practices regarding gender responsive questionnaires and training of enumerators.
d) The level of analysis for census data is sex, not gender while policy interest tends to be on the gender differentiated needs of men and women, the relational socio-cultural construct, not sex, the biological concept (see Chapter 2.A for extended definitions). As sex-disagreggation is merely a first step to making gender-based analyses, additional effort is needed to unearth women’s and men’s different needs and aspirations as well as the power differentials and relational factors that explain women’s and men’s access to resources and services. For instance, the gender pay gap is a measure of earnings differentials between women and men. Even if censuses ask about individual income (many don't), this only tells us how much, in monetary terms, women take out of employment compared with men, as a male/female difference expressed as a percentage of male earnings. To make more specific statements about gender and inform policy-making, this indicator not only has to be further disaggregated – by age groups, by occupation, by part-time, full-time, etc. – but other factors have to be considered such as the availability of child care, social norms about child-rearing and female employment, and the gendered division of labour in routine housework. While the census can be helpful in the former, it does facilitate the latter.
e) The census data may be outdated or of low quality (e.g. due to underreporting on women): Many countries, particularly countries prone to humanitarian crises, can not respect the 10-year interval for census-taking. Even in those countries that do, the last census may be several years old and its figures may no longer reflect the lived reality of women and men, girls and boys. Complex projection and estimation techniques are needed to estimate the actual situation on the ground. This is not the case in those countries (especially in the European Union) that rely heavily on continuously updated population registers. In terms of data quality, underreporting on women is a well-documented phenomenon, especially in countries like China where female births are often hidden to get around the official one child policy. In some parts of South Asia, unmarried women are less likely to be counted, whereas the under-counting of young male adults and young babies of either sex is widespread in much of the world. Female household headship, numbers of children under age 5, and numbers of young male migrants are also routinely under-reported, whereas age data by single ages may be inaccurate. Finally, many censuses suffer from incomplete coverage such that the results have to be adjusted before publication. Under such circumstances, the raw data files contain information that is different from the adjusted and published information, rendering reconstruction of information for smaller geographical units problematic. In particular, the imputation of missing data or the correction of inconsistencies in the information may be done according to criteria that are not gender-neutral and that, in some cases, actually introduce serious distortions.
f) Data access and the capacity to analyse census data in the appropriate ways may be problematic: Census data bases are typically much larger than the data sets produced by most surveys, making their analysis more difficult. Moreover, NSOs are generally not at liberty to distribute them to potential users in their raw, unabridged form, due to problems of data confidentiality. This is very different from the situation of, for example, the Demographic Health Surveys, most of which are easily accessible to individual researchers. To deal with this situation, NSOs typically adopt one of three strategies: i) They analyse as much of the data as they can in-house; ii) They prepare user samples of 1%, 5% or 10% that have been processed so as to make the identification of individual households impossible, for use by academic and other research institutions; or iii) They distribute the data to the general public in the form of data bases such as REDATAM which allow users to prepare their own tables without having direct access to the micro-data. Each of these strategies has potential limitations. If the gender analysis of census data is carried out in-house, it will usually be guided by the need to produce certain essential tables, but depending on the analytical capacity of the NSO, it will often not go into in-depth studies of particular relationships, particularly if they involve multivariate analysis. Preparing user samples can be costly2 and may run into limitations on the user end if users need to produce detailed analyses of very specific population groups. Information management systems such as REDATAM, on the other hand, are extremely useful for the flexible production of tables based on the entire population, but they generally do not allow for multivariate data analysis.
International Conventions and Conferences cited:
International Conference on Population and Development (ICPD, 1994)
Beijing Platform for Action (BPA, Fourth World Conference on Women, 1995)
Convention on the Elimination of All Forms of Discrimination Against Women (CEDAW, 1979)
United Nations Millennium Declaration (2000) on http://www.un.org/millennium/declaration/ares552e.htm
Millennium Development Goals (MDGs) on http://www.un.org/millenniumgoals/
Gender Equality, UN Coherence & You – Glossary: Definitions A-Z on http://www.unicef.org/gender/training/content/scoIndex.html
2010 World Population and Housing Census Programme (UN Statistics Division) on http://unstats.un.org/unsd/demographic/sources/census/censusquest.htm
UNFPA Census Portal on http://www.unfpa.org/public/op/edit/home/sitemap/pid/6734
Download 3.92 Mb.
Share with your friends:
The database is protected by copyright ©sckool.org 2022