Setting out the objectives of surveys or censuses
The integration of a gender perspective into data collection should be taken into account from the stage of planning the data collection and setting out the objectives of the survey or census. Usually, the objectives of a survey or census are set out based on several factors: topics and policy issues that need to be addressed; review of previous data collections within the same programme; information available from other data sources; international statistical standards; country’s institutional capacity for collecting data; financial and other resources available. It is important that a gender perspective is incorporated in the review of previous data collections, both in terms of topics covered and the specific implementation, as reflected in data collection instruments and materials. National statistical offices can use their plan for the production of gender statistics to decide what gender statistics are to be collected by the particular survey or census developed at the time and what is to be covered by other sources of data. It is also important that gender specialists are involved in the process of developing data collection objectives. This process is typically based on extensive consultations between data producers and data users such as technical experts, data analysts, researchers, policy analysts, and policy makers. Both gender statisticians and gender specialists representing the point of view of data users and policy makers should be part of the team.
Questionnaire design and testing
Within the topics agreed to be covered by the survey or census, relevant gender issues should be identified and gender-specific conceptual and measurement issues should be taken into account (as discussed in chapter 2). These elements should be reflected in the questionnaire design, interviewer’s manual, and training of the interviewers and supervisors.
The questionnaire design should ensure that gender-specific conceptual and measurement issues are adequately reflected in the questions. It should benefit from consultations with a wide range of specialists, such as subject matter specialists, classification and coding experts, field supervisors, data processing staff, and data analysts. It is important that members of the team designing the questionnaire are knowledgeable of gender issues.
The language, terms, or phrasing of the questions should not induce gender biases. In particular, it is recommended that:
the questionnaire contains very short explanatory notes for the interviewer when needed, with more elaborate instructions, explanations of terms, or, in some cases, definitions and key concepts, provided in the interviewer’s manual.
probing questions are used in order to reduce under-reporting related to women, both to help respondents remember something that they may have forgotten and to help interviewers properly code the answers of some questions.
questions are written out in detail, with the reference period clearly specified. In some cases it may also be helpful to give examples of responses or the complete list of categories of answers.
the potential answers to questions should be categorized and pre-coded in such a way that answers related mainly to women are given the same importance as those mainly related to men.
questions should be kept as short and simple as possible, free from ambiguity, using common every day terms, so that all respondents, regardless of their educational level, have no difficulty in understanding them.
questions should not influence answers or be leading. Keywords in the questions should not apply exclusively to one of the two sexes (for example, “housewife” or “fisherman”) and they should maintain their meaning when translated into major languages of a country.
The questionnaire should be field-tested to ensure that both women and men understand the questions in the same way and to detect potential under-reporting or other bias related to either women or men.
Gender-related measurement issues and gender stereotypes should be addressed in the manuals for interviewers and supervisors. Manuals should have elaborate explanations on questions that may lead to underreporting or sex-selective underreporting (for example, domestic violence or economic activity); instructions and examples on how to use probing questions or lists (for example, in measuring economic activity); and, where applicable, instructions on how to code the answers (for example, in measuring self-employment or detailed marital status). The general language should be free of gender-based biases or other stereotypes related to the characteristics measured, and the examples given should not reinforce gender stereotypes.
Samples used should cover all groups of population, households, agricultural holdings or economic units known to have distinct gender patterns. The sample design should also ensure that reliable statistics are produced for both women and men in sufficient detail and allow disaggregation by other characteristics as required for meaningful gender analysis. For example, the sample of a survey measuring status in employment should be large enough to allow for the data to be analysed separated for female and male groups of employers or any other categories of self-employed, as well as further disaggregated by age group, rural/urban areas and educational attainment.
Selection and training of interviewers and supervisors
Selection and training of interviewers and supervisors are an important element in obtaining reliable gender statistics. Gender-related measurement issues and gender stereotypes should be addressed in the training for interviewers and supervisors. For example, the training should cover the situations when multiple respondents within the household need to be interviewed to avoid indirect reporting (for instance in recording literacy) or when information needs to be collected from household members that are most knowledgeable of the issue (for instance, household food consumption, number of children ever born). How to handle the interview environment when sensitive questions need to be asked, such as in the case of violence against women, or even in the case of women’s earnings, should be also included in the training. In addition, training should emphasize understanding of general gender issues related to the topics covered by the survey or census and how the data collected will address those issues, so that they can cope with issues and problems not specifically addressed in the manuals or training.
It is important that the field staff is selected on the basis of competence, and that both women and men are recruited as interviewers or supervisors. Certain types of surveys – such as violence against women surveys – need more careful selection and more extensive training of interviewers. The sex of the interviewer often play an important part in obtaining certain types of information from the respondents. Women, for example, are more likely to disclose information on sensitive topics such as violence against women or reproductive health to women interviewers than to men interviewers.
Data coding and data editing
It is important that gender bias is not introduced into the data at the stage of data coding and data editing. Data coding and data editing are data transformations that improve internal consistency and the conceptual soundness of data. Whenever possible, pre-coded responses are used in the questionnaires, and some of the data coding can be done by the interviewers directly in the field, by coding the respondent’s answer into the questionnaire. Other coding needs to be done by specialized coders using code books or computer programs, and some of the data errors may need to be fixed through data imputation. It is important that classification and subject matter specialists with training in gender issues are involved in formulating rules for data coding, data editing and data imputation, so that assumptions based on gender stereotypes are avoided.