Women are key players in development and a decade of CIET experience in 42 countries has produced several lessons worth sharing and many mistakes worth avoiding as we work to build the community voice in planning. CIET is a South non-governmental organisation consisting of an international cadre of professionals from a variety of disciplines who bring scientific research methods to the community level. CIET had developed a method of sentinel community surveillance (SCS) which involves communities in information gathering and analysis. The method facilitates the gathering of both quantitative and qualitative data, its analysis in term of impact, coverage and costs, and community-led solutions that are sustainable and locally relevant. The CIET method of facilitating community access to appropriate measurement technology, builds national and local evaluation capabilities in reiterative cycles.
This article distills the experience with gender issues using the SCS approach. The five identifiable methodological steps may be of relevance to other methods: the first is the analysis of existing data in terms of gender; second is the stratification of responses, analysing differences by sex of respondent; third is the processing of key findings by female focus groups, to obtain their interpretation of the data even when respondents are men; fourth, the epidemiological backbone of SCS permits analysis of gender-related risk and resilience; and, fifth, logistics of fieldwork are configured to maximise participation. With support from the World Bank, UNICEF and the IDRC, the methods have been adapted for use in public sector reform, national campaigns against corruption, and improving the effectiveness and transparency of the police, judiciary, environmental programmes and urban transport.
In many societies, it is difficult even to speak to the women. In some, it is barely acceptable to promote “the community voice”, both men and women, much less is it possible to begin with strengthening the voice of the most oppressed groups.
The CIET approach is to do whatever possible and sustainable to promote the full participation of both women and men. This makes good development sense, incorporating more energy and ideas into development and not only laying the burden on women. Also, as donor agencies increasingly realise just how detrimental the limited access of women has been to development, they place increasing emphasis on gender analysis and effective strategies to overcome the barriers to women’s access and participation. In some cases, a strong gender component is a precondition to funding.
The CIET method known as Sentinel Community Surveillance (SCS) has been implemented in over 40 countries around the world in the last decade (Andersson, 1985; Anderson et al. , 1989; CIETinternational, 1997). Many continue to use the approach and some have institutionalised it to the extent that it requires no external support. In many situations where this approach is used, there are considerable challenges to overcome the political, social and legal barriers to involving women.
The SCS method
SCS has proved useful in increasing representation of the community voice in both epidemiological and participatory senses. Household and community-level evidence produced on the impact, coverage and costs of development includes people from all walks of life and from all types of community. Ordinarily, SCS focuses on the use of evidence in local or national planning (Ledogar and Andersson, 1993). This may be at the level of a municipality, a city, a state, a number of provinces or an entire country (Arostegui, 1992; McTyre, 1993; Munroe, 1993). More important, perhaps, the participatory nature of the exercise allows people to have a say in the interpretation of data, helping them determine the way things are run in the future. At local level and in the way they aggregate to complement national information systems, repeated cycles could represent an indelible link between human well-being and governance (Andersson, 1996).
The reiterative process addresses one set of issues at a time. Each cycle consists of a cross-design of qualitative and quantitative techniques that permits a holistic picture of — and locally desinged solutions to — a particular problem (Ruijter, 1991). It is a cost-effective way to collect community data, presenting them in an appropriate form for planning at local, regional and national levels.
Central to SCS is interaction with the research partners — the communities. The product is the aggregation of data from the epidemiological analysis distilled through interaction with communities. Feeding back information to the communities, dialogue for action is stimulted within households, in communities and between communities and local authorities. The ideas behind evidence-based planning were first implemented in health related projects specifically those dealing with child survival and development. Over the last decade, SCS has been appliedin a wide range of development concerns: education, food security, child rights, environment, land mines, economic structural adjustment, urban transport, justice and corruption in the police and judiciary. This flexibility across sectors, considering the very real differences across sectors and countries, has become almost a defining characteristic of SCS. The very diversity of these initiatives — from almost mechanical vaccination programmes, through water and sanitation and education — led to the development of a pragmatic cross-design synthesis of methods.
A cross-design included critical review of existing data, household surveys, physical measurement, rapid anthropological procedures and institutional reviews. With the household survey as the methodological cornerstone, SCS allows the application of powerful modern epidemiological tools to look behind the indicators. Thus a combination of quantitative and qualitative fact-finding instruments produce reliable, actionable data, all the while building the national capacities to make the scheme sustainable in the longer term.
The CIET approach to gender
Some societies have accessible and strong women’s networks, agitating to improve their position; partnerships with these groups make implementation relatively straightforward. However, in many settings where CIET works, the fact that the CIET Research Fellow might be female or that the design group is made up of women does not guarantee a full gender analysis. In this sense, SCS is a learning process where a suitable gender approach is continuously being tested and refined in each of the setting where SCS is applied. So far, five methodological steps make a pragmatic starting point.
The first is a gender stratified analysis of existing data; simply asking for existing data to be broken down by sex can provide a starting point. Second, in households and key informant questionnaires, whatever the subject matter, it is standard CIET practice to stratify all responses by sex of the respondent; this prevents a numerical bias in favour of men translating into a gender bias in the analysis.
Third, in societies where men frame the questions and the answers, female focus groups are used in SCS design, in interpretation of findings and to develop appropriate strategies for change. Fourth, gender is also treated as a factor in risk and resilience analysis, focusing on the differences in sex and generational categories. Fifth, logistics are given special attention and local women researchers are almost always recruited to the CIET team. In countries where women and men cannot travel together, for example, an extra budget allocation is required for transportation and often, recruitment and security.
Analysis of existing data
Each SCS cycle begins with the critical review and analysis of existing studies and data from routine sources. Much of the data available is only partially processed. Part of the reason for non-use of these data up to now is the lack of analytical capacity next to the volume of data produced. Specific steps can be taken to improve this, and the SCS links with existing data are contemplated in this light: even if the data from routine sources are partial, beginning to use them and to compare them with other sources is a first step to improved capacity for their management and, over time, the improvement of the data themselves.
A first step to any CIET analysis is disaggregation by sex. For many traditional data sets, national and sub-national, simply requesting the data in this form will produce it.
Disaggregation of responses
The CIET Mali survey of access to services illustrates a disturbing but typical gender issue in community-based fact finding, and one way of managing it. In rural settings in this predominantly Muslim country, it is difficult for men to interview women at household level, and even more difficult to have women interviewers move around alone in remote rural communities. Some 91% of the household questionnaire respondents turned out to be men. Reducing the data by simplistic average, pooling male and female responses, produces the obvious problem of a bias towards the male response.
With CIET methods, a disproportionate number of male responses does not represent an insurmountable problem: all responses are gender stratified before combining the results (Mantel, 1963; Mantel, 1959). In practical terms, this means looking at what female respondents said separately to male responses. Provided there are sufficient female responses to analyse, the issue is not if the responses make up 80%, 50% or 10%;it is a question of how different they are to the men’s responses. If they say the same thing, data are pooled. If they do not say the same thing, responses are presented separately by sex of the respondent and the difference becomes an issue for further analysis.
In the Mali survey, the responses of men and women were similar throughout with the exception of health, which more women identified as important (65% vs 50%). There was also a difference in the type of service that women found unsatisfactory. Women were twice as likely to report problems in dealing with administration issues, whereas men were twice as likely to report problems dealing with police.
Viewed as the first step in a longer term process, the fact that one in ten responses was female in Mali is actually a positive achievement. Many surveys of similar communities have returned much lower female participation rates. The usual SCS approach is to ensure some female responses; or at least some male responses if the respondents are predominantly female, as was the case in Nepal and Nicaragua (National Planning Commission, 1996; Andersson and Arostegui, 1995)
A realissue for further investigation is the difference in conditions of those women who do or can respond, compared with those who do not.
A particular type of gender bias in responses is evident in surveys in South Asia. A study of gender and primary school drop-out in one of these countries revealed that many households simply did not report the female children when asked about school attendance. Overall, six out of ten children in the study were boys. In some communities, as many as eight out of ten were boys. In this case, it is inadequate simply to analyse drop out separately for the two sexes. Households that do not report presence of female children, it can be argued, may be more likely to prevent girls attending school. Ignoring their exclusion from the survey underestimates the female drop out rate, and any solutions will be, at best, partial.
There is no simple remedy to this non-reporting. Such efforts as could be made, were made. In the questionnaire, interviewers specifically asked about female children; but if respondents consider that girl children “do not count as people” and do not report them, the survey will be irremediably biased. A clue about the possible size of the bias was obtained by analysing only those households where both male and female children were reported; this was compared with households where only female children were reported. This does not solve the problem of non-reporting, but it does begin to focus attention on it as an issue for further consideration.
It was also possible to code communities as high, medium or low levels of non-reporting female children, using the male:female ratio as an indicator. Female dropout from primary school was much higher in communities where non-reporting of female children was high, confirming perhaps the non-preference nature of these communities.
In the Balochistan survey of gender gap in primary education, an effort was made to quantify the undeclared girls in two ways. First, based on the assumption that there should be a similar number of boys and girls aged five to 12 years, the number of girls was reinflated, considering none of the undeclared girls as enrolled (Figure 1). Secondly, the answers of male and female respondents are considered separately. Male respondents, it turned out, reported that 59% of children in their households were boys; female respondents reported that 50% of children in their households were boys. These data confirm the notion that male respondents do not count girls. There was significantly higher female non-attendance and drop-out in households where there was a female respondent. Part of the explanation of this could be that these are different, less advantaged, households. Another part of the explanation is that some male respondents ignore girl children.
Focus groups interpret results and design strategies
One recurring problem experienced in implementing SCS is that official counterparts are usually men. Effort is usually made, frequently withoutsuccess, to include effective women’s participation in these design stages.
An example of how this affects the work comes from one South Asian country, in an SCS cycle investigating the causes of childhood malnutrition, in which the women’s social position, labour and wife-beating were included as issues for inquiry. In the piloting, it turned out, a staggeringly high proportion of women reported being beaten repeatedly over the previous year, and indications were that this might have something to do with the way they see themselves and their offspring. No woman who was asked in the pilot expressed this sentiment, many of them providing more detail than was requested. Yet at the very final stage of government clearance, this question had to be omitted as “socially unacceptable”.
A key to the success of SCS is that national planners, service workers and community members all feel comfortable with the reiterative contacts and with the way the data are obtained and handled; therefore the question of wife beating could not be included in the household questionnaire. A substitute for the question was permitted in another instrument, the focus group. Here the question, vetoed by the government steering committee, was put to 144 groups of 8-10 women, one in each sentinel community, who discussed the experience with beating and how they felt this impacted on their child-rearing practices. Although losing the quantitative data, it was thus still possible to link opinions emerging from focus groups with local levels of malnutrition.
Even in this indirect form, the impact was patent: in communities where women’s focus group concluded that wife-beating was a particular problem, children under the age of three years were more than 50% more likely to be stunted (more than 2 standard deviations below median height for age) (National Planning Committee, 1996b). As comfort levels of the key players increases, it will hopefully be possible to get a more direct measure of this important determinant of malnutrition and its link to the abuse of women.
The difficulty of contacting women and the potential bias toward the male view in the quantitative data collection was very clear in Mali; there focus groups were run solely with women. Thus the women’s perspective, although not as widely representative as the men’s, was brought out and differences between women’s and men’s concerns identified. The challenge in successive cycles in Mali is to find other ways to increase the women’s voice.
Gender risk and resilience analysis
In every survey since the inauguration of CIET in 1986, highlighting inequalities has been a cornerstone of the analytical framework. It is standard practice to look at malnutrition, vaccine efficacy, costs of corruption and land mines, access to education, justice and transport — everything — as potential gender issues. For example, in a survey of indigenous youth on-Reserve in Canada, a male youth is 40% more likely to resist smoking than a female (95%CI 1. 2-1. 6); female youth were more resilient to peer pressure in their smoking uptake (CIETcanada/Wunska, 1996).
In Bosnia, there were different child care practices for male and female children in the different ethnic groups during the four years of war (CIETinternational, 1996 and 1994). There was a reduction of breast feeding over the past three years, to less than one half of 1989 levels, after an improvement early in the crisis (1991-2). Unlike most other son-favouring societies, male children were more at risk of never breast feeding than females (odds ratio 1. 4, 95%CI 1. 1-1. 8, p=0. 07). This gender gap in favour of females was larger among Muslim and Croat children, and reversed among Serbs. The reason for this is not immediately clear; breast feeding patterns were not worse in communities near the conflict lines, and there were no notable differences among residents, host households and displaced-headed households.
In Nepal, literacy (reported ability to read and write a simple letter) is highest among 11-15 years old (77% in males and 56% in females) (National Planning Commission, 1996). It falls steadily with increasing age, down to 29% in males and 2% in females of 61 years or older. Overall, females have nearly four times the risk of males of being illiterate. Households reported a drop-out rate of children enrolled into school of only 3% (weighted value), with boys being at half the risk of girls of dropping-out. Pursuing the analysis into the possible solutions, female teachers are associated with higher attendance and lower repetition and drop-out in classes one and two and for both boys and girls.
In Balochistan Province in Pakistan, a similar survey demonstrated a gender gap in school enrolment of children, more pronounced in rural than urban areas (Figure 2) (CIETinternational, 1996b). The analytical framework applied in this analysis, geared to actionable results without major additional injection of resources, was to focus on communities without gender gaps and to compare them with those with a gender gap. The immediate implication of this analysis is that, under prevailing conditions in rural Balochistan, it is possible to have no gender gap. The communities without a gender gap are, in this sense, held up as best case scenarios. In the Balochistan survey, it was found that Parent Teacher Associations (PTAs) and religious leaders (Pesh Imams) could have a positive effect on school enrolment; however, despite enrolment rates being higher when either of these is active in primary education, the gender gap tends to be larger where we see more boys than girls being enrolled.
The head of household is a very real economic and repressive category in many countries of the South. The SCS concern is that, in male-dominated societies, female-headed households have quite different life chances; some of these, if well understood, can be redressed by development programmes. The usual CIET risk and resilience analysis bases categories on demographic characteristics of the household (defined usually as those who eat from the same “plate”). We group together those households that do not have males over a certain age, the age (typically around 20) depending on the country. For example, in Nepal, children from households with a female head were more likely to be enrolled into school. Household composition is a strong risk factor for access to basic services and food security.
An interesting finding was returned from gender stratification of school attendance in Afghanistan (CIETinternational, 1997). Perhaps predictably in the context of the suppression of female education under the Taliban, very few girls are at school in the south and centre of the country. However, even in the heartland of Kandahar, women have organised illegal underground schools to educate their daughters the numbers are not great, but they are sufficient to show a few percent attending school when none are supposed to.
To build the community voice into planning requires more than manipulation of the data. CIET methods are defined by the capacity-building — of women and men — implicit in the fieldwork. However, there are sometimes considerable cost implications of a gender approach to community-based measurement: sometimes we have to invest twice or three times as much per cycle for security and cultural reasons. For women to be part of the CIET team, in some countries they must be accompanied by guards and twice as many people requires more vehicles. There investments, are a source of tension in the discussion with donors.
An example of how CIET approaches these operational issues in studying the gender gap in a challenging environment is the “Gender in Primary Education” cycle in Pakistan. Considerable discussion was held with local and Government counterparts as to who should be interviewed. Some of the issues grappled with included the following:
- If male attitudes lie behind girls not going to school or dropping out, and this was the general feeling in the design group, then it was important to understand their perspective to work out ways to address it. Thus the men should be interviewed.
- The women’s view is also important because they can perhaps provide clues on how to deal with the men. Thus the women must also be interviewed. A women’s focus group should also be conducted to concentrate opinions on what can be done about the gender gap in primary education.
- It is difficult to have both sexes traveling together in Pakistan. Yet, in order to get the women’s views in focus groups, these must be facilitated by women. In order to get the men’s views in focus groups, they must be facilitated by men.
The operational decision was to recruit for each team one male supervisor and five male interviewers (the supervisor facilitated the male focus group (FG)); one female FG facilitator did household interviews, to get at least 1/6 answers from women; one male journalist interviewed the Pesh Imam and monitored the male FG; one female journalist interviewed the Parent Teachers Association (PTA) and monitored the female FG. In order to facilitate the travel of two women with seven men, two vehicles per site (one small, one large) will be needed, and this of course will increase costs. This operational formula was applied across 250 communities that make up the national community voice framework in Pakistan.
The last 12 years of systematising an approach to gender issues in SCS has produced several lessons worth sharing and many mistakes worth avoiding. The five methodological points identified might have a broader application in clarifying gender issues. A perhaps bigger contribution is in the substrate SCS provides for increasing women’s voice in planning.
SCS offers a pragmatic way to obtain community-based evidence for planning and to encourage public services to be increasingly beneficiary-driven, and to involve women increasingly in this process. Once workable instruments are designed, SCS follows a tightly focused process of fact-finding, analysis and dissemination of results. The large samples of household permit a baseline of occurrence rates, needs and perceptions. Insofar as women’s voice is woven into this process, their role as citizens and clients will be recognised increasingly.
Building women’s voice into planning involves several simultaneous learning curves. Challenging as it is, the easiest to deal with is the skill to conduct the community-based measurement and feedback. Then there is the challenge of building a ‘culture of evaluation’. Successive cycles that generate clear and compelling evidence make it increasingly normal to ask what works and what does not, to have information on who has access to which services. As decision-makers at policy, service and household level find practical application of the evidence, (ideally) they begin to see evaluation as integral and indispensable to programmes. A third learning curve has to do with the gender gap. Consistent questioning of the differentials between males and females, or of the disadvantaged social position of women can help to draw attention to these inequities. Constant demands to have routine data produced with gender break-outs will eventually make this standard practice. Solutions posed consistently in terms of gender, by the same token, might over time increase consciousness of the validity of this perspective.
Our modest gains in systematising an approach to gender are more than matched by the challenges that remain. In some cases, like the under-reporting or non-recognition of girls, the discrimination is so deep seated that repeated and increasing insistence is the only way forward, attempting on each successive cycle to reduce the reporting gap. Identifying gender differences in responses is important, but the real issue is to address the difference in conditions of those women who can respond, compared to those who cannot.
Bringing the women’s voice into planning would not be possible without the commitment and hard work of the thousands of community-based researchers trained by CIET in dozens of countries. Insistent support for the process has come from UNICEF, UNDP, UNHCR, World Bank and the Economic Development Institute of the World Bank, World Food Programme, CIDA and IDRC Canada and various foundations, NGOs, and universities.
There are also the CIETistas out there, committed and moving from country to country at a moments notice to provide technical support, often in difficult circumstances.
Andersson, N. (1996) Public participation in decision-making. In Governance, public participation, decentralisation and integrity, ed. P. Langseth and K. Galt, pp. . . . . Danish Ministry of Foreign Affairs and EDI World Bank, Copenhagen.
Andersson, N. And Arostegui, J. (1995) Nicaragua Service Delivery Baseline Survey Report. CIETinternational, Managua.
Andersson, N. , Martinez, E. , Cerrato F. , Morales E. And Ledogar RJ. (1989)The Use of Community-based Data in Health Planning in Mexico and Central America. Health Policy and Planning 4(3), 197-206.
Andersson, N. (1985) Impact, coverage and costs: an operational framework for monitoring child survival, emerging from two projects in Central America. UNICEF, Guatemala.
CIETcanada/WUNSKA. (1996) First Nations Youth Inquiry into Tobacco Use. Medical Services Branch, Health & Welfare Canada, Ottawa.
CIETinternational. (1997) Service Delivery Surveys: Applying the Sentinel Community Surveillance Methodology Country Overviews. Economic Development Institute of the World Bank, Washington D. C.
CIETinternational. (1996) The 1996 Vulnerability Survey. CIET, New York.
CIETinternational (1996b)Gender Gap in Primary Education in Sindh Province, Pakistan. Education Department, Government of Sindh/UNICEF /CIET, Karachi, Pakistan.
CIETinternational. (1994)Food security, Social support and Agriculture in Central Bosnia & Herzegovina. CIET, New York.
Ledogar, RJ and Andersson, N. (1993)Impact Estimation Through Sentinel Community Surveillance: An affordable epidemiological approach. Third World Planning Review 15(3), 263-272.
Mantel, N. and Haenszel, W. (1959)Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 22, 719-748.
Mantel N. Chi-square tests with one degree of freedom: extensions of the Mantel Haenszel procedure. J Amer Stat Assoc 1963;58:690-700.
Massoud, N. (1996) Enquête sur les services publics au Mali. CIETinternational, New York.
National Planning Commission. (1996)Nepal multiple indicator surveillance. Primary education. Final Report of the Second NMIS Cycle. NPC/CIET/UNICEF, Kathmandu, Nepal.
National Planning Commission. (1996b) Child feeding, development milestones and nutrition; Nepal Multiple Indicator Surveillance Cycle 4 Report. NPC/CIET/UNICEF, Kathmandu, Nepal.
Ruijter, Jose M. (1991) Sentinel Site Experience in Angola. Convergence XXXIV(4).
Schacter, Mark. (1995) Mali Service Delivery Survey: Summary Report of Preliminary Results World Bank, September 1995 and World Bank EDI: SDS Highlights (4 January 1996).