NOTES AGRICULTURE & RURAL DEVELOPMENT 69532 ISSUE 64 JUNE 2012 Filling the Data Gap on Gender in Rural Kenya WHY GENDER DISAGGREGATED DATA? HOW GENDER DISAGGREGATED DATA? Agriculture is one of the most important sectors in Because data collection methodologies typically Kenya and its performance greatly affects the poor. focus on heads of households, and because most In addition to its importance as a source of food and household heads are men, women’s views on income, the sector directly accounts for 24 percent agriculture have been largely underreported. This of Kenya’s GDP , and for another 25 percent indirectly is a serious drawback because women are often through linkages with other economic sectors. It the primary farmers in their households. Failing to provides about 70 percent of rural employment. capture information from them leads to a distorted Kenyan agriculture is dominated by smallholder understanding of farming operations. farmers, pastoralists, and fisher-folk who together The general constraints women face in agriculture comprise around 4 million households. Farms are have however been comprehensively documented small, averaging one hectare. The sector faces many by the World Bank, FAO and IFAD in the Gender challenges including low productivity, poor market and Agriculture Sourcebook, the 2012 World access, low levels of commercialization, inadequate Development Report Gender Equality and infrastructure, and increasing weather variability. Development, and the 2010-2011 State of Food and Agriculture: Closing the Gender Gap for The Government of Kenya (GoK) with financial Development. These publications have affirmed that support from the World Bank is implementing the while women perform a very substantial proportion Kenya Agricultural Productivity and Agribusiness of agricultural work, they generally have less access Program (KAPAP) which aims to increase agricultural than men to a variety of resources. productivity and smallholder income by improving agricultural technology systems, empowering men QUESTIONNAIRE DEVELOPMENT and women farmers, and promoting agribusinesses. AND SURVEY DESIGN Women farmers in particular operate well below their The questionnaires and survey were designed to potential. Improving their capacity to accumulate align data collection with the needs of the agriculture resources and to retain income are important objec- sector, and to contribute to the development of a tives of the KAPAP . The project also seeks to provide sector-wide approach to gender-disaggregated rural women with a voice in decision-making bodies. diagnostics. They were also intended to inform the However, a major challenge quickly presented itself – a lack of existing information on gender gaps. Much of the information which is available is out-of-date, and most of it is based on case studies. This made it necessary for KAPAP to collect a unique set of gender-disaggregated baseline data to provide guidance on critical gender gaps. This information will contribute to an evidence-based gender policy dialogue in Kenya’s agriculture sector. Although appeals for gender-disaggregated data are frequently heard, the process is complicated and costly, and entails the need to overcome a number of methodological hurdles. This ARD Note describes the process of gender-disaggregated data collection that has been employed by KAPAP , and presents the key lessons learned from the preliminary results of the data analyses. Balancing many responsibilities. Photo: Asa Torkelsson. agriculture sector gender policy which is being developed by Kenya’s thirteen-ministry Agricultural Sector Coordination Unit STEPS IN GENDER-DISAGGREGATED (ASCU). DATA COLLECTION AND ANALYSIS Three questionnaires were designed: a household, individual, Number of interviews: 4,100 and community questionnaire. Two respondents were inter- Approximate cost: US$500,000 viewed in each household. The household questionnaire was Time frame: April 2010-May 2012 geared towards the ‘primary farmer,’ as self-reported by the 1. Development of survey instruments (April 2010) and household, and was used to collect information about activities first round of pre-testing (June-August 2010) that all household members engage in. The other key contribu- 2. Recruitment of Firm to undertake data collection tor to farming, usually the spouse, would then respond to the (January 2011) individual questionnaire. The household and individual question- naires were similar in content and partly overlapped. 3. Data collection: – Sampling of households (one week) PRE-TESTING – Preparation of list of target and control households to be interviewed (one week) The questionnaires were finalized using a highly iterative – Appointments for interviews (one week) process to ensure relevance across the varied farming condi- – Enumerator recruitment, shortlisting and interview- tions that are found in Kenya. The questions themselves were ing (one week) designed to be easy to respond to, effective in capturing the – Enumerator training (ten days) intended information, and easy for coding and inputting the – Questionnaire training and re-cap (four days) information. A number of duplications and misunderstandings – Field data collection (eight weeks, May-June 2011) were identified and weeded out. 4. Data entry, cleaning and database creation (August- October 2011) In collaboration with a capacity building initiative of Gender 5. Descriptive analysis (October-November) Focal Points in the water sector, they were pre-tested in peri- urban and rural settings in the Coast Province. A team from 6. Analysis of material, including development of Gender KAPAP and the World Bank piloted the questionnaires in the Policy Note (January-May 2012) North Eastern and Eastern Provinces. Important challenges were encountered related to specification of measurements, particularly of time and quantities. Pre-testing also helped ENUMERATOR RECRUITMENT AND TRAINING to refine codes and identify omissions (for instance creating Call for applicants was placed in the daily newspaper, and 54 codes for both sweet and food banana in the crop inventory enumerators with bachelor’s degree in agriculture-related disci- and adding livestock ‘lost or stolen,’ in addition to those which plines were selected for a two-week training which involved go- had died). It also taught the team to be as specific as possible. ing through the questionnaires question by question, clarifying For example, ‘registered groups,’ referred to those registered the meaning of each question and the information sought, and by the Ministry of Gender, Children and Social Development. using practical exercises on how to ask questions, probe for and record the responses. A team of 45 enumerators—32 men SAMPLING STRATEGY and 13 women—was finally selected for the data collection. To generate a sample with the necessary statistical power to Commitments were made to have a balanced representation represent a robust evidence base, sampling was based on a but it was difficult to fully meet such a target, since especially random selection of households representing a proportion of young mothers would find it difficult to be out of home for such regional households in Project areas. Most of these fell into a long stretch of time. Asymmetry in the gender of enumera- a set of panel households that had been generated earlier. tors and that of respondents could be a problem in some areas However, due to the new district set-up, several households of Kenya, but did not impact data collection areas according to previously interviewed as controls now fell within project areas the longstanding experience of the firm. and therefore the newly added households were sampled as controls. Multi-stage sampling methods were used in the DATA COLLECTION non-Project locations, using random selection from a list of all Data collection was carried out between 1 May 2011-25 June village households identified together with each village elder 2011 by nine teams, each comprising five enumerators, one or area assistant chief. Before data collection began, appoint- supervisor and a driver. Supervisors managed activities and ments on interview dates for each of the sampled households made spot checks to ensure data quality. In the first week of was made through area assistant chiefs and village elders to data collection, researchers visited the teams to provide techni- maximize the response rate. cal backstopping. 2 Overall, 4,141 interviews were conducted, comprising 2,529 novation, but it was challenging to identify two respondents for households (1,799 target and 730 control), 1,523 individuals, simultaneous interviews in a household, despite the appoint- and 89 community interviews. A majority (54%) of all respon- ments made prior to the visit. This resulted in frequent call- dents in the household survey were women suggesting they backs and more intensive fieldwork. are primarily responsible for farming. Around 31% of women In some areas, notably in the North-Eastern Province, it was a respondents headed their households while nearly 93% of the challenge to identify translators for the interviews and trans- male respondents headed their households. lation also prolonged interviews when respondents did not The individual questionnaire was collected from 566 men and understand Swahili or English. 957 women. In spite of great effort to locate the two partners for the individual interviews, many households were managed PRELIMINARY RESULTS by one primary farmer alone, and in most cases this was A majority of the communities reported a declining trend in a woman. The reasons for failing to interview an additional many of the aspects of community welfare during the last household member even after repeated visits was primarily five years, except for input availability and revenues gener- due to the fact that the additional household member could ated from raising livestock. Drought, increases in farm input not be found (45.3% of the cases), followed by there not being prices and food prices were cited to have adversely affected another qualified member to respond (36.8% percent of the the livelihoods of smallholder farmers. The average age of the cases) and the household being run by a single person (13.4% respondents of 48 years—with men being four years older than of the cases). women on average—suggests a rather aging farming popula- tion. A higher proportion of women (30%) than men (12%) had DATA ENTRY, CLEANING AND ANALYSIS no formal education and could not read nor write and gender Data entry, cleaning and analysis was done in the Statistical gaps were accentuated at the higher educational levels. The fol- Package for the Social Sciences (SPSS). Preparation of data lowing key findings are singled out from the survey report: entry templates began immediately upon start of field work • A higher percentage of men (81%) compared to women and data entry began in the second week of data collection, (19%) owned land individually. Areas of land owned by men and continued until mid August, 2011, engaging a team of four were about four times larger than those owned by women, clerks. Data cleaning took a period of two months, and involved and men also farmed larger parcels. a team of ten research assistants and three enumerators who checked for and corrected errors and/or omissions in the • The majority of women concentrated mainly in the produc- entered data. tion of food crops and farmed smaller land holdings than men who grew the same crops. Women had higher yields SURVEY CHALLENGES AND for selected crops (Irish potatoes, bananas and tea) but men LESSONS LEARNED registered higher yield for all other crops. A higher percent- age of men than women owned all types of livestock except “For the eight years I have been a research analyst, chicken. I thought I had learned it all but now I am learning again“ • A higher percentage of men than women sold crops. Men Research Coordinator, Consulting firm. decided on the use of revenue from the sale of most crops. Regarding livestock, women made decisions regarding Several challenges were faced during the design and data chicken only. collection. First, the survey instruments were revised several times to ensure they captured relevant information and for • Few men (27%) and women (13%) actually sought exten- the interviews to not be longer that 1.5 hours. This called for pa- sion advice. Half of the men and 36 percent of the women tience, technical inputs and experience to get the right balance who sought extension actually received. The main reason and the solid experience of the consulting firm came in handy. given was that it was time consuming or that extension The determination of the sample size to generate statistical agents were not available. Most respondents were satisfied robustness within a tight budget was another major challenge. with the extension advice they had received, and most had applied the advice. For those who did not, the main reason Secondly, the timing of fieldwork coincided in some areas with given was that putting the advice into place was costly. the long rains which contributed to logistical challenges such as vehicle breakdown and fuel shortages, and in others it coincid- • Although the proportions of women and men who were ed with the peak season in farming, and some interviews had members in groups were similar, larger proportions of men to begin later than scheduled to allow respondents to finish than women held leadership functions in groups. their farm work. • The rural gender resource gap was validated, as shown in Interviewing two members in a household was a useful in- the graph below. 3 ACCESS TO RESOURCES OF RURAL WOMEN AND MEN IN KENYA Having savings account Chicken 100 Receiving ownership 90 extension service 80 70 60 50 Having 40 Use of improved communication 30 maize seed equipment 20 10 0 Owning farm Mean income machinery Transportation Land ownership equipment Cattle ownership Men Source: Preliminary analysis of Individual Survey Data. Note: Percentages of wo/men Women respondents with access to the specified assets. Women’s income as a percentage of men’s. • Mean income for men was three times higher than for had actually learnt a lot regarding the costs and benefits of women. A higher percentage of men was engaged in off- their farming enterprise. farm activities compared to women and they earned twice We also learned that partnership and collaboration to collect as much income as women earned from these activities. gender-disaggregated data is a great way to overcome the Over half of the men had a savings account, whereas a increased costs involved in a quantitative survey approach. smaller proportion of women had an account. About a third of men and a fourth of women had applied for credit, with a high success rate for both. Men’s credit volumes were however larger. References . “In Kenya, Survey of Female Farmers Uncovers Challenges“ CONCLUSION World Bank. http://go.worldbank.org/ETKDJPYK70 This note has summarized the lessons learned from a gender- World Bank, Food and Agriculture Organization (FAO), International disaggregated survey in Kenya. The distinction between a Fund for Agricultural Development (IFAD). 2008. Gender in Agriculture ‘primary farmer’ and a ‘head of household’ proved to be Sourcebook. World Bank/IFAD/FAO. relevant and useful because women in most cases were the World Bank. 2012. World Development Report: Gender Equality primary farmers in their households, but seldom headed their and Development. World Bank. households. Often when implementing surveys we are cautious to avoid Prepared by authors: Andrew Karanja (AFTAR) and Asa respondent fatigue. This time, the team actually learned that Torkelsson (PRMGE). Reviewed by: Victoria Stanley, participating as a respondent in a survey can actually have Pirkko Poutiainen, and Eija Pehu from the Gender in Rural an empowerment function as well. Feedback suggested for Development Thematic Group. Edited by: Gunnar Larson (ARD). example that respondents, through the interview, said they 1818 H Street. NW Washington, DC 20433 www.worldbank.org/ard