Doing research and ‘doing gender’ in Ethiopia’s agricultural research system

Abstract This article presents data from an experimental gender audit that was completed of the national agricultural research system in Ethiopia. The results show how agriculture researchers understand and practice gender in Ethiopia and what the research outputs say about Ethiopian Institute of Agriculture Research (EIAR)'s ability to incorporate gender across the research process. The article applies gender mainstreaming theory and feminist research methods and methodologies to frame the analysis. This body of scholarship highlights crucial questions that go beyond the counting of male and female participants in research surveys and the recruitment of female informants, a problem at EIAR. The article argues that focusing on gender mainstreaming ignores the power relations that currently maintain the status quo and prevent a substantial approach to ‘doing gender’ from evolving. In order to improve gender equality in Ethiopia’s agriculture sector, EIAR needs to expand the methodologies currently used and include and raise awareness for feminist methods. This study adds to the emerging literature on Ethiopian women studies, gender mainstreaming, institutional reform, and gender research methods.


Introduction
The Ethiopia government has made significant gender commitments into consecutive national development plans: Sustainable Development and Poverty Reduction Program (SDPRP) (2002/3-2004/5); the Plan for Accelerated and Sustained Development to End Poverty (PASDEP) (2005/06-2009/10); the Growth and Transformation Plan (GTP) I (2010/11 -2014/15) and GTP II (2015/16-2019/2020). These plans consider gender inequality as an obstacle to development and advocate for gender mainstreaming as the tool to address equitable poverty reduction and economic growth.
Many lessons were learned during the first Growth and Transformation Plan (GTP) (national development plan) that culminated in a redoubling of efforts during the GTP II period. GTPII places women and youth as a cross-cutting sector which means it should be included in all government programs. Consequently, gender is increasing in importance in Ethiopia's agriculture sector and in national policies where there is a desire to improve food security for all citizens. GTPII also stipulates that to 'strengthen women's empowerment and enhance their economic benefits, integrated and wellorganized awareness creation and capacity development interventions need to be pursued' (GTPII:74). These commitments highlight the magnitude of the task and that 'gender' is associated with 'women' in the policy.
However, paradoxically, the Ethiopian government passed the Charities and Societies Proclamation (No. 621/2009) which reduced the number of women's NGOs and the ability to advocate for women's rights. Article 14 (5) banned foreign and Ethiopian resident charities from working on rights issues if they receive more than 10% of their funds from foreign sources. This proclamation (along with the anti-terrorism proclamation from the same year) had a devastating effect on civil society in Ethiopia, especially on women's rights organizations. The number of registered NGOs reduced from 2275 in 2009 to 1701 in 2011 (Oxfam, 2016). Prior to this proclamation, the Ethiopian Women Lawyer's Association was a strong advocate for women's rights (Burgess, 2013:102-105).
Moreover, the 'cross-cutting' understanding of women and youth in the GTPII is insufficient to meeting gender equality. Gender is an intersectional category of analysis, which means the category of 'women' does not represent actual and fixed groupings of people (Anthias, 2002). Thus, placing women and youth as cross-cutting issues offers a narrow understanding of inequality and risks leaving those with little power out of the development process. As West and Fenstermaker (1995:12) explain gender is not just a role or characteristic it is a mechanism that can reproduce the status quo. Gender identity is developed, reinforced and overcome through multiple relationships and interactions (Connell, 1993: xi). Furthermore, as Collins (1990) notes gender is co-constituted along with other statuses such as race, religion, class, etc. While the ways social categories/statuses converge vary from context to context, in Ethiopia, a country whose administrative structures are based on an ethnic federation with a highly religious, rural and young population, gender often works in tandem with age, ethnicity and religion to produce a complex picture of gender inequality and lived experience. As Tariku and Latchanna (2016) explain such intersections are overlooked in Ethiopia, particularly in the policy realm.
This paper explores the way gender equality work is conceived and practiced in one particular government agency, the Ethiopian Institute of Agricultural Research (EIAR). Since 1966, EIAR has the responsibility to inform the government's agriculture policymaking and assist in transferring technological knowledge across the nation. There are 18 agriculture research institutes across the country with Addis Ababa as the headquarters of EIAR. All of EIAR's research could contribute to fulfilling GTPII targets, including those related to women. Through analyzing this organization, this paper explores the way gender is integrated and will examine some of the theoretical understandings of gender operating at this agency and the implication of the agency's conception in its gender mainstreaming efforts.
EIAR sits under the Ministry of Agriculture and Natural Resource Management (MoANR) in the government structure but has not implemented MoANR's mainstreaming guidelines as yet. MoANR's Gender Mainstreaming Guidelines explain that: 'mainstreaming is not an end in itself but a means to achieve the goal of gender equality. It involves ensuring that gender perspectives and attention to the goal of gender equality are central to all activities; policy development, research, advocacy/dialogue, legislation, resource allocation, planning, implementation and monitoring of projects and programs.' The guideline further defines gender mainstreaming as 'an approach or strategy, which ensures that: All development efforts are geared towards addressing the experiences, needs, and priorities of both men and women at all levels. Developed outcomes benefit women and men equally. Gender disparities are not continued or made worse.' EIAR is also connected to the Agricultural Transformation Agency (ATA), an advisory agency for MoANR and EIAR. ATA developed a strategy for gender in the agriculture sector that includes strengthening the national agricultural research system to develop problem-solving technologies related to climate-smart agriculture and the needs of female farmers and it calls for stronger gender mainstreaming and integration. Although ATA initially requested training for EIAR, a consortium of gender experts decided that a gender audit would be a necessary starting point. Gender audits are the usual method used to assess a 'development institution' 1 from a gender orientation.
The focus of this article is on the research side of EIAR and how agriculture researchers understand and practice gender and what the research outputs say about EIAR's ability to incorporate gender across the research process. The limited available literature suggests that female farmers are considered 'co-farmers' 'by those individuals with significant influence in research, extension and development positions' (Tsehai, 1991). As it is necessary to retain a focus on improving women's access to government institutional offerings (Rees, 1998:41), this paper questions whether focusing on mainstreaming is the best approach for EIAR to take. The structure of this paper is as follows: research design, literature review, introduction to Ethiopia's context and gender mainstreaming, results (approaches to gender; resources; communication and knowledge management: documentation analysis and bibliometric analysis) and conclusion.

Research design
In order to examine gender mainstreaming and gender research capacity, this paper employs a gender audit method. A gender audit is a form of assessment or evaluation for gender mainstreaming. A gender audit can help set a direction for change and help an organization to identify some of the hidden gender dynamics that may work against equality either internally (e.g. operations) or within an organization's programs and partnerships. 2 The results of effective gender audits can be used to measure progress once the audit is repeated and to set a direction for organizational change.
However, EIAR did not have an institutional gender strategy to measure the audit against. Other gender audit manuals were consulted 3 but none had been designed for a research institute. Therefore, instruments were developed to conduct a general institutional assessment of a research institute. Given that the institute is a scientific research institution, the audit should have quantitative components, but feminist methods require participatory and reflective methods, therefore, the data collection methods used were mixed. The instruments were developed collaboratively with the audit team. Given that not all eighteen institutes and sectoral units in EIAR could be assessed because of time limitations and capacity, EIAR chose four directorates (Crop, Natural Resource Management, Livestock and Agricultural mechanization) and two internal units (HR and Communications) at the Addis Ababa headquarter level to be audited due to institutional interest and a desire to capture internal processes as well as research. An online survey was developed to reach as many people as possible, along with some key informant interviews and focus group discussions. Like with many other gender audits, a documentation analysis was included. The bibliometric analysis was a new addition to the audit methodology in order for the results to appeal to a research institute's core businesspublished research outputs. Consequently, the methodology and sample used in this paper are experimental.
A mixed methods approach was developed involving: key informant interviews (KII) -9 Female (F)/9 Male (M), focus group discussions (FGD) -4F/6M 4 , online survey (a link to the online survey was sent multiple times by the Director General (DG) to all 18 research centers resulting in 70 responses (8F/62M) out of 3000 staff and reached most of EIAR's regional centers), bibliometric analysis (from 2011 to 2015 using the Scopus citation index, a general search was made using the term 'Ethiopian Institute of Agricultural Research' in the Affiliation field and then key search terms of 'gender' and 'women'); and, A documentation analysis of 55 internally produced documents voluntarily collected from units and assessed by an audit team comprising of the International Maize and Wheat Improvement Center (CIMMYT) and EIAR staff. Checklists were developed collaboratively for the documentation analysis (and iteratively as EIAR produces many types of documents).
The audit team involved three men and eight women. Two audit team members were internationals (1 Australian; 1 Ugandan) and nine were Ethiopians. Five were from EIAR, four were from the International Maize and Wheat Improvement Center (CIMMYT), one from the International Livestock Research Institute (ILRI), and one from The International Center for Agricultural Research in the Dry Areas (ICARDA). The Australian national was the lead investigator but all instruments and data collection was a collaborative process. Moreover, the focus group data collection was done in Amharic to allow participants to more freely discuss their opinions and to 'enhance rather than diminish' the respondents' sense of self-hood and agency (Mama, 2007:21). The interview respondents were given the option of being interviewed in English by the international lead researcher or in Amharic by CIMMYT staff as a way to minimize power inequities in communication and between researcher and participant. The documentation analysis was a participatory process involving the EIAR gender audit team plus CIMMYT team discussing and assessing each document. However, the majority of the analysis was undertaken by the lead investigator and shared with EIAR's gender director for input. Moreover, a number of presentations on the results have been given to EIAR generating a rich discussion about potential pathways for change and reflections on the methodological approach. This paper only presents the results relating to gender in research. Aspects relating to the workplace and capacity building are presented elsewhere (see Drucza, Tsegaye, Abebe, Giref, & Abebe, 2017).

Literature review
Gender mainstreaming has been an active development approach since the 1990s. With a gender mainstreaming approach, every part of an organization is responsible for ensuring an even impact on women and men (Benschop & Verloo, 2006). However, debates about the effectiveness of gender mainstreaming as an approach to development have grown (Alonso, 2017;Bacchi & Eveline, 2010;Clisby & Enderstein, 2017;Eveline & Bacchi, 2005;Moser, 2006). Some of the criticisms about gender mainstreaming concern the way it is donea technocratic, once off fix when sustained engagement, refinement, and capacity building is required (Daly, 2005;Verloo, 2001). Donors who fund gender work in low-income countries have a mixed record (Moser & Moser, 2005) and yet, it has rapidly risen as the gender approach in international development. For Charlesworth (2005:16) this rapid uptake reflects gender mainstreaming's 'ambiguities, weakness, and lack of bite'. Indeed, gender mainstreaming is often vaguely or inconsistently defined within a sector, organization or a country and lacks standardization (Moser & Moser, 2005:585;Walby, 2005:455). Few results for women have come from gender mainstreaming (Milward et al., 2015). Milward et al. (2015) summarized the criticisms surrounding gender mainstreaming into the following: a. 'de-politicization' whereby gender as a feminist category involving power hierarchies is subverted; b. 'technocratic' an endless supply of tools and checklists are developed without cultural nuance; c. 'gendered organizations' situates an organization and its structures as masculine and in need of reform for lasting change to ensue; d. losing sight of the people who are meant to benefit from the changes; e. 'governmental technology'gender mainstreaming will only ever achieve for women what the institution seeks (or is able to) to achieve within the wider project of neoliberal governance.
In short, gender mainstreaming overlooks power relations and the structures of inequality that keep women unequal to men.
What would make gender mainstreaming a more effective approach? Hankivsky (2005) suggests gender mainstreaming has to connect to feminist theory as well as debates about diversity and participatory processes to have any relevance. Connell (2014) argues that Southern feminist researchers, especially from Africa, should spend time theorizing gender in their countries and voicing their opinions to contextualize the meaning of gender in their country before mainstreaming can hope to be effective.
While a considerable body of African feminist literature exists, more is needed, because Africa is a large continent, with extreme heterogeneity within and between countries. For example, in Nigeria, being a feminist is advocated for by Adichie (2015) and yet, Oyewumi (1997) cautions against a Western feminist approach as one that advances colonial views of African women's subordination. Literature reviews like those by Adomako  should be more regularly updated, as some African feminist literature remains unpublished for reasons outlined by Mama (2007). Similarly, the collection by Cole et al. (2007) advocates for a discourse on gender in Africa that goes beyond the conventional dichotomies and polarizing identity politics that plague attempts to 'do gender'. Meanwhile Akin-Aina's (2011:72) discussion on the African feminist movement and its 'distortions, misperceptions and silences' raises the point that perhaps gender mainstreaming will never serve all African women. For Akin-Aina (2011:86) African feminism must include a diversity of women, and diverse advocacy and negation strategies.
African feminists call for more funding for feminist research, more employment of female (and feminist) researchers within research institutes, and a revision and update on the way women's studies are taught in African universities (Tamale and Oloka-Onyango, 2000;Mitchell et al., 2010). They also lobby for certain research methods and critique the false idea of scientific neutrality (Adomako Tsikata 2001). Quantitative methods, large-scale household surveys, and formal interviews are overly used (Imam, Mama and Sow, 1997) when qualitative, participatory and actionorientated research methods are needed to unpack an African notion of gender equality (Blystad, Haukanes, & Zenebe, 2014;Gana, 1998;Tsikata, 1989). Similar arguments are made by Western feminists about a male bias in both science and research and how dominant, positivist science standards are historically tied to imperialism and colonialism (Harding, 1991(Harding, , 2011Wajcman, 2004). Park (2018) highlights the need to adequately address the meaning of gender equality from the perspective of different African women living in different contexts.
In terms of Ethiopia specifically, the 'depoliticized national women's discourse' (Biseswar, 2008a:405) leads to a male bias in the agriculture scientific literature and in extension practices which is perhaps best reflected by the lack of gender-focused publications available. Buchy and Basazne (2005) demonstrate that the gender bias of individuals within agricultural organizations as well as the gender bias of organizations, not only prevent the development of a gender-equitable workplace, but also ultimately result in gender discriminatory outreach. Cohen and Lemma (2010) found that (usually male) agricultural extension workers prefer to work with the (usually male) household decision-maker and cultural taboos restrict their interaction with women 5 . Home gardens and poultry tend to be defined as a part of 'home economics' and thus women's domain, which excludes women from other kinds of agricultural extension advice, training, and credit (Ogato et al., 2009). Blystad et al. (2014) explain challenges situated around the language of women's development. In short, Ethiopia has a predominantly patriarchal culture, and so too does the government extension and research system.
Feminist methods 'subvert various aspects of the "scientific method" in order to produce better accounts of gendered lives' (Krook 2007:15-16). West and Zimmerman (1987: 147) discuss 'doing gender' whereby gender is an action that is situated within a particular context. Women and men are not passive victims of structural conditions because they consciously or unconsciously subscribe to power hierarchies and social norms of behavior and 'perform' these ideals (West and Zimmerman, 1987). Thus, 'the performance of gender is both an indication of and a reproduction of gendered (as well as raced, classed, generational, and sexed) social hierarchies' (West and Zimmerman, 1987:147).
The hidden power dynamics that structure women's lives are exposed through feminist methods, and the interests and tendencies of subalterns become a central site of study in order to counter the dominant group's power to define, assert and control (Leurs 2017).
Not only do feminist methods raise epistemological questions but they also offer ways to safeguard against the reproduction of power in research design and interpretation. Self-reflexivity is a central feminist methodological concept that allows researchers to interrogate their own subjective role as an expert (Ackerly and True, 2008:698). Hence 'doing gender' in EIAR requires understanding: (i) how agriculture researchers understand and practice gender, (ii) their own methods and 'expertise' and (iii) what the research outputs say about EIAR's ability (and willingness) to incorporate gender across the research process. Moreover, the researchers involved in this study worked to reduce power inequities by giving respondents options of participating in different data collection processes, some of which were participatory, gave respondents the choice of language used and the researcher, and by having a mix of internal EIAR staff, and external (albeit partner) researchers involved in the process.

Ethiopian context and gender mainstreaming
The history of gender mainstreaming and equality in Ethiopia is merged with religion and politics and is best described as a torturous paradox. The Orthodox church plays a dominant role in Ethiopian society and church leaders have contributed to women's subordinate position (Biseswar, 2008b). In the early records of the Orthodox church, high-ranking women had political and administrative roles (Crummey, 2000: 94-99) and 'the few influential Ethiopian queens and other noble Christian women were reportedly very active in building and patronizing churches and monasteries' (Tsehai and M€ uller, 2015:144). However, Tsehai and M€ uller (2015:144) identify the role of the Orthodox church in suppressing women's equal rights to education and excluding women from certain professions. More recently, Østebø and Østebø (2014) discuss how local religious leaders pretend to accept gender equality when talking to development practitioners but continue traditional practices as usual.
Ethiopia's political history has paradoxically suppressed and promoted women's rights. Ethiopia's Derg regime (1974 to 1987) suppressed almost all forms of collective organization and human rights (Rahmato, 2004). Nevertheless, Zewdie (2014) documents the actions of a number of revolutionary women fighters and protesters, making the point that women were not only victims during this period of oppression but actively worked for reform. Veale (2003) estimates that there were between 13,350 and 17,800 female fighters when the Derg regime ended. Various rebel groups and resistance fighters spread throughout the country during The People's Democratic Republic of Ethiopia (PDRE) (1987)(1988)(1989)(1990)(1991) which was only marginally better than the Derg as it had some of the same leaders (Clapham, 1989). The Tigrayan People's Liberation Front advocated for 'political development and social progress that included gender equality' and involved approximately one-third (40,000) female fighters (Mjaaland, 2013:39 andVeale, 2003). The Ethiopian People's Revolutionary Democratic Front (EPRDF), combined rebel groups and gained power from the PDRE in May 1991. Known as the transition government (1991)(1992)(1993)(1994)(1995), it developed new laws, including a constitution that laid down a democratic federal structure.
This was a turning point for women's rights and the country's gender policies. The Constitution of the Federal Democratic Republic of Ethiopia (1994) clearly states that there is equality between men and women in the social, legal, economic and political realms and suggests affirmative action to ensure women can compete fairly with men. This followed the first women's policy in 1993. In addition, several laws, such as the Family Law (2000) and the Criminal Law (2005) have contributed to the protection of Ethiopian womens rights, and the banning of any type of discrimination against them. The Ethiopian Women's Lawyers Association played a significant role in these changes (Girma, 2013, Burgess, 2013and Cochrane & Birhanu, 2018. The Federal Democratic Republic of Ethiopia remains EPRDF-led in 2018, and despite promises to improve human rights, the government tries to control civil society with dire consequences for women's rights (Yeshanew, 2012).
The result of these government-led paradoxical changes is uneven progress towards gender equality. For instance, Ethiopia lags behind other African nations at a similar level of growth in terms of gender-equality indicators despite a progressive expansion of women's rights in policy and legislation. Although Ethiopia is part of the Africa Union and signatory to many regional and international gender equality commitments, as Sosena and Tsahai (2008) note, GoE has a mixed record when it comes to adopting and implementing regional and international gender commitments. Moreover, several national policies have been inspired by international commitments on gender equality, but up to now, there has not been a full contextualization of these commitments nor even alignment between national (gender) policies (Drucza & Rodriguez, 2018). The domestic policies only weakly incorporate 'gender' choosing instead to focus on women's integration into the development process. According to Biseswar (2011:v) 'the political regime of the Derg scared an entire population to the extent that despite the currently proclaimed "freedom" of the EPRDF ruling party, women remain reluctant to step forward and claim their rights.' White (2011) explains how students are even afraid to write about Ethiopian feminism. This fear is not unwarranted.
This recent disturbing history is in part affected by a nascent domestic women's movement. Women's lack of leadership for gender equality, along with a mis-association with the term 'feminism' and a deliberate distortion of the 'women's question' within Ethiopia are seen as inhibitors to a women's movement building (Biseswar 2008b(Biseswar , 2011). Yet, with a state that suppresses women's rights (and all human rights) when its power is threatened, it will be hard for a women's movement to gain momentum. Burgess (2013) writes about an emerging women's movement until the Charities and Societies Proclamation enabled the government to target and decimate women's civil society organizations and particularly the Ethiopian Women's Lawyer's Association. Consequently, women's contribution to democratic reform remains largely hidden (Burgess, 2013). Even at the grass-roots level, there is an element of control that affects gender relationships, as explained by Berhanu and Poulton (2014) and Tobias and Kjetil (2012). The politicization of agriculture extension even affects the quality of gender advisers at the local level who are appointed to offices based upon political affiliation, rather than technical capacity (Blystad et al., 2014:38).
Nevertheless, the Ethiopian government is shifting. In April 2018, a reform-minded Prime Minister, Dr. Abiy Ahmed, was appointed and has shown strong leadership support for a more inclusive and gender responsive government. The PM appointed a slew of women into various government posts after this study, paving the way for a state entity like EIAR to make a substantial difference to rural women's lives. This new operating space could, potentially, overcome the paradox of the gender politics of the Ethiopian government and underscores the importance of this study.

Approaches to gender
This experimental gender audit method explores the way gender equality work is conceived and practiced in the Ethiopian Institute of Agricultural Research (EIAR) that informs government agricultural policymaking. This section indicates there is substantial room to improve on past efforts to mainstream gender but that there is a willingness to improve gender mainstreaming. Efforts to integrate gender into research projects have been observed across all work areas. These efforts typically include: Counting female participation rates and setting targets; Inviting women to attend training and meetings; Holding single-sex focus group discussions or using participatory methods such as Participatory Rural Appraisal (PRA); Asking division of labor, decision-making, or gender-focused questions in surveys, Seeking women's opinion and Aiming to empower women.
In addition, each institution employs a gender focal person and the gender directorate at the Addis Ababa headquarters employs a senior gender staff member. These are commendable efforts to scale up because EIAR over-relies on household surveys as the primary data collection method.
The collection of sex-disaggregated data is a minimum standard for gender responsive research. Most online respondents do collect some form of sex disaggregated data. However, reporting on sex-disaggregated data is a problem -31% of online respondents never report data in a sex-disaggregated manner. The annual performance reports inconsistently report sex-disaggregated data and under-report genderrelated work and achievements. The documentation review of EIAR's performance report towards GTPI constitutes the bulk of data and figures on EIAR interventions, however, not all data is disaggregated by sex. The report has a gender section indicating efforts made towards gender-responsiveness but this is usually populated by the gender directorate unit. In order to mainstream gender, every unit should have a gender component and each unit would be responsible for populating this section with their progress on gender.
The standard response to mainstreaming gender in EIAR's research is counting the number of female participants. Of the online survey respondents, 45 percent believe that targets for women's participation are frequently or always set, 25 percent occasionally and 26 percent seldom or never set targets. This is a short-term indicator of attendance, not an outcome of active participation. Importantly, the data collected, and its analysis should inform future programs however 74% of online respondents have never done this. It is hard to design a new gender responsive program when women's participation rate is the only gender data collected. Some respondents were also dissatisfied with current efforts, as an FGD respondent explains, we are mainly gender-blind. The only thing that we are trying to really check is whether the participation of important female headed households, as well as the youths, are there.
Most researchers talked about cultural barriers preventing women from attending and learning. Taking additional steps or putting in additional effort to involve women is rare. Comments were heard that it was harder to reach women because enumerators have to travel to different locations and meet women in appropriate places such as their homes (rather than at a Kebele 6 headquarter). This made interviewing women more expensive in terms of time and travel costs, but this was not factored into research plans from the beginning, nor was it deemed a cost-effective use of the budget. An online survey respondent explained; 'the participation of women is almost none as it is costly and laborious.' As one key informant interview (KII) respondent explains, 'When we were doing research on skills gaps with farmers, we tried to include women in the research. The numbers of women participating are very few. If the participants are fifty, the number of women among them will be two or three.' The interviewer asked, 'so when this happens what do you do?' and the respondent replied, 'We do not do much. We let them participate equally and we have an informal discussion to increase their motivation. We discuss it with the women but there is nothing special that we do.' Researchers misguidedly perceive that inviting women to participate is enough. The KIIs and the online survey respondents explained that because achieving government mandated participation rates for females was very difficult, researchers fabricated the participation figures to avoid scrutiny and perceptions of failure. The inability to meet female participation rates is associated with attitude, knowledge and capacity.
One strategy that helps researchers understand cultural and gender norms is a gender analysis. Gender analysis is a crucial aspect of gender-responsive research as it helps to identify many gender-related barriers such as roles and responsibilities, communication issues with women and norms around interviewing and community practices (Drucza et al 2017:1). However, 63% of online respondents had never conducted a gender analysis nor analyzed the gender roles and responsibilities in a targeted community while twelve out of fourteen KII respondents have never conducted a gender assessment before designing research. This means that most projects run on assumptions and gender-blind facts. Therefore, it is not surprising that researchers have troubles reaching women when they do not understand the gender basics of a community or sector. One respondent explained that usually only donor-funded projects require a gender analysis, 'my project is a government-funded project. Since it is time-bound, we work in a hurry. So, we do not have time just to make an analysis for gender. We do not have time to assess. We are asked how many women should be involved and we put some figure, but this figure is not based on assessment.' Funding to perform gender analysis work across the country by crop/livelihood and sector is needed. This would then raise the knowledge base on gender norms and enable the results to be incorporated into new projects. Furthermore, this would make the case for government funding of gender analysis.
Another strategy to reach more women would be through hiring female enumerators. Mobility restrictions and gender norms mean it is inappropriate for a married man to speak to a single woman, and women feel more comfortable talking to other women. Thus, the more female enumerators, the more women would be reached and incorporated into the research process. Only 17% of online respondents aim to recruit a gender balance in enumerators; 22% occasionally, 34% never; and 20% seldom. Female enumerators are essential to reduce gender bias in research and to enable female respondents to feel comfortable answering research questions. More effort is needed to recruit female enumerators and to involve more women in research for development activities.
Gender friendly evaluations and impact assessments are rarely done. This is crucial to ensuring EIAR's work benefits both men and women. Many respondents did not know what a theory of change was and were not involved in developing indicators. One KII respondent questioned, 'How do you work out the different impact on women and men?' This suggests that a culture of reflection and learning is rare.
Despite what might be described as a weakness in the approach taken to implementing gender mainstreaming, most online survey respondents feel personally responsible for incorporating gender in their work but ten percent need awareness raising. Seventy-three percent of online survey respondents agree that the promotion of gender equality fits into the image of the institute, (48 percent agree, and 25 percent strongly agree). This presents a tremendous opportunity to push gender equality at EIAR. Moreover, the majority of the online respondents felt that EIAR could do more than it is currently doing to institutionalize gender equality, which is a sobering call to action.
Some respondents felt lost about what they should be doing in terms of gender (outside of reporting numbers of female participants). For example, one respondent said, 'The center asks us [about gender] in the report phase and in the preparation of our action plan. Every year we categorize the number of women who will benefit from the training and technology demonstration. So, based on that we try to report that way but there is no strong way or balance about how much you plan, or how much you should achieve. But we try.' Other responses confirm that gender is not done properly and that there is a lack of support and guidance on how gender should be incorporated, 'First we do not even think about it. The reason we do not include gender in our work is because we do not know how to incorporate gender in our job. We are not opposing it, we just do not know how to incorporate it.' While basic research proposal guidelines were found during the documentation analysis, their gender-focused sections were insufficient to guide research teams on how to design, budget and conduct gender responsive research. Each Center has a gender focal person but they are internally recruited and usually work part-time on gender issues. Given that EIAR recruits scientists (and not gender experts), there is no one internally who could help the institution substantially on gender issues.
The answers to the question, 'What changes would you like to see in EIAR in terms of gender?' resulted in the following responses: recruit more women; provide special support for pregnant women; provide more opportunities to work at higher levels; recruit a younger workforce with more liberal attitudes to gender; work with 'extreme religious groups' who do not accept women as equal; 'it needs to be properly mainstreamed because gender focal people are not enough;' and work with the issue at the woreda and community level. This fairly good list of options demonstrates that people think about 'doing gender'.
This section shows that gender mainstreaming has begun, and that people generally see the need to mainstream gender. The majority of staff feels that the institution should do more to improve gender responsiveness across the organization and are dissatisfied with the way gender is currently conceived and supported. The responses reveal an emerging understanding of what gender equality should mean in Ethiopia.

Resources for gender
Funding is needed to address any of the above issues. Many respondents raised the lack of resources for gender as a barrier. Only two out of fourteen KII respondents had any budget earmarked for gender within their research programs. Some had never heard of the idea of reserving budget for gender activities, 'There is no budget allocation for gender. There might be things included when projects are designed but there is no serious allocation like other areas.' Others incorrectly assume that if research benefits all then it should count towards a gender budget. In FGD a participant mentioned not being sure if gender required a separate budget, 'There is no budget for gender and during planning, people try to consider gender but rarely budget for it. But is the lack of budget a problem? I don't know because if gender was mainstreamed would it need a separate budget?' During the EIAR gender audit presentation, this common misunderstanding was raised. The audience response indicated a lack of capacity for gender mainstreaming.
Gender must have a separate budget to reach women, address their unique needs, deal with power inequities and build awareness and capacity for gender equality. The lack of gender budgeting makes it difficult to account for spending on gender equality as well as to re-allocate funds within budgets for gender-focused activities. Gender under-resourcing signals the low priority that is given to gender work within EIAR. Government ministries are mandated to report on gender budgeting to the Ministry of Finance, including the MoANR, yet the EIAR has not been asked by the MoANR to report on gender budgeting. This would be a necessary starting place for EIAR to improve its gender responsive data.

Knowledge management/communications (KMC)
The previous section explained the lack of resource allocation toward gender-focused research. The knowledge management/communications (KMC) section organizes, designs and records publications. Outside of academic publications, they are responsible for EIAR's outputs. This section is split into two parts: the documentation analysis and the bibliometric analysis results.

Documentation analysis
Existing publications need to improve to be gender responsive. The documents reviewed from the communications unit (brochures, newsletters, magazines) mostly had pictures of men in them. Other publications only show women in stereotypical roles (e.g. holding a baby or churning milk) and thus do a disservice to gender equality. The low representation of women in glossy publications perpetuates the stereotype that women do not farm, do not work at EIAR and are not scientists.
Most of the documents reviewed (Table 1) lack a gender perspective and either use gender-neutral language or very gender stereotypical and/or exploitative terminology. For example, in a research report on 'small scale farmer-based hybrid maize seed multiplications', a sentence in the conclusion section states that ' … since the farmer uses his family labor and makes frequent follow-up of the day to day activities, quality seed can be produced with a relatively low cost.' Encouraging labor exploitation of family members breeches (inter)national human rights conventions signed by the Ethiopian government, including the Convention on the Elimination of all Forms of Discrimination against Women (CEDAW) and the Convention on the Rights of the Child (CRC). 7 This kind of language should not appear in a government report. Table 2 reveals that most documents reviewed scored zero on their gender sensitivity out of a possible score of five (Gender is mainstreamed throughout the document consistently).

Bibliometric analysis
Analyzing data by gender requires an even higher level of capacity than collecting it because people must understand gender to analyze gender differences and draw conclusions about what the data means. The bibliometric analysis 8 reveals a shortage of gender publications. While some projects refer to gender as a cross-cutting activity, it is rare to have gender appear in publications or comprehensively in reports. Of the global production of EIAR affiliated Scopus listed publications (475), 16 were found to have the term 'gender' in the article's title, in the keywords, in the abstract or even in the title of cited articles (references). If these results are limited to documents from  A two-part quality review assessed the gender details of the 17 articles. Table 3 presents results from the first quality review on a range from zero (No mention of gender or sex) to 100% (Gender and/or women are the primary focus of the entire research process, from design to analysis). None of the 17 articles focused primarily on gender or women and only one article asked a gender-focused question, collected data from men and women throughout AND completed a gender analysis (Nebiyu, Beyene, Giorgis, Kassa, & Kass, 2012), two explicitly analyzed sex-disaggregated data and its meaning from a gender perspective (Kolech et al., 2015 andSeyoum, Teketay, Dugo, &Wodafirash, 2015), eight had sex-disaggregated data but without gender analysis and six should have been excluded from the bibliometric analysis because the word women and/or gender was only found in the abstract or in references and not in the body of the article. This means EIAR had three gender related articles from 2011 to 2015. Only nine included sex-disaggregated data out of 48 Scopus listed publications.
The second part of the quality review assessed the conclusion and its relevance for policy or practice on a scale of 0-4 (0 ¼ no mention of gender in the conclusion; 4 ¼ highly relevant gender aware conclusion). The highest score was two, meaning two articles had sex but not gender conclusions (Nebiuyu et al., 2012;Seyoum et al., 2015). The rest of the articles did not discuss gender nor what the results meant for women compared to men. Will gender mainstreaming be the right strategy to increase gender-focused publications?
The lack of access to gender journal articles is part of the challenge facing EIAR researchers who wish to publish on gender. Most online survey respondents have limited access to gender publications. This is significant because most reputable journals require a literature review section. The lack of access to gender publications inhibits EIAR from publishing gender results. It also reduces the contribution EIAR researchers can make to the gender in agriculture literature.

Conclusion
The focus of this article is on the research aspect of EIAR and the way agriculture researchers understand and practice gender. The EIAR has made some attempts, however rudimentary, to integrate gender sensitivity into its various projects including data disaggregation of female participants, outreach to women for training and meetings, targeted focus groups and addressing the needs of women through surveys. However, the EIAR has yet to fulfill its mandate of gender mainstreaming. The reasons include institutional constraints, gender biases, lack of understanding of gender concepts and gendered analysis among researchers, lack of resources and the absence of guidance from oversight managers and agencies such as MoANR. Nevertheless, the results offer hopeful insights. An overwhelming percentage of online survey participants agree that the promotion of gender equality fits into the image of the institute and EIAR has a gender focal person in every Center. This presents a tremendous opportunity for improving gender equality at EIAR. Moreover, a new Director General, who is a gender advocate, began in EIAR in 2018. Recommendations from the audit have already begun to be implemented, and henceforth the value of undertaking gender audits is that they can provide a clear direction for change.
A number of theorists are skeptical about gender mainstreaming's potential. It can dilute the power and politics associated with gender inequality and results in gender being a checklist or a burdensome reporting requirement. This is highly relevant to a research institution like EIAR where gender needs to be seen as an experimentor a topic for scientific enquiryif a change is to occur. The current gender mainstreaming approach has resulted in counting numbers of women, partly because people do not know what else to do. As researchers, EIAR staff understand that to attempt gender mainstreaming requires a lot of effort and resources, something EIAR lacks as a government-funded institution in a lowincome country. Rather than adopting a blanket gender mainstreaming approach, this paper advises focusing on a few key areas of reform. One of these must be building gender research capacity.
More technical gender research support is needed because mainstreaming gender is unlikely to improve the quality of gender research or the number of genderfocused publications in the short term. As previously noted, one of the serious limitations of gender mainstreaming is its loss of potency to diagnose power as it dislodges from feminist theory and implants in structures of institutional bureaucracies. Hence, EIAR researchers should be trained on methods that integrate gender analysis in all aspects of their research, and their awareness about feminist research methods/methodologies should be raised. Moreover, EIAR should use various categories of analysisnot just household headto really understand the relevance of gender to the agriculture sector.
As African feminist scholars highlight, narrowing the gap between researchers and researched can be done via participatory research approaches and other feminist methodologies more successfully than focusing on gender mainstreaming. Perhaps the emphasis needs to be placed on creating a learning organization that experiments with different research methods and engages in regular self-reflexivity about power, gender, intersectionality and the appropriateness of research methods. More data on the various ways women contribute to agriculture and how they miss out on opportunities to learn new techniques to increase the productivity in Ethiopia's most important economic sector may help to reduce the mis-association of women's rights with threats to the state. This would contribute to escaping the paradox that surrounds the history of gender equality in Ethiopia while contributing to sound policy debate. If EIAR is serious about the capacity for integrating gender as an analytical lens in research design and implementation, it can make a substantial impact on the country's food security and gender equality. Notes 1. The word 'development institution' is used to mean an organization that operates in low income countries and manages donor funded projects. 2. Adapted from Interaction's gender audit manual: https://www.interaction.org/sites/default/ files/Gender%20Audit%20Handbook%202010%20Copy.pdf 3. Interaction, ILO and SNV. 4. Only three FGD were held as it was difficult to obtain commitment from senior managers. 5. N.B. the Ethiopian government recognizes this and since 2015 has employed more women as extension agents. 6. Amharic for village. 7. See: https://www.abyssinialaw.com/blog-posts/item/1483-ethiopia-s-human-rights-treatyreporting-to-the-un-treaty-bodies 8. A bibliometric analysis is the attempt to quantitatively assess the academic quality of journals or authors by statistical methods such as citation rates. 9. Some of the reviewers felt that a couple of documents should be given a half point.