By Dr. Nguyen Van Bao
Picture this. You are running a household survey in a shrimp farming village in the Mekong Delta. Your sample looks balanced, roughly equal numbers of men and women. The interviews go smoothly. You pack up and feel good about the data.
Then you sit down with the numbers and notice something disconcerting. Almost every woman you interviewed gave her husband’s answers. She described his decisions, his plans, his preferences. Yet, she was the one who fed the shrimp twice a day, monitored the water, handled local sales, and kept track of the costs. She just did not seem to think any of her work counted as “deciding”.
That gap, between what the data recorded and what was actually happening on the farm, is where Gender Equality, Disability, and Social Inclusion (GEDSI) work in Vietnamese aquaculture really starts. It is not just a question of ticking off a checklist. What is needed is a mirror which asks: whose reality does the data actually reflect?
This question matters more than it might seem. Vietnam’s aquaculture sector runs substantially on women’s labor. Women account for over 60% of the workforce in fish trades and services, and for more than 80% of those employed in seafood processing (Kha 2020). If the data does not capture what they actually do, then no training program, credit scheme, or climate adaptation plan is likely to fully reflect the realities on the ground.
Why This Happens: The Survey Was Not Designed To Find It
The data gap does not happen by accident. It comes from specific choices made about how surveys are designed, when they are conducted, and whose voice is treated as the main source of information.
The most basic problem is the decision-making question. Most household surveys in Vietnam ask a single question: who makes the major decisions in the household? In communities where gender norms are strong, the answer is almost always the husband, not because he makes every decision, but because that is the socially expected answer. The survey does not ask who decided which fingerling supplier to use this week, who checked the water at 5am, or who negotiated the price with the trader at harvest. It records the data, and the data reflects expectation more than reality.
The interview setting makes this worse. In many rural Vietnamese communities, women are unlikely to speak openly when their husbands are in the room, which is exactly when most household surveys are conducted. The woman says what seems appropriate given who is listening. The enumerator writes it down. What gets recorded reflects the social situation of the interview, not what the woman actually thinks or does.
Scheduling creates another layer of distortion. Women’s caregiving responsibilities mean they are often unavailable during standard working hours. Surveys conducted at the wrong time systematically miss the women with the heaviest workloads, often the exact women most directly involved in running the farm day to day.
And there is a deeper problem: the survey was not designed to ask about certain things at all. It does not ask about the ecological knowledge that comes from years of watching water color shift before a disease outbreak. It does not ask about the informal credit networks that keep a small operation running through a bad season. These are things women in Vietnamese aquaculture communities often know best. They do not appear in most datasets simply because nobody thought to include questions about them.
What To Do Differently: Redesign From the Start
The fix is not to add more women to the sample. It is to redesign what the survey instrument is trying to find out.
The most important change is to break apart the single “who decides” question. Instead of asking who makes the major decisions in the household, surveys should ask domain-by-domain: who selects the fingerlings, who decides on harvest timing, who controls how income is spent, who reaches out to extension services. In Vietnamese aquaculture households, women often manage day-to-day operations while men control capital investment decisions. Both matter for program design. But only one tends to appear in the data. With responses from both spouses, each speaking to their own domain, tools like the Women’s Empowerment in Agriculture Index (Alkire 2013) can reveal this split clearly.
The interview setting also needs to change. Conducting interviews with women separately, ideally with female enumerators, consistently produces different data. Gender-disaggregated focus groups, rather than mixed household discussions, are more likely to provide honest answers in communities where women do not typically speak openly in front of their husbands (Alkire 2013). This is not a small operational detail. It is a design choice that changes what the data actually shows.
Fieldwork scheduling matters as well. Finding out when women are actually available, accounting for caregiving, seasonal work patterns, and community norms, and then designing fieldwork around women’s preferred times rather than around standard office hours, will substantially change who ends up in the sample. Finally, capturing informal and unpaid labor requires different tools. Time-use studies and activity diaries can expose contributions that never appear in employment statistics.
But gender alone is not enough. Women in Mekong Delta aquaculture communities are not a homogeneous group. Age, ethnicity, land tenure, distance to markets, and access to extension services all shape what life looks like for a given woman on a given farm. Disaggregating by gender but ignoring these other factors will miss the most vulnerable community members entirely (OECD 2025).
Why It Matters: Bad Data Leads to Bad Programs
Better data collection is not just a technical nicety. The data gap has real consequences for what gets funded and what gets built.
Consider training programs. If household surveys consistently identify men as the primary decision-makers in aquaculture operations, training sessions will be organized for men, scheduled at times that do not account for women’s caregiving, held in places where women may not feel comfortable, and focused on capital investment rather than on operational management. The program then underperforms because it missed the people who are actually running the farms day to day.
The same problem applies to financial access. Women in Vietnamese aquaculture often depend on their husbands for investment capital, and limited bargaining power constrains their autonomy within the household (OECD 2025). Land titles are frequently held in men’s names, which blocks women from accessing formal credit even when they are the primary farm managers. None of these barriers show up in datasets that treat the household as a unit with one economic actor.
Climate adaptation programs face the same blind spot. Women’s preference for lower-risk, more stable, farming models reflects a different set of priorities and coping strategies compared to men’s. Without accounting for these differences, climate adaptation programs may end up funding approaches that increase the very volatility women are trying to manage. They may also exclude from planning the people who often hold the most detailed knowledge of local ecological conditions.
What Policy Needs To Do
Better survey design is the starting point, but it is not the whole answer. The structural conditions that make women’s contributions invisible in data are the same ones that make them invisible in policy. Addressing the data gap without addressing those conditions will produce better-informed programs that still do not reach the people they need to reach.
The most straightforward step is to require sex-disaggregated data in national statistics. Vietnam’s agricultural census and annual sector surveys do not currently break down employment and production data by sex at the household level. Making data that is disaggregated by sex, age, and role a standard requirement would create the baseline that sector-level gender analysis currently lacks. This is a policy decision, not a technical one.
Extension services need to be redesigned. Most technical training programs in Vietnam are still organized for male household heads, at times and in formats that assume male availability. Evidence from Vietnam and Thailand suggests that women’s inclusion in aquaculture training and household decision-making can strengthen gender relations and improve the effectiveness of community-based aquaculture management (IDRC 2025). Targeting women who actually manage daily farm operations, and scheduling around when they are available, would improve program effectiveness without adding significant cost.
Land and credit access requires legislative action. Certificates of land use rights that list both spouses, financial products that do not require male co-signatories, and credit programs designed around collateral women can actually offer would address barriers that no survey redesign can fix on its own.
Finally, gender-responsive research needs dedicated funding. Gender-disaggregated data collection costs more than standard surveys. Without explicit budget lines for this work in research grants and program investments, the gap will keep reproducing itself by default.
A Small Change in How We Ask, a Big Change in What We Find
The methodological changes described in this article are not complicated. Ask different questions. Run separate interviews. Schedule around women’s availability. Pilot instruments with women first. None of this requires a fundamental overhaul of research infrastructure.
But taken together, these changes produce substantially different data. And different data leads to different programs, different financial products, different climate adaptation plans, ones that actually reach the people who sustain the farms.
Vietnam’s aquaculture sector exports over USD 10 billion worth of seafood annually, an achievement based in significant part on contributions that do not appear in the country’s own data. The woman who monitors the water at 5am and negotiates with the trader at harvest does not disappear just because the survey did not ask about her. She keeps the operation running. She just does not show up in the analysis.
Fixing that starts with a simple question: what would the survey look like if it was actually designed to find what is there?
AQUADAPT 