D1. Protein diet, chay and price data

What this appendix contains. The micro data behind Protein diet and Chay baseline, generated from the data files named in each section. Different sources measure different things (food supply, individual intake, household acquisition for home use), so we never average them.

#D1.1 Which source measures what

#D1.2 Protein supply, intake and household use

Selected rows from demand_protein_intake.csv (460 rows in all).

Household use for home consumption (VHLSS), per person per month.

By region, 2024 (pre-2025 statistical regions).

Projections (OECD-FAO baseline) and USDA consumption.

#D1.3 Retail price per 100 g of protein, 24 September 2026

Formula: VND per 100 g of protein = 100 x price per kg / (10 x protein g per 100 g x edible share). Protein values from the Vietnamese Food Composition Table unless stated. USD at 26,000 VND.

#D1.4 Food spending and eating out

#D1.5 Demand drivers: population, income, health, prices and food safety

#D1.6 The protein ladder: Vietnam and its neighbours

Animal share of protein supply, meat supply and income. Selected years from demand_protein_ladder.csv (1961 to 2023). Transfer assumption: neighbours show where diets have levelled off, not where Vietnam must go.

#D1.7 Chay practice: every survey and estimate we found

None of the surveys below is a probability sample with a diet question; online panels count tofu and soy milk as plant-based, so they measure familiarity with the incumbent. The one probability survey that asks about diet, the Diet Quality Questionnaire (2021), does not ask about chay but caps it; it is set out in D1.9.

#D1.8 Chay market: volumes, spikes, restaurants and prices

#D1.9 National diet surveillance: the Diet Quality Questionnaire

The Global Diet Quality Project (GAIN, Harvard and Gallup) asked a nationally representative sample of 1,007 Vietnamese adults aged 15 and over, by mobile telephone in Vietnamese, whether they ate each of 29 food groups on the previous day. Fieldwork ran from 13 November to 12 December 2021, an ordinary month holding two principal chay days; the design effect is 2.56 and the maximum margin of error 4.9 points. The country web page gives the fieldwork a month later; the data file and Gallup's methods table agree on November to December, and we use those ^9,10 VN-directCao bộc lộ. The Vietnamese pulse item names đậu khô, đậu nành, đậu phụ, sữa đậu nành hoặc bột đậu (dried beans, soybeans, tofu, soy milk or bean flour), so it measures the daily reach of the incumbent soy foods; the processed-meat item leaves out giò, chả and nem chua (pork rolls, loaves and fermented pork), so it is a floor ^61,98 VN-directCao. Full rows: diet_quality_vietnam.csv.

Sample: 547 men, 460 women, 768 urban, 238 rural (weighted estimates); the rural subgroup is small, and subgroup differences sit within overlapping intervals except dairy by sex. The public results give no age, income or education split, and we found no downloadable microdata ^10,99 VN-directCao.

Flesh-free shares are our calculation; food lists differ by country, so comparisons are indicative.

What it tests (our calculation) VN-directTrung bình suy luận:

• The central chay assumption is supported as a ceiling only. A chay day is flesh-free, so chay-day person-days cannot exceed 7.9% in a month without festivals. Part III's 2.5% (30% keeping 2.5 days a month) uses about a third of the ceiling. If Vietnam's other flesh-free days were at Cambodia's or the Philippines' level (3.3 to 5.8%), chay-type days would be 2.1 to 4.6% of person-days (illustrative, Low), which brackets 2.5%.
• The top of the earlier range is contradicted in practice. 6.7% would need 85% of all flesh-free days to be chay days and exceeds the lower confidence bound (5.3%); about 5% is a practical upper bound. The sensitivity range in Demand sizing (1.7 to 4.2%) sits inside the ceiling.
• "About 10% of Vietnamese are vegetarian" is contradicted. Ten per cent vegetarians plus even the low chay case would exceed the 11.0% upper bound of flesh-free days.
• Processed meat cannot be tested. The item omits giò and chả and counts days, not grams.
• Legume supply and the chay-day protein assumption are consistent. FAO balance sheets give soybeans and other pulses 6.94 g of protein per person a day in 2021; spread over the 40% of adults who eat legumes on a given day, that is about 17 g per eating day (supply, not intake), the same order as the 12 g a chay day is assumed to carry ^1,9.
• A cheap next step. Gallup holds the interview dates, so the flesh-free share on recall days that fell on the lunar 1st or 15th (about 65 of them) can be compared with other days; a doubling would be detectable ^10,83.

#D1.10 What children's school plates carry

The data behind the correction in Protein diet section 11.2: for school-age children, including upland boarders, protein and zinc are met at lunch and dinner, and the documented shortfalls are breakfast, calcium and vitamin A. Rows are selected from upland_school_weeks.csv, protein_quality.csv, protein_food_micronutrients.csv and nutrition_trials.csv; school channels are in Channels.

Reading (our inference) VN-directTrung bình suy luận: no coded upland lunch was without meat, fish or egg, so a zinc-fortified soy dish sold to these kitchens would stretch pork, as tofu already does, rather than fill a gap. The documented gaps (breakfast, calcium, vitamin A) and stunting before age two point to other instruments. A fortified plant item for boarding breakfasts or charity-funded lunches is a research option to test on calcium and vitamin A as well as zinc.

Related: Protein diet, Chay baseline, D3. Retail demand signals.