Start with the interactive report: Can a food-demand elasticity carry the weight we put on it? — a single-page survey of the evidence for this workshop, with a parameter dashboard and a Hypothes.is annotation layer.
Background note: that report is the fullest synthesis for the workshop; a shorter first-pass Claude summary of evidence on PBA penetration and taste-comparability is also available. Both are exploratory and under review rather than vetted literature reviews.
The question
Animal welfare funders face a strategic question: should they invest in reducing the cost and improving the quality of plant-based alternatives (PBAs), or focus on other interventions like corporate animal welfare campaigns? The answer depends heavily on a factual question that's surprisingly hard to answer: how much do plant-based products actually replace animal products? This is the focus of The Unjournal's Plant-Based Substitution Pivotal Question.
If a 10% price reduction in Impossible Burgers leads many people to eat fewer chickens, that suggests a case for funding PBA development. If people who buy plant-based burgers are mostly substituting away from beef (not chicken), or if PBA buyers were already eating less meat, the case weakens considerably. The path from PBA improvements to animal welfare impact runs through these substitution effects.
This workshop takes a broader view than PBA alone. We're also interested in substitution across the full meat-and-alternatives landscape: between different animal products (e.g., chicken vs. beef vs. eggs), between meat and dairy, and between conventional and emerging products like cultivated meat. We also revisit an underlying question: do any existing measurement methods reliably answer these questions at all? Scanner data, choice experiments, and field experiments seem to have serious limitations, and existing estimates are "all over the map."
One finding worth engaging (though contested): The Neuhofer & Lusk (2022) study title puts it directly: "Most plant-based meat alternative buyers also buy meat." Evidence across several studies suggests the share of PBA purchasers who are strictly vegetarian or vegan is quite small, though specific estimates vary considerably and this remains an open empirical question. If the majority of buyers are omnivores who also consume conventional meat, lower PBA prices could displace more animal consumption than a "vegans buying PBA" story would suggest — but the magnitude is genuinely uncertain.
Why this is hard to measure
Estimating substitution effects is notoriously difficult. Each methodological approach faces serious challenges of its own — traditional observational demand estimation, field experiments, and hypothetical/survey methods all have distinct limitations. Click any challenge to expand.
Price endogeneity — the core challenge for observational demand estimation
Stores don't vary prices randomly. They adjust prices based on expected demand — for instance when they expect demand to be more price-sensitive, or based on pro- or counter-cyclical pricing strategies. This distorts observed price-quantity relationships and makes standard regressions unreliable.
Instrumental variable approaches (Hausman instruments, BLP-style cost shifters) are designed to address this, but the instruments may not satisfy the exclusion restriction — especially in thin PBA markets where wholesale prices and retail promotional decisions are jointly determined.
Data coverage gaps — what scanner datasets miss
Major scanner datasets (Nielsen, IRI) cover grocery store purchases but not restaurants, which account for roughly half of food consumption. If lower PBA grocery prices shift some consumption to restaurants, this cross-channel effect is invisible to scanner data.
Coverage of PBA products in earlier waves of these datasets was also limited.
Short-run vs. long-run effects — current elasticities may miss behavioral change
Cross-price elasticities measure immediate price-driven substitution. But the most important effects of lower PBA prices may operate through longer-run habit formation, identity change ("reducetarian" adoption), or market development. Short-run elasticities from current datasets may substantially understate long-run impacts.
Functional form and heterogeneity — elasticities are not constant
Demand estimation requires strong assumptions about functional forms (AIDS, QAIDS, BLP). Different specifications yield very different estimates. The variation in estimates across studies partly reflects this non-constancy, not just data differences or identification choices.
Field experiments — avoid endogeneity, but have other limitations
Retail price randomization and placement experiments (like Bray et al. 2024) avoid the endogeneity problem. This is a major advantage. But they face their own challenges: short windows, limited contexts, and few PBA-specific trials mean extrapolating to a market-wide supply-side price change is uncertain.
Cafeteria and institutional experiments (expanding vegetarian options) face similar scope limits but are informative about within-meal substitution in captive settings.
Hypothetical and survey-based methods — stated preferences may not reflect real behavior
Discrete choice experiments (DCEs) ask consumers what they would buy at different prices. Useful for scenarios not yet observed in data — but stated choices systematically overstate real switching behavior, and results across well-designed studies conflict considerably.
Inconsistent results across all methods — estimates are "all over the map"
Across the four US-based scanner studies alone, there is no single meat category with consistent findings. Some studies find PBAs are complements to certain meats rather than substitutes — a counterintuitive result that may reflect measurement artifacts, though other explanations exist.
Bray et al. (2024) document a striking gap: randomized experimental estimates of own-price elasticities were much smaller than IV and difference-in-difference estimates on comparable observational data. It seems unlikely cross-price elasticities would be more consistent or robust given the same data and identification challenges apply. This raises questions about which to trust — and whether either approach identifies the quantity we actually care about.
The workshop will directly confront which methods, if any, yield useful insight — and what further research would most reduce uncertainty.
The evidence landscape
We've identified approximately 15 relevant empirical papers, spanning several methodological approaches:
Observational demand estimation
Using scanner/retail data (Nielsen, IRI) and macro time-series to estimate demand systems and cross-price elasticities. Methods include AIDS/QAIDS and structural discrete choice models; identification strategies vary considerably in rigor.
Hypothetical choice experiments
Survey-based discrete choice experiments asking consumers to choose between product bundles at different prices. Useful for scenarios not yet observed in data, but stated choices may not reflect real purchasing behavior.
Event studies
Examining what happens to meat purchases when households first buy PBAs. Descriptively useful, but causal interpretation requires strong assumptions about why and when households make their first PBA purchase.
Field experiments
Randomized interventions including retail placement experiments (e.g., moving PBAs to the meat aisle) and price randomization trials. Results are informative but specific to context; few randomized price experiments have directly involved PBA products, and the Bray et al. (2024) pricing trial found much smaller effects than observational IV estimates.
What we want to achieve
Our scope: We look at substitution across the full product landscape — not just PBA and chicken, but including dairy, other animal product categories, and the anticipated arrival of cultivated meat — since these are all relevant to the ultimate animal welfare question.
This workshop aims to bring together agricultural economists, industrial organization and quantitative demand estimation economists, quantitative marketing researchers, economists interested in animal welfare, and funders, industry, and/or charity practitioners to:
- Review the evidence: Which studies are most credible? What can we reasonably say about substitution effects and with what confidence — and do currently available methods and data even yield useful insight?
- Confront the measurement problem: Do scanner data, choice experiments, and field experiments give reliable answers? The working hypothesis going in: "no demonstrably-reliable method exists: prove me wrong."
- Elicit and aggregate beliefs: Use structured belief elicitation — before, during, and after reviewing evidence — to quantify expert uncertainty, surface disagreements, and update collectively. What's a reasonable range for the key elasticities? Where do experts disagree and why?
- Connect to decisions: Given current evidence, is PBA funding plausibly competitive with corporate campaigns? Would meat taxes improve or worsen animal welfare? Will innovative products like PBA and cultured meat substitute for farmed animal consumption, or will they mainly be taken up by existing vegetarians and vegans?
- Identify research gaps: What study would most reduce uncertainty? Is it feasible and fundable?
Preliminary agenda
Note: This agenda is preliminary and will be refined based on participant input. Target date is July 2026 — scheduling is open.
The workshop is planned for approximately 3–3.5 hours online, with drop-in segments — participants are welcome to join only the parts most relevant to them:
- Framing & Consumer Composition (~15 min) — Why animal welfare advocates and funders care about substitution; what a preliminary literature review suggests about who actually buys PBA products; what this means for the question
- Evidence Landscape (~25 min) — Overview of scanner data studies, choice experiments, and event studies; what methods can and can't tell us
- Methods Debate (~30 min) — Structured exchange between demand-estimation approaches and experimental/survey approaches; which approaches are more reliable for the practical questions?
- Focal Question Discussion (~25 min) — If PBA prices fell 10%, what would happen to chicken consumption? Expert beliefs elicited in advance; a structured live discussion of where expert views diverge and why
- Pivotal Questions & Beliefs Elicitation (~20 min) — Walking through the operationalized PBA pivotal questions; participants revisit the beliefs they submitted beforehand and update them afterward
- ROI and Research Gaps (~20 min) — Is PBA funding competitive with corporate campaigns given current evidence? What research would most reduce uncertainty?
Pivotal Questions
We've operationalized the key uncertainties into specific, answerable questions. These include:
- PBA_01 (Focal): If the price of high-quality plant-based hamburgers fell by 10%, how would global chicken consumption change?
- PBA_02: Is PBA research a good investment from an animal welfare perspective?
- PBA_03: Does the AW benefit of $100K to GFI exceed $100K to THL corporate campaigns?
Several of these are posted on Metaculus for community forecasting.