Octopus full-cost account · supporting analysis

The animal-welfare cost of farmed mussels, benchmarked against octopus

Applying one welfare-valuation method to both species. For octopus the welfare cost dwarfs the sale price at every setting. For mussels, under the evidence-supported assumptions, it stays below price.

Working note, 9 July 2026. Companion to the octopus full-cost account (Compassion in World Farming). Method: Animal Quality-Adjusted Life Year (AQALY). Every mussel figure is a conditional construction, not a measured cost: see the caveats.

The octopus account asks a single question: does the welfare cost of farming an animal exceed the private sale value of its meat? For farmed octopus the answer is yes by a wide margin, at every plausible setting. This note runs the identical method on mussels so the comparison is like for like, and asks whether the welfare cost is still a multiple of price or whether it is manageable.

Headline. Under the neural-weight proxy, the welfare cost of farmed mussels is about EUR 0.65 per kg against a first-sale price near EUR 1.00 per kg, so about two-thirds of price, and near zero once the low probability that a mussel is sentient is applied (about 2% of price). The same method gives farmed octopus a welfare cost of 12 to 49 times price under the neuron proxy and 430 to 1,800 times price under its species-specific moral weight.

Calibrated answer. For mussels the welfare cost is manageable, not a multiple of price, under the assumptions the evidence supports. It reaches octopus-like multiples only under a contested moral weight that most assessors would reject for a brainless, sessile filter feeder. Read every mussel figure as a conditional upper bound.

Figures

0.01× 0.1× 1× price 10× 100× 1,000× 10,000× below price above price → Octopus, species weight 430–1,800× Octopus, neuron proxy 12–49× Mussel, neuron proxy 0.65× central (0.1–1.1×) mussel, sentience-adjusted ≈ 2% of price
Figure 1. Welfare cost as a multiple of first-sale price (log scale). Bars show the plausible range; the dot is the central estimate. The octopus cost exceeds price at every setting (the most conservative corner of all, least-bad welfare with a fifth-percentile moral weight, is still about 8 times price). The mussel cost straddles price under a precautionary reading and falls to roughly 2% of price once the probability of sentience is applied. octopus   mussel.
-1.5 -1.0 -0.5 0 = death-equivalent +0.5 ← worse than death life worth living → Octopus farmed life -0.42 to -1.71/yr Mussel farmed life +0.36 to -0.72/yr
Figure 2. Welfare score of the farmed life (AQALY per year). One is the best attainable life, zero is death-equivalent, negative is a life worse than death. The octopus range lies entirely below zero, so a farmed octopus life is scored as worse than not living. The mussel range crosses zero, so under standard commercial conditions the life is close to neutral and its sign is uncertain; the central grading is just below zero. Dots mark the central case.
10⁻⁷ 10⁻⁶ 10⁻⁵ 10⁻⁴ 10⁻³ 10⁻² 10⁻¹ 1 (human) Mussel neuron proxy 9×10⁻⁷ silkworm 0.002 octopus, neuron 0.006 Octopus species weight 0.213 carp 0.09 chicken 0.33, pig 0.52 about 244,000× gap
Figure 3. Moral weight (welfare range, human = 1; log scale). Rethink Priorities scores eleven farmed taxa; no bivalve is among them. The mussel neuron proxy (about 9×10⁻⁷) is roughly 244,000 times below the octopus species estimate and about 6,600 times below even the octopus neuron proxy. The silkworm, the least-weighty animal Rethink Priorities has scored, is still about 2,000 times heavier than the mussel proxy.

Notes

The method and its equations

The welfare cost per animal is the product of three terms: the lifetime welfare deficit in the animal's own units, a moral weight that converts those units to a human-equivalent scale, and a monetary value per human quality-adjusted life year. The welfare deficit is scored with the Animal Quality-Adjusted Life Year framework (Espinosa, Browning and Treich 2025; monetisation from Budolfson et al. 2024), which grades four welfare domains on a nine-point scale and aggregates them to a welfare score per year.

The score for a welfare state, written as four grades for Nutrition, Environment, Health and Behaviour (each 1 for the best attainable to 9 for the most severe negative, 5 neutral), is:

S = 1 + βNutrition + βEnvironment + βHealth + βBehaviour   // β is the grade coefficient from the table below; S = 1 best life, 0 death-equivalent, negative worse than death

The monetary welfare cost of one farmed animal over its life, and per kilogram of product:

cost per animal (EUR) = |S| × L × w × V cost per kg (EUR) = cost per animal × n L = life in years    w = moral weight (human = 1) V = EUR 147,000 per human quality-adjusted life year n = animals per kg of product

Where no species moral weight exists, it is set by a neural-weight proxy, and the sentience-adjusted and break-even variants are:

w = neurons(species) / 86,000,000,000 // human neuron count sentience-adjusted cost = cost × P(sentience) break-even moral weight w* = price / ( |S| × L × V × n ) // w above w* means cost exceeds price

The value of EUR 147,000 per quality-adjusted life year is the lower-bound figure the framework itself adopts, so the resulting costs are conservative. The engine reproduces the published octopus benchmark to the digit before any mussel input is entered: the framework's own caged-broiler worked example (state 7/9/7/9) computes to a score of -1.157 per year.

AQALY domain coefficients (merged sample, the authors' adopted estimates)

Coefficient β added to the score at each grade. Grade 1 contributes 0; even grades (2, 4, 6, 8) are the linear interpolation of their neighbours, since the underlying survey estimated only grades 3, 5, 7 and 9.
DomainGrade 3 (mild +)Grade 5 (neutral)Grade 7 (mild −)Grade 9 (severe −)
Nutrition-0.1025-0.1502-0.2399-0.7148
Environment-0.1627-0.2179-0.3422-0.7392
Health-0.0454-0.1440-0.2976-0.8502
Behaviour-0.1148-0.2529-0.3697-0.8806

Welfare scoring: how each species is graded and how the grades become a score

Each grade maps to a coefficient in the table, and the four coefficients sum with the base value of 1 to give the score. A farmed octopus is a solitary predator held in communal tanks with cannibalism, no dens and no validated humane slaughter, so it grades as severely negative on environment, health and behaviour, and its score lies well below death-equivalent. A farmed mussel grows attached by byssal threads to ropes or poles in open seawater, filter-feeding in place for roughly 12 to 24 months; it is sessile by nature, so confinement frustrates little of its repertoire and no feed is withheld. The identifiable harms are periodic handling with air exposure and a kill that uses no stunning (boiling, steaming, freezing or live sale), which is why the health domain carries the most weight and the whole life falls close to neutral.

Octopus welfare states (farmed O. vulgaris, recirculating grow-out). Grades in the order Nutrition / Environment / Health / Behaviour.
Welfare stateGrades N/E/H/BScore per year
Top (least bad)5 / 7 / 7 / 8-0.42
Mid (central)6 / 8 / 8 / 9-1.19
Bottom (worst)7 / 9 / 9 / 9-1.71
Mussel welfare states (standard EU commercial culture: Galician suspended raft, French bouchot, Dutch and Irish bottom culture). The standard commercial row is the central case.
Welfare stateGrades N/E/H/BScore per year
Near-wild4 / 4 / 5 / 4+0.36
Neutral5 / 6 / 6 / 5+0.10
Standard commercial (central)6 / 6 / 7 / 6-0.08
Precautionary6 / 7 / 7 / 6-0.15
Severe7 / 8 / 8 / 7-0.72

The mussel central is finely balanced. Because the life scores close to the neutral line, a one-step change in the health-domain grade (7 versus 6) moves the whole-life cost roughly tenfold, from near zero to about EUR 0.65 per kg. This instability is why the mussel result is presented as a band rather than a point, and why an AQALY scoring by a bivalve welfare specialist is the proper next step before any figure is sealed.

Moral weights and their sources

Moral weight is the term that decides the result. The Moral Weight Project of Rethink Priorities scores exactly eleven farmed taxa, none of them a bivalve; its ladder runs from pigs at 0.515 of the human welfare range down through octopuses at 0.213 and silkworms at 0.002. Rethink Priorities gives the reason for the absence: octopuses are likely conscious, whereas the evidence that bivalves are conscious is thin. A targeted search confirmed that no organisation has published a numeric moral weight or probability of sentience for any bivalve, and the Birch et al. 2021 sentience review that underpins the octopus case restricted its scope to cephalopod molluscs and decapod crustaceans.

Moral weights used (welfare range, human = 1). The mussel figure is a constructed neuron proxy; the octopus carries a published species estimate.
Species and basisCentralRange / bracketsNeuron count
Octopus, species estimate (Rethink Priorities welfare range)0.2130.004 (5th) to 1.471 (95th)
Octopus, neuron proxy0.0058500 million
Mussel, neuron proxy (no species estimate exists)0.000000870.00000012 to 0.0000035~75,000 (10,000 to 300,000)
Silkworm, comparator (Rethink Priorities lowest taxon)0.002

No whole-organism neuron count is published for any bivalve, so the mussel count is built from the nearest primary data: a stereological study of a small bivalve (Tantiwisawaruji et al. 2022) reporting 12,000 to 68,000 neural cells per single ganglion, and a mytilid mussel relative (Crenomytilus grayanus, Kotsyuba et al. 2020) at about 61,714 cells in its largest ganglion. Backing out non-neuronal cells and summing across the three pairs of ganglia gives roughly 75,000 neurons, with the wide 10,000 to 300,000 bracket the primary source itself carries. The human count of 86 billion is from Azevedo et al. 2009; the octopus figure of about 500 million is from Young 1963. Two cautions travel with the proxy and are carried into the sensitivity range, not resolved: Rethink Priorities argues neuron count should not be a sole proxy for moral weight, and the proxy captures welfare capacity conditional on the animal being a subject at all, which for bivalves is itself in doubt.

Prices, counts and durations, with sources

Market and biological inputs. First-sale (ex-vessel or ex-farm) prices, 2025 basis.
InputOctopusMussel
First-sale price~EUR 10 per kg~EUR 1.00 per kg
Price range and vintage7 to 10.5 per kg (wild first-sale, 2020)0.60 to 1.70 per kg (concrete points 2008 to 2013)
Animals per kg0.33 (about 3 kg each)40 (about 25 g each; band 25 to 55)
Life scored1 year1.5 years (range 1.0 to 2.0)
Price sourceEUMOFA, Octopus in the EU (2020)EUMOFA mussel price-structure analysis; counts cross-checked to Eurostat aquaculture statistics

The octopus price denominator is a wild-caught first-sale figure standing in for a farmed product that has never been sold, an unresolved question that belongs to the octopus account rather than this note. The mussel price is corroborated across searches but not yet confirmed against the source document, because the EU market observatory domain blocks the fetch tools; a manual download is needed before it ships in a client deliverable. The mussel count carries a genuine spread: the official EU stakeholder figures imply about 25 per kg, while Eurostat-derived and biometric counts give 40 to 55 per kg. The model uses 40 per kg, a linear multiplier carried through the sensitivity range.

Worked example (the central mussel figure, end to end)

standard commercial state 6/6/7/6 → S = -0.084 per year cost per kg = |S| × L × w × V × n = 0.084 × 1.5 × 0.00000087 × 147,000 × 40 = EUR 0.65 per kg // = 0.65× the EUR 1.00 first-sale price sentience-adjusted (× P = 3%) = EUR 0.02 per kg // about 2% of price

Results

Table 1. Mussel welfare cost (EUR per kg) and multiple of the EUR 1.00 first-sale price, by welfare state and moral weight. At the central price the EUR figure equals the multiple of price. Only the first three columns use the neuron proxy; the last two bound how far a different moral-weight assumption would move the result.
Welfare state (score/yr)Neuron 10k (low)Neuron 75k (central)Neuron 300k (high)Silkworm × P 5%Silkworm, certain
Near-wild (+0.36)0.362.7310.93146,271
Neutral (+0.10)0.100.742.96851,694
Standard commercial (-0.08)0.090.652.59741,482
Precautionary (-0.15)0.151.124.501292,578
Severe (-0.72)0.745.5722.363912,775

Rows with a positive score describe a net-positive life, so the cost there is the distance from an ideal life, not a realised harm; in those cases the robust harm is the slaughter alone, about EUR 0.0003 per kg under the neuron proxy, or 0.03% of price. At the neuron-proxy central setting and the standard commercial grading the whole-life cost is about EUR 0.65 per kg. The break-even moral weight confirms the direction: at this grading the cost matches price only when the moral weight reaches about 0.0000014, above the central neuron proxy of 0.00000087, so the proxy puts the cost below price. A move to a species-style weight is what would push it above.

Table 2. Welfare-cost comparison, farmed octopus and farmed mussel.
DimensionFarmed octopusFarmed mussel
Moral weight (human = 1)0.213 species; 0.0058 neuron proxy0.00000087 neuron proxy; no species estimate
Farmed-life score per year-0.42 to -1.71 (worse than death)+0.36 to -0.72 (straddles neutral)
Animals per kg0.3340
First-sale price (2025)~EUR 10 per kg~EUR 1.00 per kg
Welfare cost, neuron proxyEUR 118 to 486 per kg (12 to 49×)EUR 0.65 per kg central (0.1 to 1.1×)
Welfare cost, species weightEUR 4,300 to 17,800 per kg (430 to 1,800×)no species estimate
Cost a multiple of price?yes, at every settingno under the neuron proxy

The gap between the two animals is roughly two to three orders of magnitude and comes from two compounding factors that both cut the mussel cost, offset by one that partly restores it. Moral weight falls by about 244,000 times from octopus to mussel, and the welfare state moves from clearly worse than death to roughly neutral. Against these, the mussel carries about 120 times more animals per kg, which holds the per-kg cost at a meaningful fraction of price rather than a negligible one. The octopus result is robust: it exceeds price by at least 12 times even at its most conservative presented corner. The mussel result is not: it exceeds price only if the neuron proxy is set aside in favour of a moral weight for which no evidence exists.

What would change the conclusion

The conclusion that mussel welfare cost is manageable depends on the neural-weight proxy and on a low probability of sentience. It moves toward octopus-like multiples only under an assumption the evidence weighs against: granting a sessile animal with three pairs of ganglia and no central brain a silkworm-level welfare range (0.002), which at a 5% probability of sentience lifts the cost to about 74 to 130 times price. A third uncertainty is more fundamental than the moral weight and does not resolve in either direction: whether a welfare instrument calibrated on birds, mammals and cephalopods can validly score a bivalve at all. This is not a probability that can be discounted like sentience; it is a limit on what the whole construction can claim, and it is the reason the mussel figures are conditional upper bounds rather than expected costs.

Caveats and what was not verified

Sources

Citations resolved and cross-checked against DOIs, PubMed and publisher pages. Two items (the EUMOFA price documents) reached only at search-snippet tier because the market-observatory domain blocks automated retrieval, and are flagged for a manual confirmation before any figure ships.

Method and monetary value

  1. Espinosa, R., Browning, H., & Treich, N. (2025). Animal Quality-Adjusted Life Year (AQALY) (SSRN Working Paper No. 5255360). papers.ssrn.com/abstract=5255360 working paper
  2. Budolfson, M., Espinosa, R., Fischer, B., & Treich, N. (2024). Monetizing animal welfare impacts for benefit–cost analysis. Journal of Benefit-Cost Analysis, 15(S1), 206–223. https://doi.org/10.1017/bca.2024.19 peer-reviewed
  3. Téhard, B., Detournay, B., Borget, I., Roze, S., & de Pouvourville, G. (2020). Value of a QALY for France: A new approach to propose acceptable reference values. Value in Health, 23(8), 985–993. https://doi.org/10.1016/j.jval.2020.04.001 peer-reviewed – source of the EUR 147,000 lower-bound QALY value (VSQ EUR 147,093 at a 2.5% discount rate).

Moral weight and sentience

  1. Rethink Priorities. (2022). Welfare range estimates (Moral Weight Project). rethinkpriorities.org/research-area/welfare-range-estimates institutional – welfare-range figures verified (pigs 0.515, chickens 0.332, octopuses 0.213, silkworms 0.002); the characterization of bivalve sentience is paraphrased and its exact source page was not confirmed.
  2. Birch, J., Burn, C., Schnell, A., Browning, H., & Crump, A. (2021). Review of the evidence of sentience in cephalopod molluscs and decapod crustaceans. London School of Economics and Political Science (commissioned by Defra). LSE report page institutional
  3. Schnell, A. K., Browning, H., Crump, A., Burn, C. C., & Birch, J. (2026). Sentience in cephalopod molluscs: An updated assessment. Biological Reviews, 101, 1311–1333. https://doi.org/10.1002/brv.70125 peer-reviewed

Neuron counts

  1. Tantiwisawaruji, S., Rocha, M. J., Silva, A., Pardal, M. A., Kovitvadhi, U., & Rocha, E. (2022). A stereological study of the three types of ganglia of male, female, and undifferentiated Scrobicularia plana (Bivalvia). Animals, 12(17), 2248. https://doi.org/10.3390/ani12172248 peer-reviewed
  2. Kotsyuba, E., Kalachev, A., Kameneva, P., & Dyachuk, V. (2020). Distribution of molecules related to neurotransmission in the nervous system of the mussel Crenomytilus grayanus. Frontiers in Neuroanatomy, 14, 35. https://doi.org/10.3389/fnana.2020.00035 peer-reviewed
  3. Azevedo, F. A. C., Carvalho, L. R. B., Grinberg, L. T., Farfel, J. M., Ferretti, R. E. L., Leite, R. E. P., Jacob Filho, W., Lent, R., & Herculano-Houzel, S. (2009). Equal numbers of neuronal and nonneuronal cells make the human brain an isometrically scaled-up primate brain. Journal of Comparative Neurology, 513(5), 532–541. https://doi.org/10.1002/cne.21974 peer-reviewed – source of the 86 billion human-neuron figure.
  4. Young, J. Z. (1963). The numbers and sizes of nerve cells in Octopus. Proceedings of the Zoological Society of London, 140(2), 229–254. https://doi.org/10.1111/j.1469-7998.1963.tb01862.x peer-reviewed – source of the ~500 million octopus-neuron figure.

Welfare burden and husbandry

  1. Aquaculture Advisory Council. (2024). Recommendation on molluscan welfare. aac-europe.org (PDF) institutional – document located and live; the exact quoted husbandry wording was not pulled from the full text and needs a manual confirmation before any verbatim quote.
  2. Schotanus, J., van de Vis, H., & Steins, N. (2022). Welfare of bivalves in aquaculture? A first literature scan and plan for a comprehensive search (Wageningen Marine Research Report No. C057/22). Wageningen University & Research. https://doi.org/10.18174/577988 institutional

Prices and volume

  1. EUMOFA. (2020). Octopus in the EU: Price structure in the supply chain (Case study). European Market Observatory for Fisheries and Aquaculture Products. eumofa.eu (PDF) institutional · snippet-tier – first-sale prices at search-snippet tier only; manual download pending.
  2. EUMOFA. (n.d.). Fresh mussels: Price structure in the supply chain (Case study). European Market Observatory for Fisheries and Aquaculture Products. eumofa.eu (PDF) institutional · snippet-tier – first-sale range at search-snippet tier only; manual download pending.
  3. Eurostat. (2023). Aquaculture production by species [Dataset fish_aq2a]. ec.europa.eu/eurostat (fish_aq2a) institutional – used to cross-check EU mussel output and the animals-per-kg count.