Bluetongue in the UK: who bears the cost, and when does control pay?

Main findings

Bluetongue creates several economic problems at once: damage to affected animals, pressure on farm resources, disruption to sales and movements, and changes in prices paid and received. A measure that improves an infected farm's cash position need not produce an equivalent gain for society. Conversely, an uninfected business may still face costs even when its animals remain healthy.

Three complementary models were run for this note. In the central farm LP, vaccination improves annual contribution by £1,222 compared with BTV without added control. In the illustrative regional DP, a state-dependent programme reduces four-year discounted resource costs by £89,667. In the central UK-scale PE scenario, vaccination creates a conditional annual net domestic welfare benefit of £21.32m, after £67.39m of assumed delivery costs. These figures describe different systems and periods; they must not be added.

The central result is conditional. Lower exposure, weaker protection or higher delivery costs can remove the economic advantage. The useful policy question is therefore which animals and businesses to protect, at what cost, at what time, and with whose funding.

1. What the recorded cases establish

The common UK release reports 1,401 confirmed affected premises in the 2026–2027 season and separately lists 1,616 potential cases under investigation in Great Britain. Pending investigations are not added to confirmed cases. A confirmed premises requires at least one positive animal; this is not a count of all infected animals. The source was updated on 18 September, with the county table labelled to midnight 17 September. [1]

*Scotland reconciliation. The Scottish page updated on 19 September states 30 affected holdings, while its location list sums to 32; it also lists 19 sheep and 72 cattle confirmed. Those animal counts use a different unit from holdings. The comparative analysis retains 31 from the common UK release for the comparative chart; the discrepancy is not interpreted as growth. [2]

England and Wales have country-wide restricted zones. Scotland's south-west zone changed on 10 September. DAERA describes Northern Ireland's earlier incursion and suspended disease-free status; therefore zero newly recorded premises in this season does not imply no trade constraints, prevention expenditure or future risk. [1] [3] [4]

Why no national loss is calculated by multiplying cases by a standard loss. The public premises counts do not supply species composition, animals at risk within each holding, clinical attack rates, vaccination coverage, production system or actual financial outcomes. A mixed upland farm is not interchangeable with a large dairy. Under-ascertainment and changes in confirmation practice add uncertainty. The case data motivate the scenarios and geographic priorities; they do not calibrate the models' exposure parameters.

2. Materials and methods

Design and counterfactuals

This is a deterministic sensitivity analysis of economic decision models, with an interpolated state-transition DP. The LP and PE each compare a without-BTV reference, BTV without additional control, and BTV with vaccination. The DP compares no disease, disease with voluntary farm responses but no added public programme, and an adaptive public programme. Existing market arrangements form the reference; no additional blanket meat or milk export ban is imposed.

All runs use Model Lab engine 1.1, archived with the article. The farm LP represents monthly enterprise resources, the DP represents sequential policy decisions and farm responses, and the PE represents commodity-market welfare. Their methodological foundations and published antecedents are detailed in the research-lineage section. These are adaptations, with new BTV assumptions, rather than a claim that bTB or FMD parameters transfer unchanged. The runnable models, full inputs and numerical results are linked above.

Observed inputs and scenario assumptions

UK annual output, trade and breeding-population anchors come from Defra's 2025 agricultural statistics, updated 3 September 2026. Annual GB meat price proxies are £7.10/kg for sheep and £6.47/kg for beef; milk is £0.44/litre. These are reference averages, not September spot quotations. Multiplying model quantities by those prices does not reproduce official production-value accounts exactly. [5]

Protection is implemented with aggregate coverage multipliers. It is not an individual-animal vaccine-response model, and infection and severity effects are not product-specific trial estimates. A severity-only case and weaker protection are tested. Product, course, onset, handling cost and record-keeping requirements must be checked against current veterinary guidance before practical use. [8]

Farm LP

The linear programme maximises annual enterprise contribution subject to monthly metabolisable-energy, crude-protein, intake, forage and labour constraints, annual purchased-feed limits, and herd bounds. It has 146 decision variables: two enterprise head counts and monthly feed quantities by species and feed. The inherited farm includes 666 ha rough grazing, 40 ha improved pasture, 20 ha aftermath and 22 ha silage/hay land. These are illustrative resource accounts; their area categories are not assumed to be independent cadastral parcels.

Original output coefficients are £88/sheep and £992/cattle per year, with non-feed costs of £28 and £168. They are inherited planning coefficients, not newly extracted national accountancy estimates. Disease losses include mortality, surviving-animal output loss, treatment and fertility effects, with additional labour and nutritional requirements. Assumed untreated mortality is 4% for sheep and 0.5% for cattle conditional on exposure; output reductions are 15% and 5%. Full monthly coefficients are in the download. The original handbook provides budgeting context, not empirical validation of these BTV inputs. [9]

Farm contribution excludes common overheads, rent, support payments and unpaid family labour; it is not Farm Business Income. Herd choices are continuous planning equivalents, not a prescription to sell fractional animals. Subsidies and compensation enter the private budget, with corresponding government costs. The fixed-price resource measure is farm contribution minus government expenditure.

Market PE

Three independent markets clear domestic supply plus imports against domestic demand plus exports. Isoelastic curves are calibrated to the reference quantities and prices. Infected and uninfected producer supply are represented separately. For affected output, sheep production falls 15% and cattle production 5%; added marginal costs are £0.12/kg sheepmeat, £0.08/kg beef and £0.01/litre milk. Supply elasticities are 0.30, 0.25 and 0.20; demand elasticities are −0.50, −0.50 and −0.30. Import and export elasticities are 1.5 and 1.0 throughout. These are assumptions, not econometric estimates.

Consumer/derived-demand and producer-surplus changes are integrated from the curves. Real control and administration costs reduce welfare; subsidy and compensation payments transfer money between producers and government. Vaccination costs cover 14.78 million breeding sheep, 1.29 million beef cows and 1.85 million dairy cows at 80% coverage, yielding £67.392 million per annual programme. This is an assumed breeding-stock strategy, not whole-herd coverage or observed spending. The same protection multipliers apply to affected market output, despite using breeding-stock numbers for programme costing. This strong simplifying assumption may overstate protection per pound spent if youngstock also need direct vaccination. Youngstock vaccination and additional delivery overheads could increase cost; the double-cost sensitivity is therefore particularly important.

Policy DP

The illustrative region contains 10,000 sheep, 2,000 beef cattle and 1,000 dairy cattle. Assumed loss per infection is £60, £200 and £300 respectively. These independent reduced-form loss assumptions are not generated by the LP. The DP has 150 combinations of disease-pressure, vaccination and natural-immunity states. It applies quarterly seasonal weights of 0.05, 0.25, 0.50 and 0.20 across a prospective cycle beginning in a low-activity quarter; it is not a September-start forecast.

Backward induction minimises discounted disease damage, control resources, administration and movement disruption. Farm responses choose among no action, vaccination, other protective measures, or both using a one-quarter look-ahead; this is bounded behaviour, not a full multi-agent equilibrium. The government chooses no added programme, vaccination subsidy, vaccination plus vector support, or targeted movement/testing support. The discretionary ceiling is £50,000 per quarter. Forced policies are comparison paths and are not constrained by that ceiling.

Default quarterly vaccine and natural-immunity waning are 15% and 5%; pressure carries at 0.95 within the cycle and 0.40 at overwintering, with feedback 0.50 and external pressure 0.03. State interpolation conserves probability. These transition assumptions are uncalibrated. The finite horizon has zero terminal value, which may favour less late-horizon investment; it is not evidence that protection should stop after four years.

Appraisal and verification

Net benefit equals the reduction in disease and market damage minus incremental resource costs. The PE benefit–cost ratio divides avoided welfare damage before control costs by programme resource costs. The DP ratio compares avoided disease damage with incremental control and public-resource cost, including private action already present in its baseline. Transfers are not counted twice. LP solutions were checked against SciPy/HiGHS; the original LP baseline was reproduced. DP forward costs reconcile to Bellman values and PE markets and accounts balance. These are numerical checks, not epidemiological or empirical validation.

3. Farm results: exposure and spare capacity determine the return

Without BTV, the model chooses approximately 481.4 sheep and 55 cattle, generating £64,544. At central exposure, contribution falls by £4,462. Vaccination costs about £1,883 and recovers £1,222 of contribution after paying those costs. The implied avoided damage and resource pressure is approximately £3,105, giving an illustrative benefit–cost ratio of 1.65.

At 5% exposure, intervention reduces contribution by £1,104; at 40%, it improves it by £4,371. With all other coefficients held fixed, the model's break-even exposure is approximately 12.1%. This is a property of this budget, not a universal vaccination threshold or a recommendation to defer veterinary advice.

The original baseline herd becomes infeasible under added disease labour requirements because the initial optimum already uses a binding monthly resource. Allowing adjustment reduces central sheep numbers from 481.4 to 471.6 without control, versus 478.4 with vaccination. In a separate check, giving the original herd 10% additional monthly labour makes both disease scenarios feasible. This highlights the importance of relief labour and feed planning; it does not predict compulsory destocking.

4. Market results: a national total can hide opposite effects

In the central scenario, BTV reduces domestic welfare by £126.77m. With vaccination the loss is £105.45m, including £67.392 million in control resources. Avoided damage before those costs is £88.72m; the benefit–cost ratio is 1.32. The low-exposure programme has a ratio below one, whereas the high-exposure programme has a ratio of 2.67. This is evidence for examining targeting, coverage and price—not a claim that the real 2026 programme has those returns.

The central no-control scenario gives infected producers a £111.00m loss, while uninfected producers gain £111.92m through price effects. Buyers and derived demand lose £127.69m. In the intervention scenario, lower price pressure and broadly distributed vaccination costs reduce the uninfected group's aggregate advantage. The groups change size when infection is prevented, so these columns should not be read as the treatment effect on an unchanged group of farms.

Government spending is zero in the central PE because no subsidy, compensation or administrative programme is specified—not because public services are costless. With a 75% vaccination subsidy, the model allocates £50.544 million to government and the same amount to producers; at fixed coverage, total welfare is unchanged. The DP differs because subsidies can change behaviour. Milk buyer surplus includes manufacturing demand and cannot be interpreted solely as household consumer welfare.

5. Dynamic policy results under the specified assumptions

The adaptive central solution reduces total discounted cost from £455,268 to £365,601: a saving of £89,667, or 19.7%. Disease damage falls by £134,057, while incremental private control and public-resource costs increase by £44,390. Their ratio is 3.02. Government pays £100,407; private costs fall to £265,195. Government expenditure includes transfers and therefore is not an additional social loss on top of the total.

The first low-activity quarter selects vaccination support. Later decisions vary across possible states; the published curves average over that state distribution. A permanent vaccination subsidy costs £401,248, more than the adaptive policy. Permanent movement/testing support costs £624,718, exceeding the no-added-programme baseline under the assumed limited effect and disruption costs. Neither finding establishes what current real-world movement policy should be.

Across low, central and high initial pressure, adaptive-policy savings are £41,885, £89,667 and £35,274. They need not rise monotonically: higher pressure also changes voluntary action and the remaining opportunity for government to improve outcomes.

6. Sensitivity and comparison with published evidence

Halving both protection assumptions reverses the central LP and PE conclusions. Doubling course delivery costs also reverses them. Removing assumed infection prevention while retaining severity protection leaves a smaller positive result. Adding other protective measures with a 10% effect and £1/head cost reduces net benefit relative to vaccination alone. More measures are not automatically more economical. These variations isolate selected uncertainties; they do not represent an exhaustive probabilistic sensitivity analysis.

Published evidence supports treating control costs as part of the appraisal. The Welsh Government's July 2026 BTV-8 assessment cites unpublished Defra modelling with central movement-control losses typically £5–£29 million relative to no restrictions; its conclusion uses the wider £5–£40 million range. The underlying analyses and scenario boundaries are not reproduced in that public summary. This is directionally consistent with testing disruption costs in the DP, but it does not validate the £624,718 regional scenario, concern the same population, or measure observed 2026 BTV-3 damage. [6]

The French ministry describes nearly €22 million for state-funded vaccine doses under its 2025 BTV-8/BTV-1 strategy. That is announced supply financing, not an audited realised-loss estimate and not a new UK programme. It illustrates why payer, serotype, target population and delivery scope must accompany every spending comparison; a numerical comparison with the £67.392 million UK breeding-stock scenario would otherwise be misleading. [7]

Current UK guidance differentiates authorised vaccine indications and species. It supports checking product-specific protection rather than applying a single universal efficacy number. The present sensitivity ranges are therefore exploratory; this note is not a comparative vaccine assessment. [8]

What remains outside the estimates

The three models are complementary, not an integrated multiscale forecast. The LP has fixed prices; the PE has independent markets without substitution between meats, a full feed-market response or a separate processing sector; the DP has assumed transitions and simplified behavioural responses. They do not value animal suffering, farmer wellbeing, environmental effects, resistance outcomes or all supply-chain disruption. Imports and exports respond in the PE, but foreign welfare and trade rents are excluded. Financing distortions and deadweight costs of taxation are not modelled. The figures are partial resource/welfare measures, not GDP impacts or complete social cost.

Country-level economic losses cannot be identified from the current premises data alone. England, Wales, Scotland and Northern Ireland also differ in livestock mix, farm structure, market access and disease history. Assigning the same average farm loss to every confirmed holding would conceal those differences. The national market scenarios apply UK reference data, while the LP and DP provide transparent representative decision exercises.

Comparison with the Scottish 2008–2009 BTV appraisal

Fofana and colleagues’ study, An ex-ante economic appraisal of Bluetongue virus incursions and control strategies, concerns the Scottish policy problem of 2008–2009; the journal citation is 2016, volume 154, pages 118–135, with online publication in 2015. [10]

The most important distinction: prevention versus mitigation

The historical appraisal compares the expected outbreak costs avoided by successful prevention with public and private surveillance/prevention expenditure. Separately, it asks which reactive strategy would leave the lowest losses if an incursion occurred. An expensive residual outbreak makes prevention look more valuable; it does not make that outbreak-control strategy the best mitigation choice. This explains why ranking prevention benefit–cost ratios as if they were incremental vaccination returns would be misleading.

The study reports approximately £141 million of discounted surveillance/prevention costs over five years and approximately £330–£500 million of potential avoided outbreak costs across scenarios. These are historical modelled aggregates, not an observed Scottish outbreak bill, a current budget, or an estimate of 2026 losses. Its broad surveillance/prevention account also differs from the narrowly incremental programme costs here. For one example in Table 2—April 2009 midge incursion, persistence after year two, no move-to-slaughter licence—the reported losses are £344.5m without vaccination, £470.5m under the border-zone option and £340.7m under 50% farm vaccination with an all-Scotland zone. Those values explain the distinction between valuing prevention and minimising residual losses; they are not new model outputs. [10]

The abstract favours 50% farm vaccination with an all-Scotland protection zone under its assumptions. The detailed results are conditional on incursion route, persistence and movement-to-slaughter arrangements, and some comparisons favour no vaccination or a different zone. These results depend on the counterfactual and do not establish 50% coverage as a transferable optimum. BTV-8 biology, historical legislation, assumed demand effects and vaccine prices cannot simply be relabelled as BTV-3 in 2026.

Coverage and persistence sensitivity analyses

The LP and PE are re-solved at six animal-coverage levels; the DP compares two five-year persistence assumptions. These analyses examine questions highlighted by the historical study. They do not reproduce its spatial epidemic model. Interactive comparisons in the Model Lab use the entered assumptions, and the resulting inputs and outputs can be downloaded.

The model’s economic response to coverage depends on entered exposure, protection and costs. It does not include a calibrated herd-immunity threshold or a between-farm network. A maximum on this six-point grid is therefore only the best of these tested scenarios—not an epidemiologically established optimum. A 50% farm-coverage strategy could have quite different animal coverage where holding sizes differ.

The first DP comparison extends the entered central assumptions to five years. The second uses lower within-cycle pressure carry-over (0.60 instead of 0.95), lower overwinter carry-over (0.10 instead of 0.40), and no external pressure throughout. These deliberately contrasting assumptions are not the historical paper’s fade-out-after-year-two trajectory and do not demonstrate actual containment. Both re-optimise public policy and farm responses; the useful comparison is sensitivity of investment returns to persistence, not agreement in monetary magnitude.

Attribution: the research behind LP, DP and PE

LP. The extensive-sheep modelling tradition was developed by Alistair Stott and colleagues. Stott, Milne, Goddard and Waterhouse (2005) provide foundational economic/welfare work; Stott, Vosough Ahmadi and colleagues (2012) provide the later joint publication, Interactions between profit and welfare on extensive sheep farms. The present implementation draws on this research tradition but does not reproduce either published model exactly. The LP’s explicit labour constraint and fixed-herd feasibility comparison remain important: a profit gain is not itself proof of an animal-welfare gain. [11] [12]

DP. The joint research includes Vosough Ahmadi, Morgan and Stott (2009), a suckler-cow replacement working paper, and the peer-reviewed Vosough Ahmadi, Nath, Hyslop, Morgan and Stott (2017; online 2016). The latter estimates transitions from herd records and uses parity, calving period and body condition in a 150-state replacement DP. The current BTV programme also has 150 states, but they describe exposure and protection—not the same states or calibrated transition probabilities. Its finite-horizon public-policy objective and bounded farm responses are a distinct decision problem. The cited studies support the methodological lineage; their empirical validation does not extend to the present BTV transition assumptions. [13] [15]

PE. The directly relevant welfare reference is Alyson S. Barratt, Matthieu H. Arnoult, Bouda Vosough Ahmadi, Karl M. Rich, George J. Gunn and Alistair W. Stott (2018). Their Johne’s disease paper distinguishes infected producers, uninfected producers and consumers. The present analysis uses that distributional perspective with separate BTV assumptions: loss to an infected producer is not necessarily the whole market’s loss, and unaffected producers can benefit from higher prices. The current BTV implementation adds separate sheepmeat, beef and milk markets, trade curves and intervention accounts; it is not a reproduction of the Scottish dairy application. [14]

What this changes for interpretation

First, appraise prevention, outbreak mitigation and compensation as different decisions. Second, test coverage, persistence, delivery cost and movement disruption together before choosing a programme. Third, maintain separate private, government and welfare accounts. Fourth, build the next calibration around farm-level losses, vaccination uptake and a validated regional exposure model—not a multiplication of recorded premises by historical losses. The older study strengthens the analytical questions and attribution, while the contemporary evidence must still determine the answer.

Edition note: Tables 9–10 and Figure 7 extend the main scenarios through coverage and persistence sensitivity analysis. Historical nominal costs have not been used as current parameter estimates.

7. Conclusions and recommendations

1. For farms: price the actual decision. Combine a veterinary assessment of prospective exposure with species, animal value, production stage, delivery quotations and spare labour. Compare avoided loss with the complete cost of protection. Use the LP to identify resource bottlenecks as well as margin changes.
2. For livestock businesses: consider both infection and market exposure. Healthy animals do not insulate a business from price changes, delayed movements, customer disruption or prevention costs. Review supplier concentration, sale timing and liquidity; do not equate the model's uninfected-producer gains with guaranteed business profits.
3. For regional and national policy: test timing and additionality. Compare adaptive, targeted support with permanent programmes and voluntary uptake. Identify the intervention-induced change in behaviour before claiming a public subsidy creates an equivalent social benefit. Reassess movement measures against their specific risk reduction and disruption cost.
4. For compensation design: separate relief from prevention. Compensation may support affected farms and improve reporting incentives, but a payment is not itself recovered production. Show the farmer, taxpayer and resource accounts separately, and examine behavioural consequences.
5. For the next empirical study: collect the missing denominator and outcomes. Link premises reports to species, herd/flock size, within-holding morbidity and mortality, vaccination date and coverage, milk/growth/fertility outcomes, treatment, labour and trading costs. Include uninfected comparison holdings and protect farm confidentiality.
6. For public communication: publish a dated result with its assumptions. Update surveillance separately from model recalibration. Revise this article only when a new evidence set and rerun are available; do not allow a live case counter to silently change a published research conclusion.

The strongest conclusion is that the economic case for control is specific to exposure, protection, delivery cost and timing. Transparent models can make those dependencies visible. They cannot turn a premises count into an observed national bill without the intervening evidence.

References and reproducibility

1. Defra/APHA. Bluetongue: latest situation. Updated 18 September 2026; checked 19 September.
2. Scottish Government. Latest situation. Updated 19 September 2026.
3. Welsh Government. Bluetongue virus: current restrictions. Checked 19 September 2026.
4. DAERA. Bluetongue: Northern Ireland status and trade context. Checked 19 September 2026.
5. Defra. Agriculture in the United Kingdom 2025, chapter 8. Updated 3 September 2026.
6. Welsh Government. Changes in the approach to handling BTV-8 cases: integrated impact assessment. 17 July 2026.
7. French agriculture ministry. FCO situation and vaccination strategy. Page dated 10 September 2026; spending refers to 2025.
8. Defra/VMD. BTV-3 vaccination guidance. Updated 15 September 2026.
9. SAC/FAS. Farm Management Handbook 2025/26. Budgeting context referenced by the original LP; coefficients here are inherited model assumptions.
10. Fofana, A., Toma, L., Moran, D., Gunn, G. J., Gubbins, S., Szmaragd, C. and Stott, A. W. (2016). An ex-ante economic appraisal of Bluetongue virus incursions and control strategies. Journal of Agricultural Science, 154(1), 118–135. Online publication 2015.
11. Stott, A. W., Milne, C. E., Goddard, P. J. and Waterhouse, A. (2005). Projected effect of alternative management strategies on profit and animal welfare in extensive sheep production systems in Great Britain. Livestock Production Science, 97, 161–171.
12. Stott, A. W., Vosough Ahmadi, B., Dwyer, C. M., Kupiec, B., Morgan-Davies, C., Milne, C. E., Ringrose, S., Goddard, P., Phillips, K. and Waterhouse, A. (2012). Interactions between profit and welfare on extensive sheep farms. Animal Welfare, 21(S1), 57–64.
13. Vosough Ahmadi, B., Morgan, C. A. and Stott, A. W. (2009). Trade-offs between conflicting animal welfare concerns and cow replacement strategy in out-wintering Scottish suckler herds. SAC Land Economy Working Paper 34; AgEcon Search record 61122.
14. Barratt, A. S., Arnoult, M. H., Vosough Ahmadi, B., Rich, K. M., Gunn, G. J. and Stott, A. W. (2018). A framework for estimating society’s economic welfare following the introduction of an animal disease: The case of Johne’s disease. PLOS ONE, 13(6), e0198436.
15. Vosough Ahmadi, B., Nath, M., Hyslop, J. J., Morgan, C. A. and Stott, A. W. (2017). Trade-offs between indicators of performance and sustainability in breeding suckler beef herds. Journal of Agricultural Science, 155(1), 156–170. Published online July 2016.

Evidence checked and models run: 19 September 2026. Article version 1.2. Exact executable engine archived as model-engine.js. Publication data remain fixed to this edition. This is a focused comparison of published research and verified official evidence, not a systematic literature review. Download reproducibility bundle · Full machine-readable assumptions and results · Archived model engine · Scenario specification · Results tables (CSV) · Model equations and lineage.