What does BTV cost?
What is intervention worth?

Research lineage and the Scottish BTV study

The extensive-sheep modelling lineage includes Alistair Stott and colleagues (2005), and Stott, Bouda Vosough Ahmadi and colleagues (2012). The DP lineage includes Vosough Ahmadi, Morgan and Stott (2009) and Vosough Ahmadi, Nath, Hyslop, Morgan and Stott (2017; online 2016), on suckler-herd decisions. The welfare framework is informed by Alyson Barratt, Arnoult, Vosough Ahmadi, Rich, Gunn and Stott (2018), on Johne’s disease. These citations establish methodological ancestry, not identity between published code and the current browser implementation.

Fofana et al. (2016; online 2015) evaluated hypothetical Scottish BTV-8 incursions in 2008–2009. Its prevention benefit–cost ratios compare avoided outbreak costs with surveillance/prevention expenditure. The present intervention net benefits compare BTV with and without additional control. Different counterfactuals, serotypes, geography, price bases and periods prohibit direct comparison of the totals or ratios.

Sensitivity comparisons: LP and PE now re-solve at 0%, 25%, 50%, 75%, 80% and 100% animal coverage. DP offers two five-year persistence stress tests. The lower-persistence case sets carry-over to 0.60, overwinter carry-over to 0.10 and external pressure to zero throughout, rather than replicating the paper’s post-year-two fade-out. None of these settings is estimated from current cases. Historical vaccine prices, protection-zone rules and the paper’s farm-coverage optimum are not imported into current defaults. Read the comparison and results.

1. 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.

   Historical Scottish BTV-8 prevention appraisal; 2008–2009 incursion scenarios, not a 2008 publication or measured outbreak loss.

2. 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.

   Foundational extensive-sheep economic and welfare modelling by Stott and colleagues; precursor literature, not verification of identical current coefficients.

3. 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.

   Published joint extensive-sheep modelling lineage. Labour and welfare matter alongside profit. Cambridge's 2023 online-hosting date is not the original publication year.

4. 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.

   Verified joint DP antecedent: infinite-horizon cow replacement, not a published bTB policy model. The present finite-horizon BTV government DP is a different application.

5. 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.

   PE welfare lineage: Alyson Barratt and co-authors, including Bouda Vosough Ahmadi and Alistair Stott. Infected/uninfected producers and consumers; not a BTV calibration.

6. 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.

   Peer-reviewed DP lineage: empirically parameterised cow replacement using parity, calving period and body condition. Its 150 states are not the 150 exposure/protection states in this BTV adaptation.

Model structure and adaptation

These BTV implementations adapt existing farm-planning, policy and welfare models. Disease mechanisms, economic accounts and numerical implementation differ from the antecedent applications; methodological lineage does not imply empirical validation for BTV.

LP: a full monthly resource problem

Decision variables are sheep and cattle head counts plus daily dry-matter allocations for each of six feeds, each species and each month: 146 variables. The objective maximises gross output less non-feed costs, feed costs, expected disease damage and intervention cost, plus transfers received. Constraints enforce monthly metabolisable energy and crude protein requirements, maximum dry-matter intake, concentrate limits, monthly forage and labour availability, annual silage/hay and purchased-feed limits, and herd minimum/maximum bounds.

A two-phase simplex solver allows non-zero minimum herd sizes and reports infeasibility explicitly. Each scenario is optimised separately. A second comparison retains the disease-free optimal herd and re-optimises feeding, exposing the value and limits of enterprise adjustment. Continuous head counts are planning equivalents; rounding a result requires a new feasibility check.

Expected exposure under intervention = entered probability × (1 − coverage × infection protection) × (1 − additional exposure reduction). A separate average severity modifier applies to mortality, survivor output loss, treatment and fertility losses. This is an expected-coefficient LP, not a stochastic recourse model. Severity uses a population-average approximation; it does not reconstruct individual breakthrough infections. Feed requirements are not reduced after mortality, so this represents maintained/replacement breeding capacity rather than a detailed within-year flock inventory.

Additional fertility losses must exclude lost output already captured by the output-loss percentage. Extra labour consumes the entered resource allowance; unpaid labour is not additionally monetised. Common fixed costs, rent, support payments and unpaid family labour are outside the contribution measure. Farm Business Income is not interchangeable with this objective.

DP: policy, incentives and repeated decisions

The state is vector exposure pressure (six grid points), vaccine-protected share (five) and naturally immune share (five): 150 combinations per quarter. The horizon is 1–10 years, with four seasonal steps per year. The Bellman recurrence is V(t,s) = min [social_cost(t,s,a) + discount × Σ transition_probability × V(t+1,s′)]. The terminal value is zero; no arbitrary bTB eradication penalty or trade-freedom bonus is imported.

Government actions are no added programme, a vaccination subsidy, vaccination plus vector support, and targeted movement/testing support. Farm types choose no extra action, vaccination, other exposure-reduction measures, or both. Their response minimises current private cost plus one-quarter anticipated disease cost. This is a bounded-lookahead farm response inside a government DP, not a second full farm-level lifetime DP or a fully strategic equilibrium.

Exposure probability is derived from a declared seasonal pressure using 1 − exp(−pressure × seasonal_weight × modifiers). It is not fitted from confirmed premises. Existing vaccine and natural protection modify exposure; vaccine protection also modifies severity. New courses protect from the next quarterly step. Protection can wane, susceptible stock can be replenished, and future pressure depends on carry-over, declared external exposure and previous infections. Populations are held constant, with loss per infected animal covering the chosen replacement and production consequences.

Off-grid next states are distributed to surrounding grid states by trilinear interpolation. These weights are a numerical approximation, not empirically estimated transition uncertainty. Forward charts propagate the full state distribution and match the Bellman value, rather than following a single rounded path. Vaccine and natural protection are represented as overlapping aggregate shares under an independence assumption. The three farm types share that aggregate protection state. The model is regional and non-spatial; there is no estimated between-region transmission network or multiple-serotype interaction.

Government chooses the least-cost feasible policy within the specified budget in each state/quarter. No-added-programme expenditure, including any hypothetical compensation, remains payable even if it exceeds the discretionary cap. Fixed-policy comparators are unconstrained and may exceed the budget. Zero BTV is shown as zero incremental disease/control cost, not zero agricultural operating cost. The no-added-programme comparator allows private voluntary responses.

PE: market clearing and distribution

Each commodity solves a separate market equilibrium. Domestic uninfected supply is (1−f)Q₀(P/P₀)^εs; infected supply is fQ₀(1−loss)[max(P−cost,0)/P₀]^εs. Demand is D₀(P/P₀)^εd, imports rise with price and exports fall with price, with an optional explicit export-demand shock. The price is solved so domestic production + imports = domestic demand + exports. No bilateral trade flows, cross-commodity substitution or general-equilibrium feedback is assumed.

Consumer/derived-demand surplus change is minus the integral of demand between baseline and scenario prices. Producer surplus is the integral of supply, calculated separately for infected and uninfected cohorts. Added marginal disease costs enter the supply curve; do not add the same treatment costs again. Control programme resources are subtracted once, with subsidies and compensation allocated between producers and government. Public administration is additional resource cost. The reported domestic partial welfare change excludes foreign welfare, importer/exporter rents, processing margins, taxes, environmental impacts and non-market animal welfare. It is not GDP loss or an all-society comprehensive valuation.

Intervention changes the affected share, so infected and uninfected groups have different membership between scenarios. Higher prices can benefit uninfected producers. Production value at risk is contextual, not added to welfare loss. A farm-level LP loss must not be added to PE welfare loss: these are alternative analytical perspectives with overlapping mechanisms.

The multi-year PE total repeats the annual scenario and discounts years 0 to H−1. It is an appraisal sensitivity, not a dynamic disease trajectory. Programme animal populations are explicit eligible breeding cohorts; the link from their coverage to the affected output share is a scenario assumption. Milk quantities include demand for manufacturing, so the milk surplus is a derived-demand measure rather than solely household consumer welfare.

Evidence and assumptions

All disease probabilities, severity parameters, efficacy values, behavioural responses, elasticities, waning, programme costs and subsidies are editable scenario assumptions. Efficacy defaults are not estimates for any named product. Set infection protection to zero to explore a severity-only vaccine scenario. Additional vector and movement effects require independent evidence and are not universal guarantees.

• Defra AUK 2025, chapter 8, updated 3 September 2026: production, prices, trade and breeding populations. Volumes are UK; meat prices are GB proxies; annual averages are not current quotes. Rounding can prevent exact reconciliation.
• FAS / SAC Farm Management Handbook 2025/26: budgeting context cited by the original LP. The inherited coefficients are source-informed planning assumptions, not independently re-extracted handbook values.
• Defra/APHA BTV-3 vaccination guidance: product-specific protection and timing must be checked with your vet.
• Defra/APHA prevention guidance: context for control scenarios.
• Official situation: premises observations are not model infection probabilities.
• Welsh BTV-8 impact assessment, 17 July 2026: economic costs of controls depend on counterfactuals and scope; historical modelled figures are not imported as measured 2026 losses.

Government subsidy and compensation defaults of zero mean “no transfer in this scenario”, not a verified claim that no payments exist. Missing measured damage is not treated as zero disease loss.

Interpreting and reporting results

Download the assumptions and complete outputs for each selected run. First agree the geography, population, period, price basis and intervention definition. Then discuss changes against the same without-BTV reference, distinguish cash expenditure from resource cost and welfare, compare infected and uninfected producers, and test the assumptions driving the result. LP, DP and PE totals must not be summed.

The original programmes and these adaptations are research tools. Numerical tests check optimisation, accounting and implementation; they do not establish empirical calibration. Farm-specific budgets, vaccine quotations, independent parameter evidence and peer review are still needed before presenting the outputs as decision-grade estimates. A dated exploratory research note, with assumptions and sensitivity results, is available in Publications.