What Does Prevention Actually Buy? Mass Dog Vaccination and Rabies in Ethiopia

A worked case through the OHS Prevention Ledger. Vaccinating Ethiopia’s dogs against rabies looks worth twelve times its cost in the central case — and 0.6 times in the least favourable one. The range, not the ratio, is the finding.

Why rabies, and why Ethiopia

Rabies is the textbook case for prevention at the animal source. Almost all human cases come from dog bites; the disease is almost invariably fatal once symptoms appear; and it is entirely preventable, both by vaccinating people after exposure and by vaccinating dogs so that exposure does not happen. The global Zero by 30 plan, adopted by WHO, FAO, WOAH and the Global Alliance for Rabies Control in 2018, aims to end human deaths from dog-mediated rabies by 2030, and treats mass dog vaccination as its core intervention.

It is also a case where the economics are unusually well studied. That makes it a good first test for the OHS Prevention Ledger, which this issue introduces. If the Ledger can show the working clearly on a well-evidenced intervention, it should be useful elsewhere. If it produces a misleading answer here, that would say something about the method.

Ethiopia is one of the countries with the highest rabies burden in the world, a national elimination strategy running to 2030, and — through a series of studies led by Tariku Jibat Beyene, Henk Hogeveen and colleagues — a body of Ethiopia-specific evidence on burden, costs and the cost-effectiveness of dog vaccination in particular districts. What it does not yet have is sustained national dog vaccination. Coverage is reported to be below 20% in urban areas and close to non-existent in much of the countryside.

This article asks a simple question of that evidence: what would a national programme actually buy?

What this is, and what it is not

The analysis below is an illustrative national scaling of published district-level evidence. It is not a costed government plan, and it is not a new epidemiological model. It takes figures from the literature, states where each comes from, classifies the evidence behind it, and runs them through the Ledger’s chain from investment to economic consequences. Where the evidence is weak, it says so. The completed workbook is published alongside this article so that every number can be checked and changed.

Three simplifications matter. First, the Ledger uses average effects over the ten-year horizon rather than modelling transmission year by year, so the ramp-up period is absorbed into the average reduction. Second, sources span several years and are not adjusted to a single price year. Third, costs borne by dog owners — the time taken to bring a dog to a vaccination point — are not included. Each would matter in a full economic evaluation. None changes the shape of the result.

Step 1: Investment

Ethiopia’s dog population is not counted. The most commonly cited national estimate, 11.7 million, is an extrapolation from human-to-dog ratios. Vaccinating 70% of that population each year means about 8.2 million dogs.

The cost per dog vaccinated is the least certain input on the cost side. Ethiopian studies report campaign totals for particular districts but not a reliable national unit cost; one policy paper gives a national estimate of about US$17.5 million a year, which implies roughly US$2 per dog vaccinated, but the same paper also quotes a figure of US$0.20 per dog that is inconsistent with it. We therefore used a range drawn from comparable settings:

Low scenario (least favourable): US$5.00 per dog — the upper end of fully costed fixed-post campaigns in N’Djamena, Chad.

Central: US$2.75 — recent community-led and team-led campaigns in Mara, Tanzania.

High: US$2.00 — the implied Ethiopian national estimate.

Adding a modest set-up cost for training, cold chain and surveillance (assumed at US$1–3 million), and discounting at 3% over ten years, gives a present-value cost of US$194 million in the central case, US$141 million in the high scenario and US$352 million in the low.

Step 2: Capability

The capability the money buys is dog vaccination coverage. The widely used target is 70%, a figure that rests on work by Coleman and Dye in 1996 showing that 70% coverage would prevent a major outbreak on at least 96.5% of occasions. Estimates of rabies’ basic reproduction number in dogs are low — around 1.2 in rural Tanzania and below 2 elsewhere — so the true critical threshold is lower, perhaps 40–60%. But dog populations turn over quickly, so coverage decays within a year and campaigns have to be repeated annually.

That is the capability’s defining weakness. It does not last. A campaign that achieves 70% one year and lapses the next buys very little. The Ledger records this in its capability sheet: coverage is the link between spending and risk, and it has to be bought again every year.

Step 3: Counterfactual and risk change

Rabies in Ethiopia is endemic, not episodic. Without a national programme, the burden continues more or less as now. In Ledger terms, the probability of the “event” is 100% with and without the intervention, and what changes is its size.

The central estimate of the current burden comes from Beyene and colleagues’ 2018 study, which traced dog-bite victims in three districts — urban, rural highland and rural lowland — and extrapolated nationally: about 3,000 human deaths a year, around 194,000 disability-adjusted life years, and 97,000 exposed people needing PEP. The national rabies strategy cites up to 2,700 deaths a year. We used 3,000 as the central value, 2,000 as the low and 4,000 as the high, the last reflecting population growth since the district data were collected.

The contrast with official surveillance is striking. Between 2018 and 2022, Ethiopia’s passive surveillance system recorded an average of 59 human rabies deaths and about 7,600 suspected exposures a year. The gap between 59 recorded deaths and an estimated 3,000 is a reminder of the flagship’s point about invisible returns: a burden that is barely recorded is a burden whose reduction would barely be recorded either.

For the effect of vaccination, the strongest observed evidence comes from Tanzania, where campaigns reaching 61–74% of dogs in the Serengeti district reduced dog rabies by 70% after the first campaign and 97% after the second. In Ethiopia, Beyene and colleagues’ 2019 dynamic model of two districts found vaccination at 70–80% coverage to be the optimal strategy. We assumed an average reduction in dog-mediated rabies over the ten years of 70% in the central case, 90% in the high and 40% in the low, the last reflecting incomplete coverage, patchy delivery and reintroduction from unvaccinated areas and neighbouring countries.

We rated counterfactual confidence as Moderate. The burden is empirically estimated; the effect of sustained national campaigns in Ethiopia has not yet been observed.

Step 4: Outcomes

Over ten years, the central case implies:

Human deaths: 30,000 without the programme and 9,000 with it — 21,000 deaths potentially avoided (8,000 in the low scenario, 36,000 in the high).

PEP courses: 970,000 without and about 825,000 with the programme.

Dog rabies: falling to about 30% of its current level on average.

The PEP figure deserves explanation, because it is where intuition most often goes wrong. Vaccinating dogs does not stop dogs biting people, and bite victims — rightly — continue to seek PEP whether or not the biting dog was rabid. Evidence from Chad shows that PEP demand persists long after dog rabies falls. We therefore assumed PEP use falls by only 15% in the central case, 5% in the low and 40% in the high, the last assuming Ethiopia adopts integrated bite case management, in which the biting dog is assessed and PEP targeted accordingly.

Step 5: Economic consequences

Three consequences were monetised.

PEP costs. Beyene and colleagues found the cost of a sufficient PEP course in Ethiopia to be US$23 in an urban district and US$31–40 in rural districts, most of it borne by households in travel and lost time. These are low by international standards, partly because Ethiopia has long relied on an inexpensive, locally produced nerve-tissue vaccine. WHO recommends replacing nerve-tissue vaccines with modern cell-culture vaccines; as that happens, PEP will become more expensive, and the savings from reducing it larger. We used US$23, 30 and 40.

Cattle losses. Rabies also kills livestock bitten by rabid dogs. A household survey by Jibat and colleagues found an annual rabies incidence of about 2% in cattle and average losses of around US$50 per herd per year. A national figure of US$210 million a year is cited in the policy literature, but we could not trace its calculation. We used it as the high value, US$100 million as central and US$25 million — an eighth of the cited figure — as low, and classified the evidence as transferred.

Deaths. Valuing lives is a normative choice, not an empirical finding, and different conventions give very different answers. We followed the 2019 reference-case guidelines for benefit–cost analysis in global health by Robinson, Hammitt and colleagues, which suggest valuing a statistical life at between 20 and 160 times gross national income per capita, with 100 times as a common central value. Ethiopia’s income per head was about US$930–1,130 in 2024–25; rounding to US$1,000 gives values of US$20,000, US$100,000 and US$160,000. We did not also count lost productivity, which is already reflected in the value of a statistical life; counting both would double-count.

The Ledger weights each consequence by the change in the burden, discounts it over ten years, and sums the included rows.

The results

Four things stand out.

The central case is strong. A programme that costs US$194 million and is expected to avoid consequences worth US$2.4 billion is, on the face of it, very good value. The result is consistent with Beyene and colleagues’ district-level finding that dog vaccination is cost-effective in Ethiopia and, in rural areas where cattle losses are counted, cost-saving.

The range is enormous. The Prevention Ratio runs from 0.6 to 46.5 — a factor of more than seventy. The low scenario is not implausible in any single assumption; it simply combines every unfavourable one at once: the lowest burden, the smallest effect, the highest cost per dog, the lowest value of a life and the smallest cattle losses. The high scenario combines every favourable one. Quoting the central ratio alone would conceal that the answer depends heavily on assumptions that nobody can currently verify.

Almost nothing is money saved. On PEP savings alone, the programme repays about 2% of its cost. This matters, because the most intuitive argument for dog vaccination — that it pays for itself by reducing the need for human vaccines — does not hold in Ethiopia’s current circumstances, where PEP is cheap and demand persists. Evidence from Chad points the same way. There, PEP costs society far more — about US$198 a course — yet comparing dog vaccination plus PEP with PEP alone, break-even took an estimated six years in an early model and ten to fifteen years in a later empirical study. The case for dog vaccination in Ethiopia is a case about lives and livestock, not about health-system savings.

The evidence is thin exactly where the value is. The Ledger reports that essentially none of the central avoided value rests on observed or empirically estimated evidence. About three quarters rests on modelled values — the deaths avoided, combining an empirically estimated burden with a modelled effect and a conventional valuation — and a quarter on the transferred national cattle-loss figure. The burden is reasonably well documented; the effect of national campaigns in Ethiopia and the scale of cattle losses are not. The Ledger flags this, correctly. It does not mean the programme is a poor investment. It means the most valuable next step is evidence.

Who pays, and who benefits?

In Ethiopia, dog vaccination is delivered through regional veterinary offices under the livestock ministry. PEP is delivered through the health system, but most of its cost falls on households. Deaths and cattle losses fall on households and communities.

Allocating the central results accordingly — and assuming, generously, that the health system bears 40% of PEP costs — gives:

Government: pays about US$194 million and receives about US$1.5 million in avoided health-system PEP costs. Its own ratio is about 0.008.

Households and communities: pay nothing directly and receive almost all of the US$2.4 billion.

Even within government, the payer and the beneficiary are different ministries: the livestock ministry funds the campaigns and the health ministry receives the saving. The Prevention Paradox article in this issue describes the gap between the ratio for society and the ratio for the payer as the place where much underinvestment lives. Rabies in Ethiopia shows it at its most extreme: an intervention worth twelve times its cost to society is worth less than a hundredth of its cost to the budget that funds it.

That is why, across Africa, dog vaccination has so often depended on donors, foundations and NGOs. In N’Djamena, campaign costs were split between the national livestock ministry and a Swiss foundation. It is also why Ethiopia’s national strategy — like most national rabies strategies — has struggled to move from pilot zones to national scale.

What would have to be true?

Because the range is so wide, a more useful question than “what is the ratio?” is “what would have to be true for the programme to pay?” The £0 Outbreak article in this issue argues for this kind of break-even analysis. For this case it gives a clear answer.

At the central cost of US$194 million, with the central burden and central values:

the programme breaks even if it reduces dog-mediated rabies by about 6% on average over ten years, counting lives and cattle;

about 8%, counting lives alone;

about 14%, counting cattle and valuing each life at only 20 times income per head;

about 23%, counting cattle losses alone and placing no value at all on lives.

The Tanzanian campaigns reduced dog rabies by 70% after one campaign. Unless national delivery in Ethiopia performs far worse than district campaigns elsewhere, the central-cost programme clears its break-even point comfortably under almost any reasonable valuation.

The low scenario tells a different story. With the highest cost and every other assumption at its least favourable, the programme needs to reduce rabies by about 64% to break even, against an assumed 40%. The question that decides whether the low scenario is realistic is therefore not about valuation. It is about delivery: can Ethiopia achieve and sustain high coverage at a reasonable cost per dog?

What is not in the ratio

Several material effects are recorded in the Ledger but not monetised:

Children. About 60% of human rabies deaths in Ethiopia are in children under 15. Valuing each death equally understates what many people would regard as the true loss.

Fear and psychological burden. Exposure to a possibly rabid dog, and the uncertainty that follows, carries costs that no ratio captures.

Animal welfare. Fewer rabid animals, and less pressure for dog culling as a control measure.

Wildlife. Rabies spillover from domestic dogs is a threat to endangered species, including the Ethiopian wolf.

Distribution. The burden falls hardest on rural and pastoral households far from treatment, where the share of exposed people receiving sufficient PEP in Beyene’s study fell as low as 29%.

Capacity. A national vaccination platform builds veterinary delivery capacity that could support other One Health programmes.

Almost all of these point in the same direction: towards a larger case for the programme than the ratio shows.

What this case teaches about the Ledger

The exercise was designed as a test of the method as much as of the intervention. Three lessons emerge.

The Ledger does what it is meant to do. A headline ratio of twelve would be a persuasive advocacy number. The Ledger refuses to let it stand alone: it shows the range, reports that the value rests largely on modelled and transferred evidence, and exposes the gap between society and the payer. Each of those is more useful to a decision-maker than the ratio.

Endemic diseases need a different reading. The Ledger was designed around events that may or may not happen, but it handles an endemic burden by setting both probabilities to 100% and recording the change in size. The results summary now says so explicitly rather than displaying a zero risk change.

The choice of what to count dominates. Whether lives are valued, and at what level, changes the central ratio from 3 to 12. The Ledger cannot settle that choice. It can make sure the reader knows it has been made.

Implications

Four implications follow for Ethiopia and for programmes like it.

Evaluate as you scale. The single most valuable piece of missing evidence is the observed effect of sustained campaigns on dog rabies and human exposures in Ethiopian settings. Pilot zones under the national strategy should be designed so that effect can be measured — including the surveillance needed to see it, given that current reporting captures perhaps one death in fifty.

Measure the capability. Coverage achieved, cost per dog and the timeliness of annual campaigns are measurable every year. They are the leading indicators that tell a funder whether the programme is on track long before any change in human deaths can be seen.

Don’t sell it on savings. Arguments that dog vaccination pays for itself through lower PEP costs will not survive scrutiny in Ethiopia. The honest case is about lives, livestock and children, and it is a strong one.

Solve the payer problem explicitly. A programme worth twelve times its cost to society and almost nothing to the budget that funds it will not be funded reliably by that budget. Joint financing between livestock and health ministries, donor co-financing tied to measured coverage, and inclusion in One Health financing instruments such as the Pandemic Fund are all ways of bringing the beneficiaries’ interests to the table. The policy analysis in this issue, Insurance Before Emergency, looks at the financing options more broadly.

What would prove this wrong?

If sustained national campaigns in Ethiopia achieved coverage well below 50%, or reductions in dog rabies well below 40%, the low scenario would be closer to reality than the central one, and the case would depend on much cheaper delivery.

If better national data showed the human burden to be substantially below 2,000 deaths a year, the programme’s value would fall proportionately.

And if a careful re-estimate of cattle losses showed them to be far smaller than even our low value, the ratio excluding the value of lives would fall below 1 in all but the most favourable scenario — although the case including lives would remain strong.

Downloads

The completed Ledger for this case — OHS Prevention Ledger: Worked Example — Rabies, Ethiopia — is available with the OHS Prevention Ledger resource. Every input can be changed, and every figure in this article can be traced to the Evidence Register on its eighth sheet.

Questions & Answers

References

1. Asfaw, A. et al. (2024) ‘Epidemiology of suspected rabies cases in Ethiopia: 2018–2022’, One Health Advances. doi.org/10.1186/s44280-023-00036-6
2. Beyene, T.J., Mourits, M.C.M., Kidane, A.H. and Hogeveen, H. (2018) ‘Estimating the burden of rabies in Ethiopia by tracing dog bite victims’, PLoS ONE, 13(2), e0192313. doi.org/10.1371/journal.pone.0192313
3. Beyene, T.J., Fitzpatrick, M.C., Galvani, A.P., Mourits, M.C.M., Revie, C.W., Cernicchiaro, N., Sanderson, M.W. and Hogeveen, H. (2019) ‘Impact of One-Health framework on vaccination cost-effectiveness: a case study of rabies in Ethiopia’, One Health, 8, 100103. doi.org/10.1016/j.onehlt.2019.100103
4. Beyene, T.J., Mourits, M., O’Quin, J., Leta, S., Baruch, J. and Hogeveen, H. (2020) ‘Policy perspectives of dog-mediated rabies control in resource-limited countries: the Ethiopian situation’, Frontiers in Veterinary Science, 7, 551. doi.org/10.3389/fvets.2020.00551
5. Changalucha, J. et al. (2025) Costs of community-led and team-led dog vaccination campaigns in Mara, Tanzania, Frontiers in Tropical Diseases. doi.org/10.3389/fitd.2025.1725443
6. Cleaveland, S., Kaare, M., Tiringa, P., Mlengeya, T. and Barrat, J. (2003) ‘A dog rabies vaccination campaign in rural Africa: impact on the incidence of dog rabies and human dog-bite injuries’, Vaccine, 21(17–18), pp. 1965–1973.
7. Coleman, P.G. and Dye, C. (1996) ‘Immunization coverage required to prevent outbreaks of dog rabies’, Vaccine, 14(3), pp. 185–186.
8. Federal Democratic Republic of Ethiopia (2018) National Rabies Control and Elimination Strategy, 2018–2030 (draft).
9. Hampson, K., Dushoff, J., Cleaveland, S., et al. (2009) ‘Transmission dynamics and prospects for the elimination of canine rabies’, PLoS Biology, 7(3), e1000053. doi.org/10.1371/journal.pbio.1000053
10. Hampson, K., Coudeville, L., Lembo, T., et al. (2015) ‘Estimating the global burden of endemic canine rabies’, PLoS Neglected Tropical Diseases, 9(4), e0003709. doi.org/10.1371/journal.pntd.0003709
11. Jibat, T., Mourits, M.C.M. and Hogeveen, H. (2016) ‘Incidence and economic impact of rabies in the cattle population of Ethiopia’, Preventive Veterinary Medicine, 130, pp. 67–76. doi.org/10.1016/j.prevetmed.2016.06.005
12. Mindekem, R., Lechenne, M.S., Naissengar, K.S., et al. (2017) ‘Cost description and comparative cost efficiency of post-exposure prophylaxis and canine mass vaccination against rabies in N’Djamena, Chad’, Frontiers in Veterinary Science, 4, 38. doi.org/10.3389/fvets.2017.00038
13. Robinson, L.A., Hammitt, J.K., et al. (2019) Reference Case Guidelines for Benefit-Cost Analysis in Global Health and Development. Boston: Harvard T.H. Chan School of Public Health / Bill & Melinda Gates Foundation.
14. WHO, FAO, OIE and GARC (2018) Zero by 30: The Global Strategic Plan to End Human Deaths from Dog-Mediated Rabies by 2030. Geneva: World Health Organization. www.who.int/publications/i/item/9789241513838
15. World Bank (2026) World Development Indicators: GDP per capita (current US$), Ethiopia. data.worldbank.org/indicator/NY.GDP.PCAP.CD?locations=ET
16. Yoak, A.J. et al. (2021) Dog owner willingness to pay for rabies vaccination in Addis Ababa, Ethiopia, Preventive Veterinary Medicine, 187, 105256. doi.org/10.1016/j.prevetmed.2020.105256
17. Zinsstag, J., Dürr, S., Penny, M.A., et al. (2009) ‘Transmission dynamics and economics of rabies control in dogs and humans in an African city’, Proceedings of the National Academy of Sciences, 106(35), pp. 14996–15001. doi.org/10.1073/pnas.0904740106