When COVID-19 swept across the globe, one thing was instantly clear: the virus was the same, but the numbers were not. Death tolls, infection rates, excess mortality and “lives saved” by public health measures varied dramatically depending on who was counting and how. Into this fog of conflicting figures steps “One epidemic, many estimates” (1EME), a research initiative at the London School of Economics and Political Science that asks a deceptively simple question: why do estimates of the same pandemic differ so widely-and what does that mean for policy, accountability and public trust?
Bringing together economists, epidemiologists, data scientists and political analysts, 1EME digs beneath the surface of headline statistics to examine the models, assumptions and interests that shape them.The project does not just compare numbers; it interrogates the machinery that produces them-how data are collected and classified, how uncertainty is handled, and how choices made in spreadsheets ripple through to decisions in cabinet rooms and health ministries.At stake is more than academic precision. Competing estimates have real-world consequences, influencing everything from lockdown timing and hospital funding to international aid and vaccine distribution. By tracing how a single epidemic can generate many “truths,” 1EME aims to illuminate the politics of pandemic measurement-and help build a more transparent, accountable approach to counting in future global health crises.
How fragmented COVID 19 modelling shaped UK pandemic decisions
Inside Whitehall and SAGE meetings, ministers were not guided by a single crystal-clear curve, but by a shifting collage of projections, each grounded in different assumptions about human behavior, viral spread and health system capacity.Epidemiological teams from universities, public agencies and self-reliant consultancies produced competing numbers on cases, hospitalisations and deaths, forcing decision-makers to navigate an evidence landscape that looked more like a patchwork than a unified map. This plurality of models did not merely complicate the science; it reshaped the politics of the pandemic response, influencing what counted as a “reasonable worst case”, which trade-offs were foregrounded and how rapidly measures such as lockdowns or school closures were justified.
As these estimates diverged, they fostered a kind of technocratic triangulation.Officials weighed different projections, sometimes informally ranking them by perceived credibility, methodological transparency or alignment with existing policy instincts. In practice, this meant that:
- Selective uptake of certain models amplified particular risks over others.
- Timing of interventions was influenced by which curves were seen as politically and socially tolerable.
- Public communication drew on simplified narratives that masked internal disagreement.
| Model Source | Key Focus | Policy Effect |
|---|---|---|
| Academic groups | Transmission dynamics | Triggered early warnings |
| Government analysts | Service capacity | Shaped NHS surge planning |
| Independent modellers | Choice scenarios | Fed media and public debate |
Inside the 1EME project comparing competing forecasts and their real world impact
Drawing on a global network of researchers, the project functions as a comparative laboratory for epidemic prediction. Teams submit independent models that differ in data inputs,statistical assumptions and ethical priorities,then these projections are benchmarked against how outbreaks actually unfold. By placing forecasts side by side, the initiative reveals which approaches are most informative under different conditions-early-warning phases, peak transmission, or long-tail decline-and which were strikingly off the mark. This systematic comparison uncovers not only technical strengths and weaknesses,but also how uncertainty was communicated to policymakers,journalists and the public. In practice, the work sheds light on why certain numbers became authoritative while others were sidelined, and what that meant for decisions on lockdowns, school closures and vaccine deployment.
To show how forecast choices shape lived experience,the project traces a line from spreadsheets to hospital corridors and household budgets. Researchers map model outputs to downstream decisions and outcomes,highlighting patterns such as:
- Resource allocation: Bed and ventilator shortages linked to underestimated case peaks.
- Economic disruption: Prolonged business closures following overly pessimistic duration estimates.
- Public trust: Shifts in compliance when predictions are repeatedly revised or contradicted.
- Equity impacts: Disadvantaged groups bearing the brunt of misdirected or delayed interventions.
| Forecast type | Policy reaction | Observed impact |
|---|---|---|
| High-case scenario | Rapid nationwide restrictions | Lower mortality, sharper economic shock |
| Moderate scenario | Targeted local measures | Patchy control, uneven hospital strain |
| Optimistic scenario | Delayed intervention | Faster spread, longer recovery |
What policymakers need from epidemiological models clarity transparency and accountability
When ministers are faced with making decisions that affect millions of lives, they need more than black-box predictions and colourful charts. They need to understand what is being assumed,how confident modellers are,and what trade-offs are at stake. Clear explanations of core concepts – such as reproduction numbers, scenario vs.forecast, or the difference between uncertainty and ignorance – are not academic niceties; they are prerequisites for legitimate public policy. Simple comparative tools help. As a notable example, a short briefing that contrasts model outputs under different behavioural assumptions can do more for informed debate than a 200-page technical appendix, provided the key messages are presented with disciplined clarity and without overstating precision.
Transparency and accountability are equally political as they are technical. Policymakers increasingly expect an open view of data sources, funding streams and governance arrangements, along with a record of how scientific advice influenced actual decisions. This is where initiatives like 1EME can shift practice: by normalising side-by-side comparison of models, documenting disagreements, and inviting scrutiny rather than resisting it. To support this, communication should foreground a small set of policy-relevant questions, such as:
- What decisions is this model designed to inform?
- Which assumptions drive the results most strongly?
- How does this estimate compare with other reputable models?
- What would change the advice tomorrow?
| Policy Need | Model Feature | Practical Output |
|---|---|---|
| Rapid decisions | Simple, well-documented scenarios | One-page briefing |
| Public trust | Open data and code | Accessible repository |
| Parliamentary scrutiny | Auditable methods | Clear methodological note |
Recommendations from LSE researchers to rebuild trust in public health forecasting
Drawing on the lessons of 1EME, researchers at LSE argue that models must be treated as public infrastructure, not black-box products. They recommend mandatory disclosure of core model assumptions, open access to historical performance, and clear separation between scientific outputs and political decisions. To make epidemic projections understandable, they call for layered communication: simple topline messages, followed by accessible technical summaries, and then full documentation for specialists. This should be supported by newsroom-style explainer graphics, press briefings that include independent model reviewers, and a commitment to publish what is unknown as clearly as what is known.
At the heart of the proposals is a shift from ad-hoc crisis modelling to a standing, accountable system of public health forecasting. LSE researchers suggest building permanent multi-model hubs that pool forecasts, standardise data inputs, and publish comparison dashboards in real time. They recommend:
- Open audit trails for model updates and code changes
- Inclusive advisory panels with behavioural scientists, ethicists and community representatives
- Pre-agreed communication protocols that ban anonymous “leaks” of unpublished projections
- Routine simulation exercises with media, ministries and modellers before the next crisis
| Area | Current Practice | LSE Suggestion |
|---|---|---|
| Transparency | Selective release | Open methods & data logs |
| Communication | Technical reports | Layered, plain-language briefings |
| Governance | Informal advisory groups | Independent, standing forecast hubs |
| Accountability | Ad-hoc reviews | Regular public performance audits |
Final Thoughts
One Epidemic, Many Estimates is less about settling on a “right” number and more about exposing the scaffolding behind every figure that shapes public debate. By tracing how assumptions, data gaps and political priorities become embedded in models, the project reminds policymakers, journalists and citizens that numbers are never neutral – especially in a crisis.
As governments prepare for future pandemics, the work coming out of the London School of Economics and Political Science offers a quiet but urgent lesson: accountability in public health does not begin and end with dashboards and daily briefings. It requires opening the black box of estimation itself – who produces the numbers, how they are made, and whose interests they serve.
If 1EME has a single message, it is that transparency is not a luxury to be added once the emergency has passed. It is indeed a condition for trust when it matters most.
