Forecast and Forgotten: The Institutional Habit of Predicting Without Accounting

In October 2022, the Office for Budget Responsibility revised its economic forecasts substantially downward. In doing so, it implicitly acknowledged that earlier projections — some issued only months previously — had been materially wrong. The revision was reported as news. The question of why the earlier forecasts had failed, what modelling assumptions had proved incorrect, and whether the same assumptions remained embedded in subsequent projections received considerably less attention. The OBR, to its credit, publishes more methodological transparency than most comparable bodies. But even there, the gap between issuing a forecast and systematically accounting for its accuracy is wider than it should be.

This is not a story about one institution or one forecast. It is a story about a structural feature of British public life: the persistent absence of longitudinal accountability for predictive claims.

The Asymmetry of Institutional Memory

British institutions are, in many respects, well-equipped to generate predictions. The Treasury, the NHS, the Home Office, Ofsted, the Climate Change Committee, and dozens of arm's-length bodies collectively produce an enormous volume of forward-looking analysis each year. Economic projections, epidemiological models, educational attainment forecasts, crime reduction targets — the infrastructure for prediction is substantial and well-resourced.

The infrastructure for retrospective evaluation is considerably thinner. Post-implementation reviews exist in principle across Whitehall, mandated by the Better Regulation Framework. In practice, research by the National Audit Office and academic analysts has repeatedly found that such reviews are conducted inconsistently, published selectively, and rarely designed to interrogate the quality of the reasoning that produced the original policy rather than merely its measurable outcomes. There is a meaningful difference between asking 'did the policy achieve its stated targets?' and asking 'were the causal assumptions that justified the policy correct?' British institutional practice tends strongly towards the former.

Why Prediction Errors Persist Across Policy Cycles

When a forecast proves incorrect and no structured review identifies why, the analytical error does not disappear — it migrates. The same modelling assumptions, the same conceptual frameworks, the same implicit theories of how social systems behave may be carried forward into the next round of policy development, unremarked and unchallenged. This is not a hypothetical risk; it is a documented pattern.

Consider the history of welfare reform predictions in the UK. The introduction of Universal Credit was accompanied by official forecasts about caseload timelines, administrative savings, and claimant behaviour that proved, across multiple dimensions, to be significantly optimistic. The NAO documented these discrepancies in detail. What is harder to establish is whether the Department for Work and Pensions subsequently revised the underlying models — the assumptions about claimant behaviour, digital take-up, and administrative capacity — in ways that were publicly scrutinised and debated. The evidence suggests that institutional learning, where it occurred, was largely internal.

Similarly, NHS England's capacity planning projections have repeatedly underestimated demand growth and overestimated the pace of service transformation. Each cycle of planning tends to acknowledge that previous targets were not met whilst simultaneously issuing new projections built on comparable assumptions. The circularity is rarely named as such in public discourse.

The Academic Sector's Parallel Problem

It would be convenient if this were solely a government problem. It is not. British universities and research institutions face comparable accountability gaps in their public-facing predictive claims. Academic economists, epidemiologists, and social scientists frequently offer projections in media commentary, government advisory roles, and parliamentary evidence sessions. The professional incentive structure rewards the confidence and novelty of predictions rather than their subsequent accuracy.

Philip Tetlock's long-running research on expert forecasting — most accessibly summarised in Superforecasting — demonstrated that domain expertise correlates poorly with predictive accuracy, and that experts who communicate with high confidence tend to perform worse than those who express calibrated uncertainty. These findings are well-known within academic communities. They have had limited impact on the conventions of British expert commentary, where epistemic confidence remains a professional norm rather than a liability.

A small number of institutions have attempted to address this. The Behavioural Insights Team has published retrospective analyses of its own intervention evaluations. Some academic journals now require pre-registration of hypotheses to prevent post-hoc rationalisation of results. These are meaningful steps, but they remain exceptions rather than features of the broader institutional landscape.

What Genuine Predictive Accountability Would Require

For students of public policy and debate, the question of predictive accountability is not merely procedural — it is epistemological. An institution that does not systematically track its own forecasting errors cannot learn from them, and cannot be held to evidential standards by those it advises or governs. The absence of a public record of past predictions and their outcomes is not neutral: it actively insulates poor reasoning from scrutiny.

A serious approach to institutional predictive accountability would involve several elements. First, all significant policy-linked forecasts should be archived in a publicly accessible, searchable format, with the modelling assumptions documented alongside the headline figures. Second, post-implementation reviews should be required to address the accuracy of the causal reasoning that underpinned original predictions, not merely the achievement of output targets. Third, advisory bodies should be expected to publish regular assessments of their own forecasting track records — not as exercises in self-congratulation, but as genuine inputs into the calibration of future advice.

None of this is technically demanding. The data exist. The analytical capacity exists. What is absent is the institutional culture that treats predictive accuracy as a matter of professional obligation rather than a retrospective embarrassment to be managed.

The Debate Worth Having

For debaters and educators, the accountability gap in British institutional forecasting raises a question with genuine normative weight: do citizens have a right to know not merely what their institutions predict, but how reliably those institutions have predicted in the past?

If the answer is yes — and it is difficult to construct a principled argument against it — then the current arrangement represents a significant democratic deficit. Decisions about resource allocation, legislative design, and public health are routinely made on the basis of forecasts whose track records are not publicly visible. The argument for transparency here is not partisan; it applies equally to forecasts that favour any political position. It is simply an argument for the conditions under which evidence-based reasoning can actually function.

Britain has the institutions, the analytical traditions, and the parliamentary mechanisms to make predictive accountability a reality. The question is whether there is sufficient political will to demand it — and sufficient public awareness to make that demand legible.