For most of the post-war era, economic forecasting was treated as a science. Central banks built models. Governments issued five-year plans. Investment committees constructed ten-year return projections. The underlying assumption was that the economy, while complex, was ultimately knowable — that with enough data, enough computing power, and enough methodological sophistication, the future could be mapped with reasonable confidence.
That assumption is under serious strain. And 2026 may be the year it breaks entirely.
The failure of the models
The evidence is not subtle. The IMF, the World Bank, the Federal Reserve, and virtually every major private forecaster missed the inflationary surge of 2021-2023 — not by a small margin, but by a factor of several multiples. They missed the speed of the subsequent rate cycle. They missed the resilience of the US labour market. They missed the depth of China's property crisis. And they have consistently overestimated the speed of emerging market recovery from successive external shocks.
This is not primarily a failure of talent or effort. The people running these models are among the most sophisticated economists in the world. The failure is structural. The models were built for a world that no longer exists: a world of stable supply chains, predictable geopolitical relationships, relatively contained capital flows, and slow-moving technological change. Every one of those assumptions has been invalidated.
What has changed
Several forces have made the economy fundamentally harder to predict:
Geopolitical fragmentation. The rules-based international order that provided a stable backdrop for economic activity for seventy years is fracturing. Sanctions, export controls, currency weaponisation, and the decoupling of major economies create discontinuities that linear models cannot capture. A policy decision in Washington, Beijing, or Brussels can reroute global capital flows and supply chains within weeks — faster than any forecasting cycle can respond.
Climate volatility. Physical climate risk is now a first-order economic variable. Crop failures, infrastructure disruption, mass migration, and the insurance industry's retreat from high-risk geographies create demand and supply shocks that are inherently unpredictable in their timing, even if their probability distributions can be estimated. The El Niño cycle of 2023-2024 cost the global economy an estimated $3 trillion. How do you model the next one?
AI-driven productivity discontinuities. Artificial intelligence is beginning to reshape labour markets, corporate cost structures, and productivity dynamics in ways that have no historical precedent. The speed of diffusion, the distribution of gains, and the second-order effects on consumption and investment are genuinely unknown. Adding an "AI factor" to existing models is like adding a new equation to a system where the fundamental algebra has changed.
Reflexivity at scale. Digital financial markets, algorithmic trading, and the instantaneous global propagation of sentiment mean that forecasts themselves move markets. The act of prediction changes the thing being predicted. At sufficient scale, this creates feedback loops that are mathematically chaotic — sensitive to initial conditions in ways that make long-range precision impossible in principle, not just in practice.
What does this mean for investors and businesses?
The honest answer is that it means less certainty, which most people already knew, and a different kind of strategy, which fewer have fully absorbed.
If the future cannot be predicted with confidence, then point forecasts — "GDP will grow 2.4% next year," "the dollar will weaken 8% by Q3" — are not just imprecise but potentially dangerous. They create false confidence. They encourage strategy built on single-scenario planning. And when reality deviates — as it will — the organisations that planned for only one future are the least prepared to adapt.
The alternative is not to abandon forecasting, but to change what we ask of it. The most sophisticated operators in volatile environments have shifted from point forecasting to scenario planning, from quarterly guidance to multi-scenario capital allocation, and from optimising for expected value to optimising for robustness across a range of outcomes.
Scenario planning over point forecasting. Instead of asking "what will happen?" the more useful question is "what are the plausible ranges of outcomes, and what is our strategy in each?" This requires developing genuine expertise in multiple possible futures rather than a refined view of a single most-likely path.
Optionality as strategy. In high-uncertainty environments, the value of preserving future choices — maintaining liquidity, avoiding irreversible commitments, building in decision points — increases substantially. The businesses that have performed best across the volatility of the past five years are disproportionately those that kept their balance sheets flexible and their strategic options open.
Real-time data over lagging indicators. Traditional economic indicators — GDP, CPI, unemployment — are released with lags of weeks or months and are subject to significant revision. High-frequency alternatives — shipping data, satellite imagery, payment flows, job postings — provide a faster, if noisier, signal. The organisations investing in real-time data infrastructure are building a genuine forecasting edge.
Geopolitical intelligence as a core competency. In an era where a single policy decision can reshape an industry overnight, political analysis is no longer a soft complement to economic forecasting — it is central to it. Businesses that treat geopolitical risk as a specialist add-on rather than a core input are systematically underestimating the most important variable in their planning horizon.
The deeper question
There is a more fundamental issue underneath the technical debate about forecasting methodology. Economic prediction serves a social function: it provides the basis for coordination. Businesses invest because they believe they can project demand. Governments plan because they believe they can project revenues and costs. Central banks act because they believe they can project inflation and employment trajectories.
If that coordination function breaks down — if everyone loses confidence in their ability to predict, and therefore loses confidence in acting — the economic consequences could be severe. Keynes called this "animal spirits": the propensity to invest, to hire, to expand, depends not just on calculation but on a basic confidence that the future is manageable.
Maintaining that confidence in an environment of genuine uncertainty is perhaps the most important economic policy challenge of the decade. It requires leaders — in government, in business, in finance — who can communicate clearly about what is known and what is not, who can articulate strategy under uncertainty rather than false precision, and who can build institutions resilient enough to absorb shocks rather than pretending they won't come.
The bottom line
We cannot predict the future with the confidence we once assumed. That is not a counsel of despair — it is an invitation to think differently. The investors, businesses, and policymakers who thrive in the coming decade will not be those with the most accurate point forecasts. They will be those who have built the most robust strategies for navigating a range of futures, who have invested in the capabilities to adapt quickly, and who have cultivated the judgment to act decisively under conditions of irreducible uncertainty.
The age of predictable economics may be over. The age of strategic resilience has begun.