In 2022, a long list of economists warned that a US recession was nearly certain within a year. Surveys of forecasters put the odds well above 50 percent. Then 2023 arrived. Unemployment stayed low, and the economy kept growing. The recession everyone expected never showed up.

This isn't a story about bad economists. It's a story about what forecasting actually is. When Queen Elizabeth visited the London School of Economics in 2008, she asked a simple question about the financial crisis: why did nobody see it coming? The honest answer is that economies have structural features that make prediction genuinely hard. Understanding those features changes how you should read every forecast.

Every economic forecast starts with a model, and every model is a simplification. An economy with hundreds of millions of people making billions of daily decisions can't be captured directly, so economists make assumptions. People act sensibly. Markets settle into balance. Banks pass money along smoothly. These shortcuts make the math workable, and in ordinary times they do a reasonable job.

The problem is that simplifications aren't neutral. They leave things out, and what they leave out tends to matter most when things go wrong. Before 2008, many of the leading models used by central banks treated the financial system as a kind of plumbing, something that quietly worked in the background. So when that plumbing burst, the models had almost nothing useful to say about it.

Think of a subway map. It's useful precisely because it ignores real distances and street layouts. But if you tried to walk between stations using it, you'd get lost. Economic models work the same way. They describe normal times well, when the things they ignore stay quiet. That means the errors aren't random. They cluster around exactly the moments when we most need accurate forecasts.

Takeaway

A model's blind spots aren't spread evenly across time; they show up together, at turning points. When a forecast feels most confident, it's worth asking what the model has assumed away.

In the 1960s, economists noticed what looked like a reliable pattern: when unemployment fell, inflation tended to rise, and vice versa. This relationship, called the Phillips curve, seemed to offer policymakers a simple menu. Accept a little more inflation, and fewer people would be out of work.

Then the 1970s happened. Governments leaned on that trade-off, and people caught on. Workers and businesses began expecting higher prices and built those expectations into wages and contracts. Inflation climbed while unemployment climbed too, a painful combination called stagflation. The economist Robert Lucas made the deeper point in 1976: historical patterns reflect how people behaved under old conditions. Change the policy, and people change their behavior, so the pattern itself shifts.

This is what makes economics different from predicting the weather. Clouds don't read the forecast. People do. If everyone hears a recession is coming, households may cut spending and firms may delay hiring, helping bring it about. Or a central bank may act early to prevent it, making the forecast look wrong precisely because it was taken seriously. The past is a guide, but it's written by people who keep revising their plans.

Takeaway

Economic data describes how people responded to the world as it was. When the rules change, old patterns can quietly stop applying, sometimes because of the forecast itself.

Most forecasting tools are built on averages and typical ranges. They assume tomorrow will look roughly like the spread of all our yesterdays. That works well enough most of the time, which is exactly why it's dangerous.

In February 2020, US unemployment was 3.5 percent, near a fifty-year low. By April it was nearly 15 percent. No model calibrated on decades of data would have produced that jump, because nothing like it was in the data. The writer Nassim Nicholas Taleb calls these events black swans: rare, enormously consequential, and easy to explain only after they happen. Pandemics, wars, financial panics, and oil shocks all fit the description.

Here's the uncomfortable part. In economics, a handful of extreme events can matter more than all the ordinary years combined. A forecast that's accurate nine years out of ten can still miss the one year that reshapes careers, savings, and government budgets for a decade. Averages smooth those events away. That's why careful forecasters increasingly talk about ranges and scenarios rather than single numbers, and why stress tests ask what happens if things go badly, not just what's most likely.

Takeaway

When rare shocks dominate outcomes, being right on average isn't the same as being useful. Preparing for a range of possibilities matters more than betting everything on the most likely one.

None of this means economic forecasts are worthless. They help us understand the forces in play and plan around the most likely path. But they're estimates built on simplified models, past behavior, and the quiet assumption that nothing extraordinary happens.

So the next time a headline announces where growth or inflation is heading, read it as a weather report for a sky that's listening. Ask what it assumes, how wide the uncertainty is, and what would change the picture.