Chapter 2 — When a Forecast Becomes a Promise

When a Forecast Becomes a Promise

In late 2022, the owner of a small manufacturing company began preparing for a recession that had not yet arrived. Her company fabricated metal components for agricultural and industrial equipment. It employed about eighty people in a Midwestern town where a canceled shift traveled quickly—from the production schedule to the diner, the school fundraiser, and the local bank.

Orders were still coming in, but almost every signal around her pointed toward caution. Interest rates were rising. Economists warned that bringing inflation down could weaken the labor market. Business headlines repeated the possibility of recession. Her lender wanted to know how the company would respond if demand fell. So she did what responsible owners are expected to do when the future becomes uncertain.

She renegotiated the company’s credit line before borrowing costs rose further. She ordered extra specialty steel because lead times remained unpredictable. She delayed the purchase of a new cutting machine. Most consequentially, she postponed adding a second shift.

The recession did not arrive in the form she had prepared for. Employment remained stronger than many historical models would have suggested. Consumers continued spending. Financial stress appeared, but the broader system did not collapse. The company’s orders slowed unevenly rather than disappearing. By the time the owner felt confident enough to reconsider the expansion, the equipment cost more, financing had become less attractive, and several workers she might have hired had gone elsewhere.

The warning had not been meaningless. The risks had been real. But the forecast had still entered the world and changed it.

This owner is a composite, assembled from common decisions made by firms during the post-pandemic adjustment: delayed hiring, precautionary inventory, postponed investment, and renewed attention from lenders. No forecast had named her company. No economist had promised that her second shift would fail. The public message had been more careful than that. The public memory was not.

A forecast becomes something like a promise when people reorganize their lives around it.

The Sentence That Begins With If

Forecasts are not prophecies. They are conditional maps. An economic forecast might estimate what will happen if interest rates follow a particular path, if energy prices remain within a certain range, if supply disruptions ease, if household spending slows, or if geopolitical shocks do not multiply. Those conditions are not technical decorations. They are part of the claim.

Central banks communicate uncertainty through ranges, projections, confidence bands, and multiple possible paths. The Federal Reserve’s Summary of Economic Projections, for example, does not present a single institutional prediction delivered with one voice. It collects estimates from policymakers whose expectations differ and whose projections depend on assumptions about how conditions will develop.¹

The disagreement is visible in the chart. It is less visible in the headline.

A forecast might begin as:

If inflation remains persistent and policy must stay restrictive, economic activity may slow enough to produce a meaningful rise in unemployment.

By the end of the news cycle, it becomes:

Recession risk is rising.

After repetition, it becomes:

A recession is coming.

Nothing dishonest has necessarily happened at any individual step. Each version is shorter than the one before it. Each removes detail so the claim can travel farther. The first sentence describes a relationship among conditions. The last describes an event waiting on the calendar.

This is the compression problem extended into time. A current statistic reduces millions of present conditions into a number. A forecast reduces millions of possible futures into a path. The present is already difficult to summarize. The future adds uncertainty, feedback, adaptation, and human response.

A forecast can change behavior before the forecasted event occurs. Businesses delay investment. Consumers postpone purchases. Banks adjust lending standards. Investors move money. Governments prepare emergency measures. These responses can make the forecast more likely, less likely, or simply different from the future that would have unfolded without the warning. The map enters the territory.

This is one reason economic forecasts are so difficult to judge after the fact. A hurricane forecast can be compared with the storm that eventually makes landfall. An economic forecast may help produce the conditions against which it is later evaluated. If a warning leads institutions to prepare successfully, the feared outcome may not occur. If a warning damages confidence, it may contribute to the slowdown it predicted. If people hear the warning but not its conditions, they may remember a promise no forecaster believes was made.

The difference between if and when is only one word. It can contain an entire theory of responsibility.

From Probability to Public Memory

In a conference room, uncertainty can remain visible. An economist can display several possible paths, explain the assumptions beneath them, and describe the conditions that would cause the forecast to change. The audience can ask questions. A range can remain a range.

Public communication works differently. A news alert cannot reproduce a probability distribution. A television discussion needs a direction. A business owner scanning headlines before a morning shift needs to know whether to expand, wait, borrow, or preserve cash. Political actors need language that can establish competence or blame. Under those pressures, a forecast is valued partly for the orientation it provides: something is coming, someone sees it, and someone is prepared to act.

This is where risk communication begins to slide toward signaling. Officials need to signal that they recognize danger. Central bankers may need markets to believe they are serious about inflation. Regulators may need banks to prepare for stress. Elected leaders may need the public to understand that difficult policy choices are not being made casually. Seriousness is itself an instrument. But once seriousness is performed through confident language, later revision becomes socially expensive.

Suppose an official says that recession risks are elevated. The statement is conditional and defensible. Markets adjust, news organizations emphasize the warning, and businesses prepare. Six months later, employment remains strong and the probability of recession falls. The official updates the forecast.

Inside the institution, this is what responsible forecasting looks like. Evidence changed, so the map changed. Outside it, the update can sound like retreat: you said a recession was coming; now you say the economy is resilient; why should we believe the next thing?

The public is not necessarily rejecting the idea that evidence can change. People update expectations constantly in their own lives. What creates distrust is the difference between how uncertainty was communicated before the revision and how the revision is explained afterward. Beforeward, the forecast often sounded like orientation. Afterward, it is defended as conditionality. The confidence traveled farther than the conditions did.

Late 2022 made the mechanism unusually clear. In the Philadelphia Fed’s Survey of Professional Forecasters, the mean probability of a negative quarter approached fifty percent for several quarters of 2023—high enough that caution looked prudent, and high enough that later headlines about resilience would be measured against warnings already lived through decisions.² The survey did not speak with one voice, and not every forecaster expected the same path. That is part of the point. A distribution still entered public memory as a direction. Firms, households, and lenders prepared against a future that many experts treated as likely enough to matter.

The forecast becomes memorable precisely where it was most compressed. People may never see the fan chart, probability range, or assumptions. They remember the sentence that reached them while they were deciding whether to hire someone, buy a home, refinance debt, or keep more money in reserve. The forecast enters memory not as a model but as advice: prepare; wait; do not risk it. A technically careful institution can therefore spend credibility through a claim it never technically made.

The Costs of a Future That Does Not Arrive

It is tempting to treat a forecast that did not materialize as harmless—nothing happened, the warning passed, the economy adjusted. But preparing for a possible future has costs even when preparation is prudent.

The manufacturer who delayed a second shift avoided the risk of hiring into a downturn. She also gave up production capacity, delayed income for workers she might have employed, and lost time she could not recover once conditions became clearer. A household that postponed buying a home may have avoided purchasing at the top of a local market; it may also have encountered higher rates later. A bank that tightened lending standards may have protected its balance sheet; it may also have denied credit to viable firms whose owners were already hearing that trouble was ahead.

None of these decisions proves that the forecast was wrong. Decisions made under uncertainty cannot be evaluated only by the outcome that happened to occur. Buying insurance is not foolish merely because the house did not burn down. But insurance has a visible premium. Forecast preparation often does not.

Its costs appear as absences: the employee never hired, the machine never purchased, the store never opened, the move never made. These absences do not appear together in a single national statistic. They remain distributed among households and firms, each too small to become the public story. The manufacturer who delayed the second shift could later say employment remained strong nationally. That did not put the postponed shift on the shop floor or restore the year of production she did not attempt.

When the worst outcome is avoided, institutions naturally emphasize resilience. The labor market held. The financial system absorbed the strain. The feared recession did not arrive. These statements can all be true.³ They do not reverse the decisions made while the warning was active. Continuity and invisible protection still leave costs; what the system holds back is not the same as what a household or firm is free to attempt.

This creates a peculiar credibility problem. The institution evaluates the forecast against the range of outcomes that might have occurred. The public evaluates it against the life reorganized in response. The institution sees avoided risk. The household or firm sees a cost already paid. Both are looking backward. They are remembering different events. A later speech celebrating soft landing can therefore sound accurate to the room that owns the chart and premature to the room that owns the unused equipment proposal.

For the economy as a system, the landing may have been softer than many alternatives. For a household, the landing might still have included higher rent, more expensive credit, a postponed move, and prices that remained elevated after the rate of inflation declined.⁴ A lower rate of deterioration is meaningful in a model and unsatisfying in a kitchen. Households do not pay the rate of change. They pay the resulting price. When public language collapses that distinction into the economy is strong, so you should feel better, people who do not feel better hear an accusation.

Revision and Conditional Credibility

Forecasts must be revised. A model that refuses to change when evidence changes is not more credible. It is simply rigid.

Employment data are revised as additional reports arrive. Inflation expectations change with energy prices, supply conditions, wages, and policy. Financial risks shift when institutions alter their behavior. Forecasting is an ongoing process of updating uncertain beliefs, not a one-time act of prediction.⁵

Yet revision has a social meaning beyond its scientific one. Inside a technical institution, revision can signal seriousness: we noticed that conditions changed. Outside it, revision can signal unreliability: you did not know what you claimed to know. The problem becomes worse when institutions communicate in long periods of confidence followed by sudden changes.

For months, the official story appears stable. Then a new forecast, revised projection, or policy turn arrives. The institution describes the change as an ordinary response to new data. The audience experiences it as a rupture. Why were the previous qualifications not more visible? What changed, and when did you know? Were you wrong, or are you protecting yourself now? These questions are not always fair. They are also not irrational.

People judge forecasts partly by how updates are handled. A leader who presents every current estimate as the most likely interpretation while openly naming what would change it creates a record of conditionality. A leader who performs confidence until the evidence becomes overwhelming creates a record of reversal. Small revisions can look like maintenance. Large, delayed revisions look like confession.

This creates an institutional temptation to avoid updating in public. If every adjustment will be called a failure, it may seem safer to hold the narrative steady until change can no longer be denied. That choice protects credibility briefly and damages it more deeply. The longer certainty is performed, the more revision resembles betrayal. A forecast can survive being wrong more easily than it can survive appearing dishonest about uncertainty.

Conditionality can be said in ordinary language: this is the path we currently consider most likely; these are the conditions that could move us away from it; this is what we will watch; this is when we expect to update the assessment. Such language does not eliminate misunderstanding. It gives the audience a way to interpret change without treating every revision as proof of incompetence or deception.

The manufacturer eventually understood this in a smaller vocabulary. Inside her company, a revised order book was maintenance. Outside it, a revised recession risk had already been spent as caution. She did not need an economist to explain that a probability could change. She needed language that left room for the update before the unused machine became the proof that someone had been wrong.

Large institutions often know more about variation than their public messages reveal. The Federal Reserve’s Beige Book collects reports from different districts about employment, wages, prices, credit, and business conditions.⁶ Inside the institution, the aggregate and the anecdote coexist. Public communication tends to privilege the model because the model offers a single direction. The problem is not that the nuance was absent. It is that people often discover it only after the headline fails their experience. A manufacturer hears that supply chains have normalized while still waiting for a specialty component. A national risk can fall before a local order book recovers. These are not caveats attached to the economic story. They are the paths through which the story becomes real.

The same two-clock discipline matters even more when speaking about the future. One tracks the national path: recession probability, projected inflation, expected unemployment. The other tracks decisions already being made in response: delayed hiring, tightened credit, canceled projects, precautionary savings. The second clock starts before the forecast is resolved. A national forecast should not be judged only by whether its headline outcome arrives. The forecast also becomes part of the environment in which households and firms act.

The Machine That Was Never Purchased

Months after the most urgent recession warnings had faded, the manufacturer returned to the equipment proposal she had postponed. The new machine still made sense operationally. It would reduce waste, increase capacity, and allow the company to bid on work it currently sent elsewhere. But the calculation had changed. The financing cost was higher. The supplier’s price had increased. The workers she might have trained were no longer necessarily available. Customers remained cautious enough that the additional capacity felt harder to justify.

She sat with the same binder of vendor quotes she had closed the year before. The quotes were out of date. The second-shift plan was still scribbled in the margin of an older staffing sheet. Nothing in the room announced a national recession. What announced itself was the cost of having treated a high-probability warning as advice.

She could not point to one forecast and say it caused the project to disappear. Inflation, interest rates, demand, supply conditions, and her own risk tolerance all mattered. That is how economic meaning usually moves—not through a single false statement, but through an accumulation of signals that alter what feels prudent. The Survey of Professional Forecasters had recorded a distribution; her equipment schedule had recorded a decision. Public memory often collapses those into one verdict: you warned us, and then you said we were fine.

Institutions had spoken in probabilities. She had reorganized the company around conditions she heard as a promise. The recession had not arrived as a national event. Its possibility had still shaped her payroll, her inventory, and the machine that never moved from the proposal page to the shop floor.

Footnotes

1. Board of Governors of the Federal Reserve System, Summary of Economic Projections and related Federal Open Market Committee materials, 2022–2024. ↩

2. Federal Reserve Bank of Philadelphia, Survey of Professional Forecasters, Fourth Quarter 2022 (released November 14, 2022), reporting survey-mean probabilities of a negative quarter near fifty percent for several quarters of 2023. https://www.philadelphiafed.org/surveys-and-data/real-time-data-research/spf-q4-2022 ↩

3. U.S. Bureau of Labor Statistics, The Employment Situation, 2022–2024; Board of Governors of the Federal Reserve System, Financial Stability Report, 2022–2024. ↩

4. Board of Governors of the Federal Reserve System, Report on the Economic Well-Being of U.S. Households, 2022–2024. ↩

5. Philip E. Tetlock and Dan Gardner, Superforecasting: The Art and Science of Prediction (New York: Crown, 2015). ↩

6. Board of Governors of the Federal Reserve System, Beige Book: Summary of Commentary on Current Economic Conditions by Federal Reserve District, 2022–2024. ↩