Rational Expectations

Rational expectations are model-consistent forecasts that use the defined information set without producing forecast errors that are systematically predictable from it.

Rational expectations are forecasts that are consistent with the economic model and the information available when the forecast is made. The assumption allows realized outcomes to differ from forecasts, sometimes substantially, but it rules out forecast errors that are systematically predictable using information already included in the model’s information set.

Rational expectations is a modeling restriction, not a claim that every household, investor, business, or policymaker knows the future. It does not require identical forecasts, perfect information, zero-cost research, instant adjustment, stable markets, or error-free decisions.

Key Takeaways

  • Rational expectations means model-consistent expectations, not perfect foresight.
  • Forecast errors can be positive or negative because new information and genuine shocks arrive after the forecast date.
  • The central restriction is conditional: information available at time t should not systematically predict the next forecast error.
  • A zero average error in a small sample does not prove rational expectations; large offsetting errors can average to zero.
  • The relevant variable, horizon, information set, data vintage, model, and loss function must be specified before testing the assumption.
  • Rational expectations does not imply that all people have the same information or beliefs.
  • It does not make anticipated monetary or fiscal policy universally ineffective; results depend on the full model, including prices, contracts, information, credit, and other frictions.
  • Rational expectations and market efficiency are related in some models but are not the same hypothesis.
  • Survey expectations and market-implied expectations are measurements that require interpretation, not direct proof that the assumption holds.
  • In finance, expectations affect discount rates, yield curves, inflation compensation, exchange rates, credit spreads, earnings forecasts, and asset values.

Formal Definition

Let x_(t+1) be a future economic variable and let I_t be the information set available at time t. A model-consistent forecast is the conditional expectation:

$$ \widehat{x}_{t+1|t}=E\left(x_{t+1}\mid\mathcal{I}_t\right) $$

The realized value can be written as the forecast plus a forecast error:

$$ x_{t+1}=\widehat{x}_{t+1|t}+\varepsilon_{t+1} $$

The rational-expectations restriction is:

$$ E\left(\varepsilon_{t+1}\mid\mathcal{I}_t\right)=0 $$

This is the correct direction of the relationship. The outcome equals the earlier expectation plus an error. The expectation does not equal the realized outcome plus an error known in advance.

The conditional-zero restriction means that a variable already in I_t should not reliably predict the sign or size of epsilon_(t+1) in the population. It does not require every realized error to equal zero.

What the Information Set Includes

The assumption cannot be evaluated until I_t is defined. Depending on the application, it may contain:

  • historical values available by the forecast date;
  • published economic releases and their release vintages;
  • the announced monetary or fiscal policy rule;
  • market prices, interest rates, exchange rates, and spreads;
  • contracts, tax rules, regulation, and institutional constraints;
  • public forecasts or surveys;
  • model parameters and beliefs about how variables interact; and
  • privately observed information when the model allows agents to differ.

“Available” does not necessarily mean free, noticed, understood, or worth acquiring. A model can allow costly information collection and heterogeneous information while still imposing rational expectations relative to what each agent knows and the incentives to learn more.

This boundary matters empirically. A forecast cannot be criticized for failing to use data released afterward. Conversely, if a public indicator was known at the forecast date and repeatedly predicts errors, the stated information set or forecasting model may be incomplete.

Rational Expectations Is Not Perfect Foresight

ConceptForecast errors allowed?Information assumptionMain use
Perfect foresightNo, within the modeled pathFuture outcomes are knownDeterministic benchmark
Rational expectationsYes, from unpredictable shocks and new informationForecast is consistent with the model and defined information setDynamic macroeconomics, policy, and finance
Adaptive ExpectationsYesForecast updates from past outcomes or past forecast errors using a stated ruleInflation and other persistent processes
Extrapolative expectationsYesRecent direction or trend is projected forwardCyclical demand, prices, or sentiment
Survey expectationsYesRespondents report point, range, or probability forecastsDirect measurement of reported beliefs
Market-implied expectationsYesPrices are mapped into expectations through a valuation modelRates, inflation, options, credit, and event probabilities

Adaptive expectations are not automatically irrational. A backward-looking rule can be effective when the process is stable and current information adds little. Rational expectations can also be a poor empirical description if people face limited attention, changing models, costly data, or learning.

Worked Example: Inflation Forecast Errors

Assume a forecaster publishes a one-year inflation forecast using information available on December 31. Define the error as:

$$ e_{t+1}=\pi_{t+1}-E_t(\pi_{t+1}) $$

where pi_(t+1) is realized inflation over the next year. Positive error means inflation exceeded the forecast; negative error means it fell below the forecast.

Forecast yearForecastRealized inflationError: actual minus forecastInformation learned later
Year 13.0%5.0%+2.0 percentage pointsUnexpected supply disruption
Year 24.0%2.0%-2.0 percentage pointsUnexpected demand contraction

The two errors average zero:

$$ \overline{e}=\frac{2.0+(-2.0)}{2}=0.0 $$

That does not prove rational expectations. The sample is tiny, the errors are large, and the explanations have not been tested. An analyst should ask whether data known on each forecast date could have predicted the errors and whether the shocks described as unexpected truly arrived afterward.

Suppose every high prior-year energy-price reading is followed by a positive forecast error. If prior energy prices were in I_t, that pattern may indicate that the model underweights persistent energy effects. If the relationship appears only after searching hundreds of variables, it may instead be data mining.

How Forecast Rationality Is Tested

A common test asks whether variables known at time t predict errors:

$$ e_{t+1}=\alpha+\boldsymbol{\beta}'\mathbf{z}_t+u_{t+1} $$

where z_t contains selected variables from the information set. Under the tested version of rational expectations, the joint null is typically:

$$ \alpha=0 \quad \text{and} \quad \boldsymbol{\beta}=\mathbf{0} $$

The test is only as strong as its design. Analysts should check:

  • whether the forecast was recorded before the outcome and later information;
  • whether the realized variable matches the forecast definition and horizon;
  • whether revised data were substituted for the real-time data available to forecasters;
  • whether errors overlap across horizons and therefore create serial correlation;
  • whether the sample spans a policy regime change or structural break;
  • whether the instruments in z_t were selected in advance;
  • whether the forecast is a mean, median, mode, or probability distribution;
  • whether forecasters optimize squared error, absolute error, business usefulness, or another loss function; and
  • whether rejection reflects irrationality, model misspecification, measurement error, or an incorrectly defined information set.

Unbiasedness and efficiency are related but distinct. An average error near zero addresses unconditional bias. Failure of known information to predict errors addresses informational efficiency. A forecast can be unbiased on average yet inefficient because errors remain predictable.

Practical Example: Expectations in a Two-Year Yield

Expectations about future short-term rates help explain longer-term yields, but they are not the only component. A simplified two-year approximation is:

$$ y_{2,t}\approx\frac{i_t+E_t(i_{t+1})}{2}+TP_{2,t} $$

where:

  • y_(2,t) is the annualized two-year yield;
  • i_t is the current one-year short rate;
  • E_t(i_(t+1)) is the expected one-year rate one year from now; and
  • TP_(2,t) is a term premium.

Assume, only for illustration:

  • current one-year rate: 4.0%;
  • expected one-year rate next year: 3.0%; and
  • two-year term premium: 0.5%.

Then:

$$ y_{2,t}\approx\frac{4.0\%+3.0\%}{2}+0.5\%=4.0\% $$

If new information raises the expected future one-year rate to 4.0% while the current rate and term premium remain unchanged, the approximation becomes 4.5%. The yield can move before the future policy decision occurs because current prices incorporate revised expectations.

This does not prove that the market forecast is rational or that the future rate will be 4.0%. The observed yield also depends on term premiums, liquidity, supply and demand, taxes, compounding, instrument conventions, and model error. See Forward Rate and Term Premium for those distinctions.

Policy Implications and the Lucas Critique

Expectations matter because households and firms can change current behavior when they anticipate a future rule. Borrowers may refinance before a known rule change, firms may alter prices or investment, and investors may reprice assets before a policy is implemented.

The Lucas Critique warns that historical relationships may change when a policy regime changes because people revise expectations and decisions. A reduced-form relationship estimated under one rule should not automatically be used to predict outcomes under a different rule.

That insight does not establish that systematic policy is always ineffective. Policy effects depend on the model, including:

  • whether prices and wages adjust immediately or through contracts;
  • whether information is complete or dispersed;
  • whether households and firms face borrowing or liquidity constraints;
  • whether taxes, regulation, or policy alter real incentives and resources;
  • whether the commitment is credible;
  • whether the action was anticipated; and
  • which horizon and outcome are being evaluated.

The Nobel Prize’s 1995 explanation of Robert Lucas’s contributions describes rational expectations as forward-looking use of available information without the systematic mistakes built into some earlier expectation rules. It also explains why policy evaluation must account for behavior changing when the policy regime changes.

    flowchart LR
	    A["Policy rule and public information"] --> B["Households, firms, and markets revise expectations"]
	    B --> C["Prices, contracts, saving, borrowing, and investment adjust"]
	    C --> D["Observed economic outcome"]
	    D --> E["Forecast error and model evaluation"]
	    E --> F["Update information set or model if errors are predictable"]

Rational Expectations in Finance

Asset valuation

Asset values depend on expected cash flows, discount rates, risk, and state probabilities. Rational expectations can be imposed inside an asset-pricing model, but the assumption alone does not identify the correct valuation model, risk premium, or future price.

Market efficiency

Market Efficiency concerns how information is reflected in prices and whether a specified trading rule earns abnormal returns after risk and cost adjustments. Rational expectations concerns the relationship between beliefs, information, and the model generating outcomes. One does not automatically prove the other.

Even in an efficient market, investors can earn different realized returns, prices can move sharply on new information, and expected excess returns can compensate for risk. Rational expectations does not imply stable or predictable prices.

Inflation and interest rates

Expected inflation influences nominal rates, real-rate estimates, contracts, wages, and valuation. Analysts use surveys and market prices, but no measure is pure. The Federal Reserve’s Index of Common Inflation Expectations combines measures that differ by respondent type, horizon, source, and inflation concept.

Market-based inflation compensation can include inflation expectations, inflation risk premiums, and liquidity effects. Survey measures avoid some market-price premiums but can be stale, unrepresentative, rounded, or affected by question wording.

Corporate and credit decisions

Revenue plans, capital budgets, loan pricing, covenant forecasts, and impairment models depend on expectations. A model-consistent forecast should connect assumptions across inflation, volume, price, wages, rates, defaults, and financing rather than selecting each input independently.

Rational expectations does not make the forecast conservative or prudent. Risk management still requires scenarios, stress tests, limits, liquidity planning, and explicit treatment of tail outcomes.

How Expectations Are Observed

Expectations are latent; analysts observe imperfect proxies.

Evidence sourceWhat it measuresMain limitation
Household or consumer surveyReported beliefs about inflation, jobs, income, spending, or creditQuestion interpretation, rounding, sampling, and limited financial knowledge
Professional forecast surveyPoint or probability forecasts from economists and institutionsConsensus can hide disagreement; forecasts may be stale between releases
Market pricePrice consistent with marginal trades and a valuation modelRisk, liquidity, convexity, collateral, and term premiums can contaminate inference
Business surveyManagement expectations for sales, prices, hiring, or investmentQualitative scales, sector mix, strategic response, and survivorship
Forecast embedded in a planAssumption used for budgeting, lending, valuation, or policyMay reflect conservatism, incentives, constraints, or approval targets rather than mean belief
Realized actionSaving, hiring, inventory, borrowing, hedging, or pricing choiceAction combines beliefs with preferences, constraints, and contracts

The Federal Reserve Bank of Philadelphia’s Survey of Professional Forecasters publishes forecast releases, historical series, dispersion, and individual anonymized responses. The Federal Reserve Bank of New York’s Survey of Consumer Expectations measures household expectations for inflation, labor markets, and household finance. These are evidence about reported expectations, not declarations that respondents satisfy a particular model.

How to Analyze a Rational-Expectations Claim

  1. Define the variable. Specify inflation measure, policy rate, earnings, default rate, exchange rate, or other outcome.
  2. Fix the forecast date and horizon. A one-quarter forecast and a ten-year average are different objects.
  3. Record the information set. Preserve releases, market data, rules, and documents available at the time.
  4. State the model. Explain how information maps into the conditional expectation.
  5. Define the forecast statistic. Mean, median, mode, range, and probability distribution are not interchangeable.
  6. Use the correct realization. Match units, seasonal adjustment, annualization, and release vintage.
  7. Calculate errors consistently. State whether error is actual minus forecast or the reverse.
  8. Test bias and predictability. Examine averages, serial dependence, and variables known at the forecast date.
  9. Check structural breaks. Separate errors caused by a changed regime from stable-model performance.
  10. Compare alternatives. Test against naive, adaptive, survey, and market-implied benchmarks.
  11. Evaluate economic significance. Statistical rejection may have little decision impact, while rare large errors may dominate risk.
  12. Document uncertainty. A point forecast without a distribution can conceal material tail outcomes.

Common Mistakes and Limitations

  • Calling rational expectations perfect foresight: Unexpected shocks can produce large realized errors.
  • Saying forecasts are correct on average without conditioning: Zero unconditional mean error is weaker than no predictable error given the information set.
  • Assuming everyone has identical information: Models can include heterogeneous and costly information.
  • Treating the true model as known: Model uncertainty and learning are central practical limitations.
  • Declaring all systematic policy ineffective: Neutrality or ineffectiveness results require additional assumptions and a specified horizon.
  • Equating rational expectations with market efficiency: They restrict different objects and require separate tests.
  • Reading a forward rate as a pure forecast: Term, risk, liquidity, collateral, and other premiums can create a wedge.
  • Ignoring data revisions: A forecast must be evaluated against definitions and vintages available at the appropriate dates.
  • Treating consensus as certainty: A mean forecast can hide substantial disagreement and individual uncertainty.
  • Using later information in a backtest: Revised data and hindsight can make a historical model appear more informed than it was.
  • Inferring beliefs directly from actions: Preferences, constraints, hedging needs, regulation, and incentives also affect choices.
  • Assuming no predictable error means a useful forecast: An unpredictable forecast can still be too imprecise for a decision.

Authoritative Sources

These sources describe specific models, surveys, or measurement methods. Forecast definitions, releases, samples, policy regimes, and market premiums must be checked for the decision being analyzed.

  • Expectations: Beliefs or probability assessments about future economic and financial outcomes.
  • Adaptive Expectations: Forecasts updated from past outcomes or past errors using a stated adjustment rule.
  • Lucas Critique: Warning that behavior and estimated relationships can change when a policy regime changes.
  • Market Efficiency: Degree to which information is reflected in prices under a specified test.
  • Forward Rate: Rate implied for a future interval by current term-structure prices.
  • Term Premium: Positive or negative wedge between a longer-term yield and the expected result of rolling shorter-term positions under the stated model.
  • Expected Return: Probability-weighted return used in investment and portfolio analysis.
  • Behavioral Economics: Empirical and theoretical study of decision processes that can differ from simple benchmark assumptions.

FAQs

Do rational expectations mean forecasts are always correct?

No. Forecasts can be wrong because unexpected information and shocks arrive. The assumption restricts systematic predictability of errors using information available when the forecast was made.

Do rational expectations require everyone to use the same forecast?

No. Agents can have different information, costs, models, and private signals. Aggregation and the exact equilibrium concept determine how those differences enter a particular model.

How can rational expectations be tested?

Analysts can test average bias and whether variables known at the forecast date predict later errors. Results depend on the information set, horizon, data vintage, error definition, sample, and assumed loss function.

Are market-implied expectations rational expectations?

Not automatically. Market prices contain information but also risk, term, liquidity, collateral, and other premiums. Extracting an expectation requires a valuation model, and the resulting estimate can still be wrong.

Does rational expectations theory say monetary policy never works?

No. Particular models can limit the effects of anticipated policy on particular real variables, especially over longer horizons. Policy can still matter through real incentives, sticky prices, credit constraints, information, credibility, and unexpected components.

This article provides general economic and financial education. It does not forecast inflation, interest rates, markets, or policy and does not provide individualized investment, trading, lending, tax, legal, or regulatory advice.

Browse Economics