Bayesian Inference

Bayesian inference updates probabilities or parameter beliefs by combining prior information with observed data.

Bayesian inference is the general method of learning from data by starting with a prior belief and revising it after seeing evidence.

$$$$

The pieces

Bayesian inference works with four core parts:

  • a prior for what was believed before the data,
  • a likelihood for how probable the observed data are under different hypotheses,
  • a posterior after updating,
  • and often a predictive distribution for future observations.

The update follows Bayes theorem:

$$ \text{Posterior} \propto \text{Likelihood} \times \text{Prior} $$

Economic relevance

Economists use Bayesian inference when agents or researchers must revise beliefs about uncertain quantities such as inflation persistence, growth trends, default probabilities, or the effect of a policy.

It also mirrors how many economic models describe learning: people do not restart from zero every period, they update from what they already believed.

Bayesian vs. frequentist framing

The Bayesian question is usually, “given the data, how plausible is each parameter value or hypothesis?” That differs from the classical frequentist emphasis on sampling distributions and repeated-sample properties.

Both approaches are useful. Bayesian inference is especially attractive when prior information is economically meaningful or when the researcher wants direct probability statements about parameters.

Knowledge Check

### What is the basic idea of Bayesian inference? - [x] Update beliefs with data instead of treating each new sample in isolation - [ ] Replace all uncertainty with certainty - [ ] Measure only historical averages - [ ] Ignore prior information completely > **Explanation:** Bayesian inference starts from a prior and revises it when evidence arrives. ### What role does the likelihood play? - [x] It measures how compatible the observed data are with different hypotheses or parameter values - [ ] It sets tax rates in a fiscal model - [ ] It replaces the prior automatically - [ ] It is the same thing as the posterior > **Explanation:** The likelihood tells the researcher how strongly the data support different possibilities. ### When is Bayesian inference especially useful in economics? - [x] When prior information matters or the model is complex and uncertainty needs to be described explicitly - [ ] Only when there is no data at all - [ ] Only for accounting identities - [ ] Only for perfect-information models > **Explanation:** Bayesian methods are valuable when the researcher wants full probability statements and a disciplined way to combine theory, prior evidence, and new data.