Autoregressive Conditional Heteroscedasticity (ARCH)

A time-series model in which current volatility depends on past squared shocks.

An ARCH model is a time-series model in which the conditional variance of a series depends on past squared errors, allowing volatility to cluster over time.

$$$$

The basic structure

In an ARCH(1) model:

$$ y_t = \mu + \varepsilon_t,\quad \varepsilon_t = \sigma_t z_t $$

with

$$ \sigma_t^2 = \alpha_0 + \alpha_1 \varepsilon_{t-1}^2 $$

where (z_t) is typically modeled as white noise.

Why economists use it

Financial returns often show volatility clustering: calm periods are followed by calm periods and turbulent periods by turbulent periods. ARCH models capture that behavior much better than constant-variance models.

What the model changes

The mean of the series can stay simple, but uncertainty around that mean becomes time-varying. That matters for forecasting risk, pricing derivatives, and measuring how shocks propagate through markets.

Knowledge Check

### An ARCH model is mainly used to model: - [x] time-varying volatility - [ ] only the sample mean - [ ] perfect competition - [ ] tax incidence > **Explanation:** ARCH focuses on changing conditional variance rather than just average levels. ### Why are ARCH models useful in finance? - [x] Because asset-return volatility often clusters over time - [ ] Because returns always have constant variance - [ ] Because prices never jump - [ ] Because forecasting risk is irrelevant > **Explanation:** Volatility clustering is a standard empirical regularity in financial data. ### In an ARCH model, today's variance depends on: - [x] past squared shocks - [ ] only today's price level - [ ] future inflation announcements - [ ] no information from the past > **Explanation:** The model builds conditional variance from earlier disturbances.