Autoregressive Models in Finance: AR, ARMA and ARIMA Explained

What Is an Autoregressive Model?

An autoregressive (AR) model is a time series model where the current value of a variable is expressed as a linear function of its own past values plus a random error term. AR models are fundamental tools in financial econometrics, used for modelling and forecasting interest rates, volatility, asset returns, and macroeconomic variables.

The AR(p) Model

An AR(p) model uses the previous p values to predict the current value: Y_t = c + φ₁Y_{t-1} + φ₂Y_{t-2} + ... + φ_pY_{t-p} + ε_t, where φ₁...φ_p are the autoregressive coefficients and ε_t is white noise. The order p determines how many lagged values are included — selected using information criteria (AIC, BIC) or the partial autocorrelation function (PACF).

ARMA and ARIMA

ARMA(p,q) combines autoregressive (AR) terms with moving average (MA) terms, capturing both autocorrelation in the level and autocorrelation in the error. ARIMA(p,d,q) adds differencing (d) to handle non-stationary series — most financial and economic time series (prices, GDP) are non-stationary and require differencing before AR modelling is appropriate.

Applications in Finance

AR models are used to model interest rate dynamics, forecast inflation and macroeconomic variables, model the autocorrelation structure of volatility (GARCH models are extensions of AR models applied to variance), and build error correction models for cointegrated financial series.

Stationarity

For an AR model to produce reliable forecasts, the time series must be stationary (constant mean and variance). Most asset prices are non-stationary; asset returns are often stationary. The Augmented Dickey-Fuller (ADF) test is the standard test for unit roots in financial time series.

Further Reading

• Risk Management Careers
• Python for Finance
• CPD for Finance Professionals

Study with Learnsignal: CPD and qualification courses for quantitative finance professionals. Browse CPD.