Time-Series Models for Financial Data › Model Identification

Michael BrenndoerferDecember 9, 20255 min read
Quantitative FinanceAll chapters
Part 8 of 15Econometrics and Statistical Modeling
Chapter 41Time-Series Models for Financial Data › Model Identification
Reading section 7 of 13 · Model Identification · 5 minSection 7 of 13 · 5 min

Book contents

Quantitative Finance15 parts · 79 chaptersView the whole book →
Active chapter 41 of 79

Section 7 of 13 · 5 min read

Time-Series Models for Financial Data › Model IdentificationOne section · Sections in this chapter7 / 13 ↓
Sections in this chapter13
Reading tools Definitions & inline links

Choose your expertise level to adjust how many terms are explained. Beginners see more tooltips, experts see fewer to maintain reading flow. Hover over underlined terms for instant definitions.

Article links

Make inline references clickable

Identifying the appropriate (p,d,q)(p,d,q) specification is both art and science. The Box-Jenkins methodology provides a systematic approach.

Step 1: Assess Stationarity

First, examine the series visually and statistically:

  • Visual inspection: Does the series trend or wander? Does variance appear stable?
  • ACF analysis: For non-stationary series, the ACF decays very slowly
  • Unit root tests: The Augmented Dickey-Fuller (ADF) test formally tests for stationarity

If the series is non-stationary, difference it and repeat. Most financial price series require d=1d=1 (returns), while some economic series may need d=2d=2.

Step 2: Examine ACF and PACF

For the stationary (possibly differenced) series:

ACF and PACF patterns for model identification.
ModelACF PatternPACF Pattern
AR(p)Decays graduallyCuts off after lag pp
MA(q)Cuts off after lag qqDecays gradually
ARMA(p,q)Decays graduallyDecays gradually

The ACF and PACF of financial returns often suggest low-order models. Daily stock returns typically show little autocorrelation in levels but significant autocorrelation in squared returns (addressed in the next chapter on volatility modeling).

Out[5]:
Visualization
Bar chart of AR(2) ACF showing gradual decay over lags 0 through 12.
ACF of AR(2) with φ₁ = 0.5 and φ₂ = 0.3. The autocorrelation decays gradually, a hallmark of autoregressive processes.
Bar chart of MA(2) ACF showing significant bars at lags 0, 1, 2 and near-zero bars thereafter.
ACF of MA(2) with θ₁ = 0.6 and θ₂ = 0.3. The ACF cuts off sharply after lag 2, the defining signature of a second-order moving average process.
Bar chart of ARMA(1,1) ACF showing gradual decay.
ACF of ARMA(1,1) with φ = 0.7 and θ = 0.4. The autocorrelation decays gradually, so this mixed process has no sharp ACF cutoff for visual order identification.
Out[6]:
Visualization
Bar chart of AR(2) PACF showing sharp cutoff after lag 2 with near-zero bars at higher lags.
PACF of AR(2). The partial autocorrelation cuts off sharply after lag 2, confirming the AR(2) order. This cutoff property is the primary tool for identifying autoregressive order.
Bar chart of MA(2) PACF showing gradual decay with alternating signs.
PACF of MA(2). The partial autocorrelation decays gradually, contrasting with the sharp ACF cutoff and completing the identification pattern for MA(2).
Bar chart of ARMA(1,1) PACF showing gradual decay.
PACF of ARMA(1,1). Gradual decay in the PACF, paired with gradual ACF decay, characterizes a mixed process that requires information criteria rather than a visual cutoff for order selection.

Step 3: Estimate Candidate Models

Fit several plausible specifications and compare using information criteria:

AIC=−2ln⁡(L^)+2k(Akaike Information Criterion)BIC=−2ln⁡(L^)+kln⁡(n)(Bayesian Information Criterion)\begin{aligned} \text{AIC} &= -2 \ln(\hat{L}) + 2k && \text{(Akaike Information Criterion)} \\ \text{BIC} &= -2 \ln(\hat{L}) + k \ln(n) && \text{(Bayesian Information Criterion)} \end{aligned}

where:

  • AIC\text{AIC}: Akaike Information Criterion
  • BIC\text{BIC}: Bayesian Information Criterion
  • L^\hat{L}: maximized likelihood
  • kk: number of parameters
  • nn: sample size

Lower values indicate better models. BIC penalizes complexity more heavily and tends to select simpler models.

Step 4: Diagnostic Checking

After estimation, verify that residuals behave like white noise:

  • Ljung-Box test: Tests whether residual autocorrelations are jointly zero
  • Residual ACF: Should show no significant autocorrelations
  • Normality tests: Check if residuals are approximately normal (though financial residuals often have fat tails)

If diagnostics fail, revisit the specification.