Autoregressive Conditional Duration¶

Model positive intervals between irregular events as a unit-mean innovation times a conditional expected duration that evolves from past durations and past conditional means.

Version: v3 · 2026-09-19 · History
Domain-specific #: 1324
Domain group: Social Sciences
Origin domain: Economics & Finance
Subdomain: Financial Econometrics → Economics & Finance
Aliases: ACD model, Autoregressive conditional duration model

Core Idea¶

Let event times satisfy \(t_i>t_{i-1}\) and define duration \(x_i=t_i-t_{i-1}>0\). An autoregressive conditional duration (ACD) model writes

\[ x_i=\psi_i\varepsilon_i,\qquad E(\varepsilon_i\mid\mathcal F_{i-1})=1, \]

where \(\psi_i=E(x_i\mid\mathcal F_{i-1})\) evolves from prior durations and conditional means. In a linear ACD\((p,q)\),

\[ \psi_i=\omega+\sum_{j=1}^{q}\alpha_jx_{i-j}+ \sum_{k=1}^{p}\beta_k\psi_{i-k}, \]

with positivity restrictions. Engle and Russell introduced the model for clustered, irregularly spaced financial transaction data.

How would you explain it like I'm…

The Waiting-Time Guesser

At a lemonade stand, sometimes customers come one right after another, and sometimes there are long waits. When it's been busy, the next wait is usually short too. This idea makes a guess about the next wait by looking at the recent waits, and the real wait is that guess made a bit longer or shorter by surprise.

Predicting the Next Wait

In financial markets, trades don't happen at regular times — sometimes they come in quick bursts, sometimes there are long gaps. The time between two trades is called a duration. An autoregressive conditional duration model predicts the expected length of the next gap from the lengths of recent gaps and recent predictions. The actual gap is that expected length multiplied by a random surprise that averages out to 1. This captures the fact that short gaps tend to follow short gaps and long gaps tend to follow long gaps.

Clustered Waiting-Time Model

An autoregressive conditional duration (ACD) model describes the waiting times between events that happen at irregular moments, like trades in a financial market. If events happen at times t₁, t₂, and so on, the duration xᵢ is the gap tᵢ − tᵢ₋₁. The model writes xᵢ = ψᵢ·εᵢ, where ψᵢ is the expected duration given everything known before, and εᵢ is a positive random factor with average 1. The expected duration is updated from past durations and past expected durations — for example ψᵢ = ω + α·xᵢ₋₁ + β·ψᵢ₋₁, with parameters kept positive so durations stay positive. This lets the model capture clustering: busy periods with short gaps and quiet periods with long ones. Engle and Russell introduced it for transaction data.

The autoregressive conditional duration (ACD) model, introduced by Engle and Russell for clustered, irregularly spaced financial transaction data, models the durations between successive events. With event times tᵢ > tᵢ₋₁ and durations xᵢ = tᵢ − tᵢ₋₁ > 0, it specifies a multiplicative structure xᵢ = ψᵢεᵢ, where ψᵢ = E(xᵢ | ℱᵢ₋₁) is the conditional expected duration given the past and the innovation εᵢ is positive with E(εᵢ | ℱᵢ₋₁) = 1. The conditional mean evolves autoregressively: in the linear ACD(p, q), ψᵢ = ω + Σⱼ₌₁^q αⱼxᵢ₋ⱼ + Σₖ₌₁^p βₖψᵢ₋ₖ, with positivity restrictions on the parameters. It captures duration clustering — short durations tend to be followed by short ones — through the dependence of ψᵢ on recent durations.

Scope of Application¶

ACD models handle trade, quote, price-change, volume, and order-arrival durations in ultra-high-frequency finance. They also provide a template for irregular event sequences in other domains. Engle embeds duration analysis in marked point-process views of complete transaction records and links surprising duration to market-microstructure quantities.

Clarity¶

Define the event, timestamp resolution, zero-duration treatment, overnight gaps, censoring, time-zone/session boundaries, and deterministic seasonality adjustment. State lag orders, recursion form, innovation distribution, covariates, positivity constraints, likelihood, and residual tests. Do not interpret a shorter predicted duration as higher activity without declaring the reciprocal intensity convention.

Manages Complexity¶

ACD models move analysis from arbitrary calendar bins to event time. A compact recursion summarizes duration clustering, while the innovation distribution separates predictable pace from unpredictable waiting. The same framework supports likelihood, hazard, forecast, and marked-event extensions.

Abstract Reasoning¶

Define and clean the event sequence.
Compute positive durations and handle ties.
Estimate or model intraday seasonality.
Select ACD lag orders and recursion family.
Choose a positive unit-mean innovation distribution.
Estimate parameters under admissibility constraints.
Standardize durations by fitted \(\psi_i\).
Test residual dependence and distributional fit.
Compare forecasts and alternative duration models.

Knowledge Transfer¶

The portable pattern is factor a positive waiting time into predictable local pace and unit-scale surprise, then let pace inherit memory from earlier waits. It transfers to irregular event-stream forecasting. The proposed immediate parent is Stochastic Process.

Relationships to Other Abstractions¶

Current abstraction Autoregressive Conditional Duration Domain-specific

Parents (1) — outgoing typed relations

Autoregressive Conditional Duration is a kind of Stochastic Process Prime

Stochastic Process is the broader abstraction this entry instantiates.

Hierarchy path (1) — routes to 1 parentless root

Autoregressive Conditional Duration → is a kind of Stochastic Process

Neighborhood in Abstraction Space¶

Autoregressive Conditional Duration sits in a sparse region of the domain-specific corpus (98^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2769 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-25