Paper: arXiv 2505.06190

Authors: Giuseppe Cavaliere, Thomas Mikosch, Anders Rahbek, Frederik Vilandt

Abstract

Integrated autoregressive conditional duration (ACD) models serve as natural counterparts to the well-known integrated GARCH models used for financial returns. However, despite their resemblance, asymptotic theory for ACD is challenging and also not complete, in particular for integrated ACD. Central challenges arise from the facts that (i) integrated ACD processes imply durations with infinite expectation, and (ii) even in the non-integrated case, conventional asymptotic approaches break down due to the randomness in the number of durations within a fixed observation period. Addressing these challenges, we provide here unified asymptotic theory for the (quasi-) maximum likelihood estimator for ACD models; a unified theory which includes integrated ACD models. Based on the new results, we also provide a novel framework for hypothesis testing in duration models, enabling inference on a key empirical question: whether durations possess a finite or infinite expectation. We apply our results to high-frequency cryptocurrency ETF trading data. Motivated by parameter estimates near the integrated ACD boundary, we assess whether durations between trades in these markets have finite expectation, an assumption often made implicitly in the literature on point process models. Our empirical findings indicate infinite-mean durations for all the five cryptocurrencies examined, with the integrated ACD hypothesis rejected – against alternatives with tail index less than one – for four out of the five cryptocurrencies considered.

Complexity vs Empirical Score

  • Math Complexity: 9.0/10
  • Empirical Rigor: 8.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper develops non-standard asymptotic theory for integrated ACD models, involving advanced limit theorems and mixed normal distributions, resulting in a high math complexity score. It is applied to high-frequency cryptocurrency data, with specific findings and testing procedures, indicating substantial empirical implementation, though it focuses on a single dataset.

Research Flowchart

  flowchart TD
  A["Research Goal: Develop unified asymptotic theory for ACD models<br>and test for finite vs. infinite mean durations"] --> B["Methodology: Quasi-Maximum Likelihood Estimation<br>Unified Asymptotic Theory for Integrated ACD"]
  B --> C["Data: High-Frequency Cryptocurrency ETF Trading Data<br>5 Cryptocurrencies examined"]
  C --> D["Computational Process: Estimation & Hypothesis Testing<br>Testing for infinite expectation using tail index analysis"]
  D --> E{"Key Findings & Outcomes"}
  E --> F["Infinite-Mean Durations Found<br>in all 5 cryptocurrencies"]
  E --> G["Integrated ACD Hypothesis Rejected<br>for 4 out of 5 cryptocurrencies<br>(against alternatives with tail index < 1)"]
  E --> H["New Framework Established<br>for inference on finite/infinite expectation in duration models"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper develops a unified asymptotic theory for quasi‑maximum likelihood estimation in ACD models, including the integrated case, and provides a testing framework for finite versus infinite mean durations.

Source sentence
we provide here unified asymptotic theory for the (quasi-) maximum likelihood estimator for ACD models; a unified theory which includes integrated ACD models.

Assumes 6

  • Durations are generated as xi = ψi εi with εi i.i.d. and E[εi]=1. model
    Source sentence
    Consider xi = ψiεi, εi i.i.d. with E [εi] = 1, and conditional duration ψi given by ψi = ω + αxi−1 + βψi−1
  • Parameters α and β are positive. model
    Source sentence
    In terms of the parameters α, β > 0 the results in CMRV…
  • The intercept parameter is fixed at ω0 = 1 in simulations. model
    Source sentence
    The data generating processes (DGPs) for the simulations are given by the ACD process as defined in equations (2.1)-(2.2), with the intercept parameter fixed at ω0 = 1.
  • The diurnally adjusted intra‑day durations are obtained by correcting raw LOBSTER durations with cubic splines (knots every 30 minutes). data
    Source sentence
    the original, or ‘raw’ intra-day durations obtained from the limit order book are corrected for intradaily patterns using here cubic splines (with knots placed every 30 minutes).
Show 2 more
  • The observation period consists of regular trading hours (9:30 AM to 4:00 PM EST) over 35 trading days. market
    Source sentence
    The observation period of 35 trading days during regular trading hours (9:30 AM to 4:00 PM EST) corres…
  • Monte‑Carlo simulations assess finite‑sample performance under both the IACD null (α+β=1) and alternatives (α+β≠1). statistical
    Source sentence
    statistics… under the IACD null hypothesis and their behavior under alternative hypotheses featuring both finite and infinite expected durations.

Data

  • Cryptocurrency ETF trade durations · LOBSTER order-book commercial
    Jan 2 to Feb 28, 2025 (35 trading days) · Five ETFs tracking cryptocurrency prices (BTC, ETH, GBTC, ETHE, BITB)
    Source sentence
    ACD models to intra-day, diurnally-adjusted durations {xi} for five different exchange-traded funds (ETFs) tracking cryptocurrency prices from January 2 to February 28, 2025 (or, 35 trading days).

Recognised sources: LOBSTER

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • Empirically, all five cryptocurrency ETFs exhibit infinite‑mean durations; the integrated ACD hypothesis is rejected for four of the five (BTC, ETH, GBTC, ETHE). statistical
    Source sentence
    the null hypothesis is rejected at the 5% nominal level for four out of the five cryptocurrency ETF considered: BTC, ETH, GBTC and ETHE.
  • The QML estimator for the integrated ACD case converges at the non‑standard rate √t/ log t and remains asymptotically Gaussian. theoretical
    Source sentence
    that for κ = 1, the rate of convergence for the quasi maximum likelihood estimator (QMLE) is the non-standard
  • Monte‑Carlo results indicate that the τt statistic is asymptotically standard normal and that test size control is good in reasonable settings, though large samples are needed for some parameter configurations. statistical
    Source sentence
    To further illustrate the validity of the asymptotic results, we consider the τt statistic which by Corollary 3.2 is asymptotically standard normal
  • Estimated α and β parameters sum approximately to one for the cryptocurrency ETFs, supporting the integrated ACD characterization. statistical
    Source sentence
    We find that the ACD model provides estimators of α and β summing approximately to one.
  • Standard inference procedures based on t‑statistics and likelihood‑ratio tests remain valid under the integrated ACD setting. theoretical
    Source sentence
    standard inference procedures – based on t-statistics and likelihood ratio tests – remain valid under the integrated ACD setting.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 31 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.