SEMINAR: The econometrics of financial duration modeling Seminar
Speaker: Giuseppe Cavaliere (U Exeter and U Bologna)
Abstract: Financial durations models are widely used in finance to model time between events such as trades, stock price movements, or other financial events. A workhorse in the literature is the classical autoregressive conditional duration (ACD) model by Engle and Russell (Econometrica, 1998), where the expected conditional duration depends on the most recent past durations. The likelihood of the model resembles the Gaussian (G)ARCH likelihood, and it is widely believed that asymptotic results for conditional volatility models carry over to ACD models. We show that this common wisdom is generally incorrect, and that the asymptotic theory for ACD is actually non-standard, and requires different machinery. In particular, the behavior of likelihood estimators in ACD models is highly sensitive to the tail behavior of the financial durations: asymptotic normality breaks down when the tail indices of the durations are smaller than unity, and estimators are mixed Gaussian with non-standard rates of convergence. These results are based on exploiting the fact that the number of observations within any given time span is random for duration data. The implications of these results for financial durations modeling and standard/bootstrap inference are also discussed.
(based on joint works with Thomas Mikosch, Anders Rahbek and Frederik Vilandt, U Copenhagen)
Host: Laura Coroneo (York)