Latest developments on heavy-tailed distributions

The recent financial and economic crises have shown the dangers of assuming that the risks are nearly Gaussian distributed. The recent financial and economic crises have shown the dangers of assuming that the risks are nearly Gaussian distributed. In particular, non-causal representations are not identified in the case of Gaussian AR processes. By contrast, in the infinite variance case, non-causal patterns can be identified and are relevant to describe different types of time series behavior. While maximum likelihood inference is known to be computationally demanding for the most popular families of heavy tailed distributions, some minimum distance approaches may be more tractable. The extended tests have negligible size distortion and more power than standard tests. The tests are applied to competing symmetric leptokurtic distributions with monthly return data on the US stock market index. These distributions are generally not picked as plausible alternatives, primarily because of the presence of skewness.
SCOPUS: cp.j
info:eu-repo/semantics/published