Kasparis, I., "Conventional Inference in the Vicinity of Generalised Nonstationarity Boundaries: Regressions with Heavy Tailed Weakly Nonstationary Data"

Title: "Conventional Inference in the Vicinity of Generalised Nonstationarity Boundaries: Regressions with Heavy Tailed Weakly Nonstationary Data" (with James Duffy - Oxford)

Speaker: Associate Professor Ioannis Kasparis, University of Cyprus

Host: Assistant Professor Alexopoulos Angelos, Department of Economics, Athens University of Economics and Business

Time: 15.30 -17.00

Room: A36

Abstact: The interaction between long memory/persistence and heavy tails leads to an expansion of the nonstationary region, effectively broadening the model space where traditional inferential  methods are not applicable. To address this, both parametric and non-parametric regression methods are explored to facilitate inference across both stationary and nonstationary environments in the presence of heavy tails. These include kernel weighted and IVX type of IV methods. For the purposes of our analysis, a new limit theory is developed for heavy-tailed weakly nonstationary processes (HT-WNPs) -processes that exist on the boundary of nonstationarity. It is demonstrated that the proposed methods enable conventional inference for a wide range of heavy-tailed covariates, including stationary long memory, weakly nonstationary processes, and strongly nonstationary long memory. Potential applications include testing the predictability of stock returns through risk measures.