Testing for breaks in variance structures with smooth changes

TL;DR

This paper introduces a new statistical method for detecting variance breaks in time series data that accounts for smooth changes, overcoming limitations of traditional tests that assume abrupt shifts.

Contribution

It proposes a novel procedure for identifying variance breaks considering smooth variations, improving detection accuracy over existing methods.

Findings

The new tests effectively distinguish between smooth variance changes and abrupt breaks.

Monte Carlo simulations demonstrate the tests' good finite sample properties.

Application to U.S. macroeconomic data illustrates practical usefulness.

Abstract

The problem of detecting variance breaks in the case of smooth time-varying variance structure is studied. It is highlighted that the tests based on (piecewise) constant specification of the variance are not able to distinguish between smooth non constant variance and the case where an abrupt change is present. Consequently, a new procedure for detecting variance breaks taking into account for smooth changes of the variance is proposed. The finite sample properties of the tests introduced in the paper are investigated by Monte Carlo experiments. The theoretical outputs are illustrated using U.S. macroeconomic data.