Recent Developments in Time-Series Methods for Detecting Bubbles and Crashes: Guest Editors' Introduction

Abstract

Financial and economic bubbles, along with their inevitable crashes, can have a significant impact on investment outcomes and a country's macroeconomic performance. From historical speculative behavior to modern asset price surges, identifying and measuring these phenomena has posed a crucial challenge for investors and policymakers alike. Examples such as the Dot-Com bubble that originated in the mid-1990s, the US housing market bubble of the late 1990s and early 2000s, and the Bitcoin price bubbles since the mid-2010s, underscore the need for robust econometric methods to detect the presence and timing of bubble behavior, either historically or in real time.

Time series modeling of bubble processes originated with methods to detect explosive behavior in asset prices. Diba and Grossman (1988) proposed using left-tailed unit root tests applied to the levels and first differences of prices, recognizing that differences of bubble processes are not stationary. In a now seminal paper on bubble testing, Phillips et al. (2011) [PWY] proposed a more direct approach, basing their test on recursive right-tailed unit root tests, with the alternative of explosive autoregression acting as a model of the bubble dynamics. Phillips et al. (2015) [PSY] extended the PWY approach to a doubly recursive test, which improves performance in a setting of multiple bubbles and crashes. In addition to methods of bubble detection, PWY and PSY also introduced techniques for dating the timing of a bubble's onset and collapse.

Building on the initial work of PWY and PSY, a rich literature has evolved in the area of detecting and dating bubbles and crashes, with developments in both econometric methodology and breadth of empirical application. Methodological contributions over the last ten years have been made in a variety of directions; these include, inter alia, alternative approaches to improve test power, methods to increase the accuracy of bubble and crash date estimates, procedures to robustify inference to features such as heteroskedasticity and/or jumps, and techniques to allow rapid detection of bubbles and crashes in real time. Applications have been made to a wide range of series, covering stocks, housing, metals, agricultural commodities, energy and cryptocurrencies, among others. The papers in this Special Issue continue the development of this important research area, and provide contributions to both methodology and empirical analysis.

We are very grateful to all of the authors who have contributed to this special issue. We would also like to thank the anonymous reviewers for their detailed comments and helpful feedback on each paper. Special thanks go to Robert Taylor, Editor-in-Chief of the journal, for commissioning the special issue, as well as to Priscilla Goldby for her invaluable help throughout the editorial process.