Advanced Techniques in Time Series Analysis: From Structural Changes to Time-Varying Networks, March 2024

On March 8th, coinciding with International Women's Day, we are pleased to offer a workshop on time series analysis. The workshop will cover current methods for analysing structural changes and network dynamics. The event also marks Jiajing 'Jane' Sun’s appointment to the School of Mathematics. Aligning this workshop with International Women's Day  underscores our ongoing efforts towards inclusivity and acknowledges the significant role of women in mathematics. The event is scheduled in Room 524 in the School of Education, and will be available as a hybrid session to facilitate broader participation and to reduce our environmental impact. Teams Meeting ID: 331 823 481 31 (Passcode: ZeUjFj).

We recognise the value of diversity in our discipline and welcome all attendees, with a special invitation to PhD students. The workshop offers a chance to engage with specialists in the field and explore the forefront of research. Supported by the London Mathematical Society, the day is set to feature discussions and the chance for collaboration, reflective of our commitment to excellence in mathematical sciences.

Schedule

10:00-10:45: Jiajing Sun (Birmingham)
Title: Sequential change point detection for time series - an adjusted-range based approach
Abstract: This talk introduces the adjusted-range based sequential change point detection, extending the work of Chan et al. (2021). We find that our approach is more sensitive in detecting structural changes and is valid even when there are 'almost undetectable' breaks present in the training sample. In contrast, Chan et al.'s (2021) approach only performs well when the training sample is 'clean'. This finding echoes those from Shao and Zhang (2010) which state that a naive  application of the self-normalization of Shao (2010) fails when directly applied to the KS test statistic. It supports the use of the adjusted-range based self-normalization (SN) as a novel approach. We establish the null asymptotic and the consistency of the proposed sequential change-point test under general regularity conditions. Simulation experiments and applications identify structural modifications in conditional heteroskedasticity within two exchange rates (GBP/USD and EUR/USD) related to the United Kingdom European Union membership referendum 2016 and the global COVID-19 outbreak in 2020.

10:40-11:30 Anindya Banerjee (Birmingham)
Title: Panel data cointegration testing with structural instabilities
Abstract: Spurious regression analysis in panel data when the time series are cross-section dependent is analysed in the paper. The set-up includes (possibly unknown) multiple structural breaks that can affect both the deterministic and the common factor components. We show that consistent estimation of the long-run average parameter is possible once cross-section dependence is controlled using cross-section averages in the spirit of Pesaran's common correlated effects approach. This result is used to design individual and panel cointegration test statistics that accommodate the presence of structural breaks that can induce parameter instabilities in the deterministic component, the cointegration vector and the common factor loadings.

12:00-12:45 Jia Chen (York)
Title: Estimating Time-Varying Networks for High-Dimensional Time Series
Abstract: We explore time-varying networks for high-dimensional, locally stationary time series using the large VAR model framework. Both transition and error precision matrices evolve smoothly over time. Two types of time-varying graphs are investigated: one contains directed edges representing Granger causality linkages, and the other contains undirected edges indicating partial correlation linkages. Under the sparse structural assumption, we propose a penalized local linear method with time-varying weighted group LASSO to jointly estimate the transition matrices and identify their significant entries. Additionally, a time-varying CLIME method is used to estimate the precision matrices. The estimated transition and precision matrices then determine the time-varying network structures. Under some mild conditions, we derive the theoretical properties of the proposed estimates, including their consistency and oracle properties. We also extend the methodology and theory to cover highly-correlated, large-scale time series. Here, the sparsity assumption becomes invalid, and we account for common factors before estimating the factor-adjusted time-varying networks. We provide extensive simulation studies and an empirical application to a large U.S. macroeconomic dataset to illustrate the finite-sample performance of our methods.

14:00-14:45 Marco Barassi (Birmingham)
Title: Panel VAR Models with Latent Group Structures
Abstract: This paper introduces the Panel Vector Autoregression (PVAR) model with grouped fixed effects (PVAR-GFE) and its extensions. The PVAR-GFE is a parsimonious structural model, allowing for easy computation which, unlike previous studies on univariate panel models with interactive fixed effects, offers greater flexibility by not requiring pre-specification of the number of groups, group membership, or the number of group factors in each equation. Additionally, it easily extends to group-specific heterogeneous coefficient PVARs. We propose an iterative least squares with K-means clustering
estimator and establish its asymptotic properties, demonstrating consistency of the estimator for large N and T. Compared to existing PVAR-GMM and PVAR-IFE models, PVAR-GFE offers several advantages. Firstly, it is shown to be consistent in large-scale settings (large N and T). Secondly, Monte Carlo simulations reveal diminishing bias and negligible  misspecification rates under similar conditions. We employ the PVAR-GFE to re-examine the relationship between financial development, economic growth, and energy consumption in 30 Chinese provinces from 1996Q1 to 2015Q4. This application demonstrates the model’s usefulness in empirical research.

14:45-15:30 Yiannis Karavias (Brunel)
Title: Multiple Structural Breaks in Interactive Effects Panel Data and the Impact of Quantitative Easing on Bank Lending
Abstract: This paper develops a new toolbox for multiple structural break detection in panel data models with interactive effects. The toolbox includes tests for the presence of structural breaks, a break date estimator, and a break date confidence interval. The new toolbox is applied to a large panel of US banks for a period characterized by massive quantitative easing programs aimed at lessening the impact of the global financial crisis and the COVID19 pandemic. The question we ask is: Have these programs been successful in spurring bank lending in the US economy? The short answer turns out to be: “No”.

15:30-17:30 Social Event
Participants and speakers in the workshop are invited to a social event at the School of Mathematics after the workshop. External speakers will have the opportunity to meet colleagues from the School of Mathematics. Additionally, this event provides research students with a venue to meet academics and discuss their research themes in a less formal setting. Our preliminary goal is to celebrate the new appointment of Dr Jiajing Sun, but this platform should also foster relationships and cultivate potential research collaborations. In particular, we aim to promote the application of mathematical statistics in addressing real societal challenges.

Evening: Dinner by invitation