Econometric Methods for Financial and Climate Time Series
Year: 2026
Type: dissertation
Abstract: This thesis proposes novel econometric methods for analyzing financial and climate time series. Although the fields of finance and climate science may seem distinct, they share common statistical modelling challenges. Despite the differences in methodology and research questions across chapters, each chapter seeks to model an underlying process generated by an unknown data-generating mechanism. A central objective throughout the thesis is to obtain consistent estimates of the parameters governing these processes, thereby providing insight into their underlying dynamics. We also seek to quantify the uncertainty surrounding these estimates to facilitate valid statistical inference. In Chapter 2, we propose a new bootstrap methodology for more accurately quantifying the uncertainty surrounding estimates of GARCH volatility processes with dependent innovations. By sampling blocks rather than individual observations, our approach preserves the dependence structure of the innovations. Interestingly, the resulting confidence intervals for the GARCH parameters can be either wider or narrower than those obtained under the inconsistent fixed-design bootstrap, depending on the underlying data-generating process. Then, in Chapter 3, we propose a seasonal quantile autoregressive model to investigate heterogeneity in persistence across the temperature distribution and its relationship with the NAO index. We find a statistically significant increase in the persistence of warm temperatures in winter in north-western Europe in recent decades. This increase is more pronounced when conditioning on the NAO index. Similarly, we document a statistically significant increase in the probability of precipitation in winter in recent decades when the NAO index reaches strongly positive values. Moving forward, in Chapter 4, we introduce a method for estimating the number and locations of change points in discontinuous piecewise linear trend models with AR(1) errors. In contrast to earlier work on detecting change points in annual global temperature time series, our approach allows for explicit control of the error rate. We assess the performance of the proposed procedure through a simulation study. When applied to annual global mean temperature time series, our method produces results similar to those reported in earlier studies. However, these results are obtained only at a false discovery rate of 50%, highlighting the limitations of drawing strong conclusions from limited data. Finally, in Chapter 5, we employ a local linear model to investigate whether global warming has accelerated. This flexible approach does not impose a predetermined functional form on the trend, allowing for a more smoothly evolving trend over time. Additionally, we include the Niño 3.4 index as a covariate to account for the influence of ENSO on global mean temperatures. In contrast to earlier studies, we find no evidence of a structural acceleration in global warming in recent years. Instead, our results suggest that the observed acceleration may be temporary in nature.
DOI: 10.5463/thesis.1863
Author: Barend Adriaan Spanjers
Language: English
Subfield: Finance
Domain: Social Sciences
SDG: Climate action
Open Access status: gold
OpenAlex ID: W7225738846
