- This event has passed.
Su-Min Lee + Shuyi Ge (External Seminar)
Talk 1: Shuyi Ge, “Text-based linkages and local risk spillovers in equity markets”
Time: 14:30 – 15:15 BST
Abstract: One stylised fact of asset returns is that the interconnectedness in idiosyncratic returns is non-negligible even in large dimensional systems. The network architecture of firms is the key to study the transmissions of local shocks. However, such linkage data is usually unavailable for researchers. This paper uses extensive text data to construct firms’ links that have not been documented in other sources. Utilising the novel text-based linkage data, I quantity the strength of local risk spillovers in the equity market by estimating a heterogeneous spatial autoregressive model (HSAR) for the de-factored (idiosyncratic) equity returns. The model outperforms several alternative methods in terms of out-of-sample fit. The estimation results show that after removing the common risk factors and industry risk factors, there is still a considerable degree of local risk spillovers, and with substantial industrial heterogeneity. By constructing spatial-temporal spillover matrices using the estimated parameters, we are able to identify the major systemic risk contributors and receivers, which are of the interest to microprudential policies. From a macroprudential perspective, a rolling-window analysis reveals that the strength of local risk spillovers increases during the crisis period, when, on the other hand, market factor loses its importance.
Talk 2: Su-Min Lee, “Learning-by-losing: Do political parties widen representation to win elections?”
Time: 15:15 – 16:00 BST
Abstract: Despite the right to vote being near universal in many countries, there still exist significant inequalities in political representation, which may have a profound influence on policy. One contributing factor may be political parties’ selection of candidates for election. I argue that parties may learn from losing elections, and potentially widen representation among their candidates. I exploit the unique case study of the Conservative Party learning from unexpected landslide defeat in the 1906 UK elections. I use hand-collected biographical data and machine-learning methods to classify over 2,000 candidates. A one standard deviation increase in the Conservatives’ 1906 defeat in a constituency is associated with a 10 percent decrease in likelihood that the Conservative candidate in the 1910 elections is from the political elite, and a 19 percent increase that they are from the working class. I find further evidence of a learning process that contributed to their recovery in 1910.