Extremes in heterogeneous economic data

Thesis

Large economic surprises occur with greater frequency than previously assumed. Popular examples include the severity of the 2008 financial crisis, the wages paid to top employees, and the damage from climate change. Their frequencies are growing more obvious, not just thanks to the passing of time, but also the greater availability of large scale, granular data. On one hand, these datasets improve the statistical power of traditional, econometric models developed for small sample sizes. On the other hand, an obsession with central moments, or ‘average effects’, obscures the role of higher moments in both theory as well as statistical forecasting. This thesis studies this systematic bias in economic models for the distribution of economic quantities.

The role of granular data is increasingly important for economic research. The first part of this thesis studies the impact of extreme values in two applications. The first application documents the impact of social contagion on stock returns using text data scraped from the WallStreetBets online forum. The sentiments of users who attract thousands of comments and follow-up conversations generate an outsized impact on stock prices relative to average sentiments, due to the heavy-tailed nature of viral social media content. The second application documents the heavy tails in the productivity of European firms. This study implies that a nontrivial number of firms generate millions of euros per employee in value added, while productivity for the typical firm remains in line with national averages, at around tens of thousands of euros. This dispersion is difficult to reconcile with the usual assumption in econometric studies that asserts a finite second moment for labour productivity at the firm level.

I formalise the relationship between the size of economic variables and their frequency in the second half of this thesis, by introducing entropy as a key measure for variety. Entropy is a useful measure of variety when constructing aggregates from heterogeneous data, such as economic output. In one example, the degree to which output is more or less concentrated is consequential to what economic growth truly represents. In finance, heterogeneity plays a crucial role in diversification. Banks decide on the maturity of their investments, and mutual funds decide on what mandate to construct a portfolio on. Entropy regulates the frequency of wealth allocated to different bonds or stocks and thus governs interest rates and stock prices both in and out of equilibrium.

Authors: Winkler, J

• DOI: 10.5287/ora-kovo82mqw
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: productivity dispersion; heterogeneity; heavy-tailed distributions; investor sentiment; maximum entropy principle
• Subjects: Complexity economics; Macroeconomics; Economic statistics; Economic measurement; Behavioural finance