CEU Electronic Theses and Dissertations, 2012
| Author | Koncz, Gábor |
|---|---|
| Title | Can Correlation-Based Networks Capture Systemic Risks in a Financial System? The Case of Lehman's Collapse |
| Summary | In this thesis, I evaluate whether systemic risks in a financial system can be captured by a network built using only publicly available data. I construct the correlation-based network of publicly traded US banks based on stock prices prior to Lehman's collapse, and I assess whether this network could have predicted which bank stocks would suffer the biggest drops in prices after Lehman's collapse. I find that a correlation-based network built using the Minimal Spanning Tree method can tell us some valuable information about systemic risk. I show that some of the stocks with the highest drops lie close to Lehman in the tree. Moreover, when I consider the length of the path between every bank's node and Lehman, I find that a 10 percent increase in the path length to Lehman is associated with a 0.081 standard deviations decrease in the price drop on average. Importantly, the network is a better predictor of the price drops than simply the correlation with Lehman. Robustness tests show that using two alternative methods to construct the tree (the threshold- and the partial correlation-based one) were unable to predict price drops. Therefore I conclude that it does matter how the network is constructed if we want to capture systemic risks. In this example not all the correlation-based networks can capture these risks, only the Minimal Spanning Tree. |
| Supervisor | Canidio, Andrea |
| Department | Economics MA |
| Full text | https://www.etd.ceu.edu/2012/koncz_gabor.pdf |
Visit the CEU Library.
© 2007-2021, Central European University