CONTROLLING THE STABILITY OF AN INTERCONNECTED FINANCIAL NETWORK
Submitted to Professor Trobaugh and the Department of Electrical and Systems Engineering
Woenho Chung, B.S. in Biomedical Engineering and B.S. in Electrical Engineering
Nathan Vogt, B.S in Electrical Engineering and Master of Business Administration
Dr. Zachary Feinstein, Project Advisor
Final Report Submission Date: April 30, 2017
Project Timeline: January 2017 – May 2017
Under a zero-shock situation for the Eisenberg-Noe model, there still existed some drop in the asset prices. The nodes experienced a 5% drop in the prices of LT investment and a 16% in Other assets. This was unexpected. However, the function used to map the price drops was hypothetical in nature and we have no idea how accurately a hyperbolic cosine function reflects the actual behaviors of asset fire sale.
Also, the function does not take into account consumer 28 behaviors. Thus, the drops will be attributed to inaccuracies in the model and the values obtained for this trial will be used to describe a perfectly stable system. For the other three models, both the Others and PP&E were unaffected. Even with LT investment and Other assets, there were no large deviations from the stable level, established in the previous paragraph.
CCP, however, showed to be most resilient to changes in the asset prices, by showing only a 0.5% drop in LT investment prices as opposed to 5% in all other models. Under an endogenous shock, all four models fared similarly. The baseline model, Eisenberg-Noe, turned out to be least resilient under an endogenous shock. As mentioned previously, the CCP model should be quite robust (and even more with guarantee funds), thus was not surprising to see the results.
The islanding model, however, was a surprise due to the large number of default banks. The islanding model does very well for an endogenous stress. If only one node fails, it is easy to simply isolate that node. When a multitude of nodes fail, it becomes difficult as the protocol of the mechanism forces isolation of default nodes, leading to a cascading effect of isolation-load reassignment-default due to the reassignment-more isolation. The exogenous stress produced a greater drop with close to 30% in the prices of other assets.
This seems to align well with the previously established conclusion; that the system is better at handing an endogenous stress than an exogenous one. Also, it was not surprising to see that both the CCP and the bailout models did better than the baseline model and the islanding model.