Abstract:
This thesis is an attempt to explain Turkish financial crisis of 2000 in the light of network theory. Our contribution is applying network theory to a very special data that spans pre-crisis and crisis periods. Analyzing time series behavior of network parameters during such a special period is can bring new perspectives for applications of network theory to financial systems. Turkish overnight money market is considered as a real world network and daily networks of the year 2000 were constructed with the help of market microstructure data. Results showed that during the year network was highly centralized due to the huge borrowing of Demirbank. Centralization around Demirbank changed the structure and broke ties between other institutions which decreased connectivity. November crisis demolished highly central structure around Demirbank and connectivity increased during the crisis. Network illustrations and time series behavior of connectivity, connectivity-volume correlation, backbone volume ratio, Google Pagerank value for the central node all show the centralization around Demirbank before crisis and the collapse of this structure during the crisis. These parameters can be watched to catch this kind of centralization and to prevent turbulences related to the central node.
