Econometric modelling of financial contagion with applications

© 2012 Dr. Jessie Xiaokang WangThis thesis develops a two-period rational expectations equilibrium (REE) model of contagion in an economy with information asymmetry and hedging activities. A new channel of contagion is proposed which is based on hedging strategies such as portfolio insurance. Due to asymmetric information about liquidation values and incomplete information of hedging activities, selling triggered by hedging strategies might be incorrectly perceived as a result of low fundamental values. This leads to uninformed investors adjusting downwards their expectations, causing prices to fall across markets. Through hedging, the model can generate contagion without the dependence of fundamentals. Consistent with empirical observations, the model also finds that contagion is larger in market downturns than in market upturns, which suggests that contagion is asymmetric. This distinction reflects the fact that hedging activities are more often implemented in a down market. Moreover, the model shows that information asymmetry can also greatly influence the strength of the contagion effect. As all other contagion channels are excluded, the proposed channel is shown to have an independent effect beyond other contagion channels. The simulation results show that correlation, coskewness and cokurtosis of asset prices are much higher during a crisis period compared to those in a noncrisis period. This is an important result as suggests that contagion may be transmitted through higher order moments. The thesis also develops a general test of contagion which is based on a comparison of conditional copula distributions of financial returns in crisis and noncrisis periods. An important feature of the approach is that a range of contagion channels are tested including the usual channels that operate through the first and second moments, as well as channels that operate through higher order moments including coskewness and cokurtosis. The finite sample properties of the test are examined using a range of Monte Carlo experiments and compared with existing tests of contagion. The empirical application applied to the recent global financial crisis shows that contagion occurs mainly through the fourth moment. A further contribution of the thesis is that the finite sample properties of a range of contagion tests are investigated using a set of Monte Carlo experiments. The data generating process is based on the Hawkes mutual jumps diffusion model of asset returns which is an extension of the recent model of Aït-Sahalia, Cacho-Diaz and Laeven (2010). The Forbes-Rigobon and Peseran-Pick tests are found to be under-sized, the Favero-Giavazzi test displays reasonable size, while the Bae-Karolyi-Stulz is over-sized. Proposed adjustments to the Forbes-Rigobon tests are found to correct the size distortion of this test. Moreover, the sizes of most tests increase in the presence of correlation in the noncrisis period. For these experiments the Chow contagion test, Pesaran and Pick and joint copula tests still exhibit the highest power. However, the Pesaran and Pick test is found to be oversized suggesting that empirical evidence of contagion based on this is potentially spurious