Abstract:
There is increasing interest and an ongoing debate on the return behavior of financial markets. The problem is that there is no sufficient evidence about successful modeling of stock returns yet. In fact, the possibility of a successful prediction model itself is still open for discussion. Financial time series are complex, noisy, and randomlooking. Linear modeling attempts have always failed whereas nonlinear ones have achieved only little. At this point, ‘chaos theory’ may provide some, if not all, answers we have been looking for. Finding a deterministic structure in a system implies that a successful prediction model is theoretically possible. Furthermore, being able to identify that structure’s characteristics, e.g. fractal dimension, means that such a model is practically possible as well. This thesis examines the return behavior of Istanbul Stock Exchange index (ISE100) in the light of ‘chaos theory’, which is almost totally missing in the current literature. The time period covered is the last eleven years, from 01.01.1998 to 16.12.2008. The main return series were created by adding one index level at every tenth second and then by calculating the logarithmic differences of the consecutive values. As a summary of the findings, there is yet no reason that prevents us from imagining the stock returns as different weather conditions. Successful short term predictions are theoretically possible but it becomes impossible to speak thoroughly about the long term. However, to become the true ‘meteorologist’ of the financial markets, one first has to develop an effective nonlinear noise filtering method which does not distort the original data and is still capable of thoroughly capturing the hidden signal in it. In the absence of such a good filtering method, the true ‘meteorologist’ becomes an ordinary ‘fisherman’ who has to rely on his/her luck at some point!
