Dragon-Kings in financial data - study at different time scales -
Tutor / director / evaluatorSornette, Didier, 1957-
Document typeMaster thesis (pre-Bologna period)
Rights accessOpen Access
Drawdowns (loss fromthe last local maximum to the next local minimum) o er a more natural measure of the financial market dynamics than fixed time-scale measures. We study the presence of Dragon Kings corresponding to meaningful outliers in the distribution of drawdowns at di erent time scales - from 1-min to daily. Our analysis comprises nine time series of prices of futures tracking major stock indexes (S&P, FT-SE, Nikkei), currencies (Yen, DM) and government bonds (Japan, US, Germany). We find no empirical evidence of the presence of Dragon-Kings for high frequency data, 1-min resolution. Nevertheless, for the largest scales, namely daily, the statistical tests applied demonstrate that the 1% quantile of the largest events of the population of drawdowns belong to a distribution significantly di erent from the rest. For the other scales, the test results are inconclusive. This results suggest that the feedback mechanisms present in the Dragon-Kings require a certain time to buildup.
Projecte final de carerra realitzat en col.laboració amb ETH Zurich