The well-known facts of financial markets support price fluctuations containing information on the complexity of the interactions among market participants. In this work, we present a new surrogate method to find the dependency of higher-order correlations on the magnitude of price fluctuations. By sorting returns into several groups with respect to the level of fluctuations, we show that large fluctuations characterize the structure of the temporal correlations of a financial time series. In particular, by investigating the positive and the negative parts separately, we confirm that the risk-averse behavior of traders is explicitly observed in financial markets.