Long memory behavior of returns after intraday financial jumps

In this paper, characterization of intraday financial jumps and time dynamics of returns after jumps is investigated, and will be analytically and empirically shown that intraday jumps are power-law distributed with the exponent 1<μ<2; in addition, returns after jumps show long-memory behavior. In the theory of finance, it is important to be able to distinguish between jumps and continuous sample path price movements, and this can be achieved by introducing a statistical test via calculating sums of products of returns over small period of time. In the case of having jump, the null hypothesis for normality test is rejected; this is based on the idea that returns are composed of mixture of normally-distributed and power-law distributed data (∼1/r1+μ). Probability of rejection of null hypothesis is a function of μ, which is equal to one for 1<μ<2 within large intraday sample size M. To test this idea empirically, we downloaded S&P500 index data for both periods of 1997–1998 and 2014–2015, and showed that the Complementary Cumulative Distribution Function of jump return is power-law distributed with the exponent 1<μ<2. There are far more jumps in 1997–1998 as compared to 2015–2016; and it represents a power law exponent in 2015–2016 greater than one in 1997–1998. Assuming that i.i.d returns generally follow Poisson distribution, if the jump is a causal factor, high returns after jumps are the effect; we show that returns caused by jump decay as power-law distribution. To test this idea empirically, we average over the time dynamics of all days; therefore the superposed time dynamics after jump represent a power-law, which indicates that there is a long memory with a power-law distribution of return after jump.