Title: Forecasting low-frequency return density using high-frequency information: MC simulation and FAR
Authors: Minjoo Kim - University of Glasgow (United Kingdom) [presenting]
Abstract: Motivated by a stylized fact that intraday returns can provide additional information on the distribution of daily returns, we propose a novel forecasting framework combining MC simulation and Function Autoregressive Model (FAR) of nonparametric density functions. First, we simulate daily returns by a continuous-time data generating process (DGP). We estimate the DGP using intraday returns and it simulates daily returns. We then approximate the empirical density function of simulated daily returns relying on the nonparametric density estimator. We replicate this procedure for each day and the empirical densities are used as a functional sample. The advantage of this approach is that the continuous-time filter is able to control serial dependence often observed in the intraday data. Next, we simulate daily returns without requiring specific distribution family on the intraday returns. This MC stimulation is the most popular and standard in the financial engineering literature. Second, we apply FAR to the empirical densities. Based on the in-sample estimation, we forecast a daily return density function. The forecast can provide various information on the financial risk such as extreme event, event probability and interval. Finally, we evaluate our approach by both simulation study and empirical evaluations. Furthermore, we introduce its usage in financial applications.