Price Forecasting and Analysis of Exchange Traded Fund


Journal of Mathematical Finance

March 2013, n°3, pp.181-191

Departments: Information Systems and Operations Management, GREGHEC (CNRS)

Keywords: Forecasting, Pricing, ETF, Exchange Traded Fund, SPY, Holt's Exponential Smoothing, Linear Regression, Multiple Regression, ARIMA Models

ETFs are baskets of securities designed to track the performance of an index. They are designed to provide exposure to broad-based indexes at a lower cost. We first analyzed why ETF should be the choice for an investment. We provide a brief history of this segment, key attributes of ETFs, and investments strategies and implementations with ETFs. The article then presents data analysis and a series of forecasting methods with data analysis techniques to evaluate the per- formance of each method. The data analysis and the forecast evaluation is to determine the best forecasting model for a single ETF; SPY. The different techniques considered include single exponential smoothing, Holt's exponential smoo- thing, simple linear regression, multiple regression and various versions of Box-Jenkins (ARIMA) models. Based on the evaluation of a decade of past historical data, we provide a guidance for the price of our ETF (SPY) using the multiple regression technique (with an R-square of 98.4%), which produced promising results (with low forecast errors of 1% across several forecast metrics), among the different techniques evaluated. Promising results were also obtained using the Multiple regression technique on several other popularly traded ETF's