THE CENTER FOR ECONOMICS JOB MARKET SEMINAR ALI HABIBNIA
Ali Habibnia, London School of Economics and Political Science
2018-02-09 at 10:15
2018-02-09 at 11:45
Economics conference room 730, SSE, Bertil Ohlins Gata 4, 7th fl.
Job Market Paper: “Nonlinear Forecasting Using a Large Number of Predictors: A Deep Learning Approach”
Abstract: We propose a category of nonlinear forecasting procedures that contrasts with current models, generally aimed at either high dimensionality or nonlinearity, We propose a nonlinear generalization of the statistical factor model (equivalent to a deep learning autoencoder’s structure), as a first step, factor estimation, we employ an auto-associative neural network to estimate nonlinear factors from predictors. In the second step, forecasting equation, we apply a nonlinear function -feedforward neural network- on estimated factors for prediction. In this paper, we show that these features can go beyond covariance analysis and enhance forecast accuracy. We also adopt statistical tests to detect nonlinearity in and between time series, which can be used as guidance in choosing the appropriate models. We apply this approach to forecast equity returns, and show that capturing nonlinear dynamics between equities significantly improves the quality of forecasts over current univariate and multivariate factor models. Empirical results on daily returns of equities on the S&P 500 index from 2005 to 2014 provide support for the out-of-sample forecasting ability of these models vis-a-vis existing approaches.
The paper can be downloaded here
To find out more about Ali Habibnia, visit his website