TY - JOUR T1 - Cross-Predictability of Industry Return in Trade Network: <em>Using LASSO</em> JF - The Journal of Investing DO - 10.3905/joi.2019.1.095 SP - joi.2019.1.095 AU - Rasha Ashraf Y1 - 2019/07/18 UR - https://pm-research.com/content/early/2019/07/18/joi.2019.1.095.abstract N2 - This study tests the predictability power of returns across economically related industries, using monthly data over 1990–2010 and a machine learning model, the Least Absolute Shrinkage and Selection Operator (LASSO). This article presents evidence that LASSO identifies significant predictors of the returns of an individual industry among its interdependent industries. The results indicate cross-predictability of industry return by showing a significant relation between an industry’s returns and the lagged returns of its LASSO-selected trade partners. The study finds that industries that are more central in the input-output network have a greater effect in predicting related industry return. It also computes out of sample (2011–2016) the mean return forecasts of the portfolio of industries based on the predictability effect of the lagged returns of trade partners. The self-financing trading strategy of buying (selling) a high (low) portfolio based on related industry lagged returns yields about 10.44 percent annual excess return (0.83 percent monthly), with an annual Sharpe ratio of about 0.86. The study shows that after controlling for the Fama-French (1993) three factors, the monthly returns from the self-financing trading strategies generate a significant alpha of 0.7 percent monthly over the period 2011–2016.TOPICS: Big data/machine learning, factor-based models, portfolio construction, performance measurement ER -