When Alternative Data Met Macroeconomics: Predicting China Fixed Asset Investments using Equity Financing Deal New

Abstract

Investment in fixed assets accounts for over 40% of China GDP. While financial press closely tracks and discusses China fixed asset investment, scientific literature has been mostly silent in studying it. With quite a limited amount of research pondering the profile [1], driving forces [2, 3] or impacts [4] of China’s investment activities, there is, to our knowledge, no research dedicated to predicting this important economic statistic. This paper is intended to fill this gap by approaching the problem in a data-driven way. After rigorously evaluating the proposed approach in terms of data signal and bias, rationale assumptions, predictive performance, method robustness, potential improvements as well as research impact, we arrive at the conclusion that the achievements of this paper have five folds. First, it fills the blank in academic research of predicting China per sector fixed asset investment. Second, it approaches a traditional problem (macro forecasting) with innovation (applying natural language processing techniques to alternative data1, specifically, equity financing news). Third, its proposed predictive indicators far outperform (with much higher R-squared) market consensus over the year of 2020, especially for the biotechnology & healthcare and manufacturing sectors as well as during covid-19 shocks. Fourth, updating on a daily basis, its resulting indicators are able to provide timely signals in day-to-day business life. Fifth, its underlying rationale of capturing early and alternative signals for macro forecasting is not limited to any typical macro forecasting problem. It should inspire more to explore alternative data to achieve better economic forecasts.