A Machine Learning Perspective on using Accounting data to forecast GDP growth
In joint work with HBS faculty, we find that accounting variables relating to profits, accrual estimates, capital raises, and capital allocation decisions are most informative for the longer-term outlook for the economy. We also verify that non-linear techniques like Random Forest can help considerably.
Privacy vs. Alpha
Macro Prediction Book chapter
Traditional macroeconomic data used by economic agents to make decisions are noisy, lack richness, and produced with considerable lag. This chapter explores how alternative, web-scale data sources (“Big Data”) can help. We present a case study using a common alternative data source- web search to predict one of the most important data releases- non-farm payrolls (NFP). We discuss the efficacy of various machine learning (ML) techniques, the live performance of alternative data prediction models and the typical problems faced in practice.
Improvised Marketing Interventions in Social Media.
Does quick marketing wit on social media matter for firm value? We investigate the online phenomenon of improvised marketing interventions (IMI) - social media actions that are composed and executed in real time proximal to an external event, using five multimethod studies and point to features that help reap online sharing and firm value benefits.