What can we learn from machine forecasting error?

Title: What can we learn from machine forecasting error?

Speaker：Rui Shen (CUHK (Shenzhen), Associate Professor)

Host：Yun Dai（Lingnan College, Sun Yat-sen University, Assistant Professor）

Date and Time: November 9, 2018 (16:00-17:30)

Location: Wang Daohan Room, Lingnan Hall

Language: Chinese + English

Abstract:

In this study, we utilize four popular machine learning algorithms to generate earnings forecasts for each firm. We find statistically significant associations between ex post machine forecasting errors and existing public information (input variables) such as Tobin’s Q, accruals, past stock returns, past return volatility and past forecasting errors. These associations are qualitatively and quantitatively similar to those documented in sell-side financial analyst forecasting errors. Our evidence regarding machine forecasting errors cannot be explained by popular theories proposed to explain analyst forecasting errors such as agency incentives or cognitive bias. We provide evidence consistent with an estimation uncertainty explanation.

The personal website of the Seminar Speaker:

http://sme.cuhk.edu.cn/zh-hans/content/7554