Forecasting the equity premium: Do deep neural network models work?
This paper constructs deep neural network (DNN) models for equity-premium forecasting. We compare the forecasting performance of DNN models with that of ordinary least squares (OLS) and historical average (HA) models. The DNN models robustly work best and significantly outperform both OLS and HA mo...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Modern Finance Institute
2023-08-01
|
| Series: | Modern Finance |
| Subjects: | |
| Online Access: | https://mf-journal.com/article/view/2 |
| _version_ | 1797294379887493120 |
|---|---|
| author | Xianzheng Zhou Hui Zhou Huaigang Long |
| author_facet | Xianzheng Zhou Hui Zhou Huaigang Long |
| author_sort | Xianzheng Zhou |
| collection | DOAJ |
| description |
This paper constructs deep neural network (DNN) models for equity-premium forecasting. We compare the forecasting performance of DNN models with that of ordinary least squares (OLS) and historical average (HA) models. The DNN models robustly work best and significantly outperform both OLS and HA models in both in- and out-of-sample tests and asset allocation exercises. Specifically, DNN models generate monthly out-of-sample R2 of 3.42% and an annual utility gain of 2.99% for a mean-variance investor from 2011:1 to 2016:12. Moreover, the forecasting performance of DNN models is enhanced by adding additional 14 variables selected from finance literature.
|
| first_indexed | 2024-03-07T21:30:25Z |
| format | Article |
| id | doaj.art-13396426982b4534adeee737b69b735f |
| institution | Directory Open Access Journal |
| issn | 2956-7742 |
| language | English |
| last_indexed | 2024-03-07T21:30:25Z |
| publishDate | 2023-08-01 |
| publisher | Modern Finance Institute |
| record_format | Article |
| series | Modern Finance |
| spelling | doaj.art-13396426982b4534adeee737b69b735f2024-02-26T16:13:35ZengModern Finance InstituteModern Finance2956-77422023-08-011110.61351/mf.v1i1.2Forecasting the equity premium: Do deep neural network models work?Xianzheng Zhou0Hui Zhou1Huaigang Long2Guosen SecuritiesTulane University & California State UniversityZhejiang University of Finance and Economics This paper constructs deep neural network (DNN) models for equity-premium forecasting. We compare the forecasting performance of DNN models with that of ordinary least squares (OLS) and historical average (HA) models. The DNN models robustly work best and significantly outperform both OLS and HA models in both in- and out-of-sample tests and asset allocation exercises. Specifically, DNN models generate monthly out-of-sample R2 of 3.42% and an annual utility gain of 2.99% for a mean-variance investor from 2011:1 to 2016:12. Moreover, the forecasting performance of DNN models is enhanced by adding additional 14 variables selected from finance literature. https://mf-journal.com/article/view/2equity premiumreturn predictabilitydeep neural networkasset allocationforecasting performance |
| spellingShingle | Xianzheng Zhou Hui Zhou Huaigang Long Forecasting the equity premium: Do deep neural network models work? Modern Finance equity premium return predictability deep neural network asset allocation forecasting performance |
| title | Forecasting the equity premium: Do deep neural network models work? |
| title_full | Forecasting the equity premium: Do deep neural network models work? |
| title_fullStr | Forecasting the equity premium: Do deep neural network models work? |
| title_full_unstemmed | Forecasting the equity premium: Do deep neural network models work? |
| title_short | Forecasting the equity premium: Do deep neural network models work? |
| title_sort | forecasting the equity premium do deep neural network models work |
| topic | equity premium return predictability deep neural network asset allocation forecasting performance |
| url | https://mf-journal.com/article/view/2 |
| work_keys_str_mv | AT xianzhengzhou forecastingtheequitypremiumdodeepneuralnetworkmodelswork AT huizhou forecastingtheequitypremiumdodeepneuralnetworkmodelswork AT huaiganglong forecastingtheequitypremiumdodeepneuralnetworkmodelswork |