What factors predict the innovation of Chinese family firms: empirical evidence based on machine learning

Purpose Prior studies on family firm innovation have focused primarily on causal explanations and isolated variable significance, but lack a prediction-oriented approach that systematically compares the relative importance of multiple predictors. Drawing on the capability–willingness–pressure framework, this paper aims to identify the key predictors of innovation in Chinese family firms and to reconcile inconsistent findings in prior research. By integrating a prediction-oriented perspective to complement traditional causal inference, this study examines how the family, firm and environmental characteristics jointly predict innovation input and output. Design/methodology/approach Based on a systematic review of 102 empirical studies, this paper constructs a multidimensional framework encompassing 30 predictors across family, firm and external environmental factors. Using Chinese A-share listed family firms from 2010 to 2024, the final samples include 10,078 firm-year observations for R&D investment and 10,145 firm-year observations for R&D output. R&D investment is measured by R&D intensity, while R&D output is measured by patent applications. The study uses the gradient boosting regression tree algorithm. Relative importance measures and partial dependence plots are used to evaluate predictive power and uncover nonlinear boundaries. Rolling-window out-of-sample tests are conducted to assess model performance. Findings The results show that the gradient boosting regression trees (GBRT) model outperforms traditional linear benchmark models in predicting both R&D investment and output. In the integrated model, GBRT achieves rolling-window out-of-sample R-squared values of 30.656% for R&D investment and 43.086% for R&D output, compared with 13.266% and 31.294% for ordinary least squares, respectively. Family ownership involvement, digital transformation, clan culture, ownership balance and industry competition emerge as the core predictors. Crucially, these features exhibit distinct nonlinear tipping points. By identifying the specific ranges where moderate levels are positively associated with innovation and excessive levels may weaken or reverse these associations, the predictive model identifies nonlinear boundary patterns that are consistent with different theoretical expectations in prior research. These nonlinear boundary patterns provide a potential explanation for previously conflicting empirical findings. Practical implications The findings offer actionable implications for multiple stakeholders. For family business managers, the results provide a specific blueprint for balancing family control with appropriate ownership checks and strengthening digital transformation to support innovation investment and patent output. In particular, moderate ownership balance may help support innovation output, whereas excessive family ownership concentration may weaken innovation incentives. For policymakers, the study highlights the necessity of tailoring regional innovation support and technological subsidies to account for localized clan cultures and varying degrees of industry competition. In regions with strong clan culture or intense industry competition, policy support may need to place greater emphasis on governance professionalization, external knowledge access and mechanisms that reduce the risk of innovation investment. Originality/value This study contributes to the literature by introducing a prediction-oriented capability–willingness–pressure framework to family firm innovation research. It moves beyond isolated variable significance to identify the relative predictive importance, nonlinear patterns and boundary conditions of key predictors, thereby helping reconcile previously conflicting empirical findings.

Authors

Institutions

Publication Details

Journal
Journal of Asia Business Studies
Published
2026-11-19
DOI
https://doi.org/10.1108/jabs-02-2026-0085
Primary Topic
Family Business Performance and Succession
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

What factors predict the innovation of Chinese family firms: empirical evidence based on machine learning

Yanzhao Tang, Nan Li
Journal of Asia Business Studies
Family Business Performance and Succession
article

What factors predict the innovation of Chinese family firms: empirical evidence based on machine learning

Yanzhao Tang, Nan Li
article en

Abstract

Purpose Prior studies on family firm innovation have focused primarily on causal explanations and isolated variable significance, but lack a prediction-oriented approach that systematically compares the relative importance of multiple predictors. Drawing on the capability–willingness–pressure framework, this paper aims to identify the key predictors of innovation in Chinese family firms and to reconcile inconsistent findings in prior research. By integrating a prediction-oriented perspective to complement traditional causal inference, this study examines how the family, firm and environmental characteristics jointly predict innovation input and output. Design/methodology/approach Based on a systematic review of 102 empirical studies, this paper constructs a multidimensional framework encompassing 30 predictors across family, firm and external environmental factors. Using Chinese A-share listed family firms from 2010 to 2024, the final samples include 10,078 firm-year observations for R&D investment and 10,145 firm-year observations for R&D output. R&D investment is measured by R&D intensity, while R&D output is measured by patent applications. The study uses the gradient boosting regression tree algorithm. Relative importance measures and partial dependence plots are used to evaluate predictive power and uncover nonlinear boundaries. Rolling-window out-of-sample tests are conducted to assess model performance. Findings The results show that the gradient boosting regression trees (GBRT) model outperforms traditional linear benchmark models in predicting both R&D investment and output. In the integrated model, GBRT achieves rolling-window out-of-sample R-squared values of 30.656% for R&D investment and 43.086% for R&D output, compared with 13.266% and 31.294% for ordinary least squares, respectively. Family ownership involvement, digital transformation, clan culture, ownership balance and industry competition emerge as the core predictors. Crucially, these features exhibit distinct nonlinear tipping points. By identifying the specific ranges where moderate levels are positively associated with innovation and excessive levels may weaken or reverse these associations, the predictive model identifies nonlinear boundary patterns that are consistent with different theoretical expectations in prior research. These nonlinear boundary patterns provide a potential explanation for previously conflicting empirical findings. Practical implications The findings offer actionable implications for multiple stakeholders. For family business managers, the results provide a specific blueprint for balancing family control with appropriate ownership checks and strengthening digital transformation to support innovation investment and patent output. In particular, moderate ownership balance may help support innovation output, whereas excessive family ownership concentration may weaken innovation incentives. For policymakers, the study highlights the necessity of tailoring regional innovation support and technological subsidies to account for localized clan cultures and varying degrees of industry competition. In regions with strong clan culture or intense industry competition, policy support may need to place greater emphasis on governance professionalization, external knowledge access and mechanisms that reduce the risk of innovation investment. Originality/value This study contributes to the literature by introducing a prediction-oriented capability–willingness–pressure framework to family firm innovation research. It moves beyond isolated variable significance to identify the relative predictive importance, nonlinear patterns and boundary conditions of key predictors, thereby helping reconcile previously conflicting empirical findings.

Journal of Asia Business Studies
Xiamen University (CN)
Industry, innovation and infrastructure
Family Business Performance and Succession
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon