This study investigates whether earnings management in its two forms (accruals and real activities manipulation) can improve bankruptcy prediction models. In particular, it examines whether special information extracted from earnings management, including potential manipulations of the reported earnings found in financial statements, might improve the accuracy of bankruptcy prediction models. It applies earnings management–based models, based on financial ratios and earnings management variables, to a sample of 6,000 French small and medium-size enterprises, then compares the classification rates obtained by these models with a model based solely on financial ratios. This study thus makes several contributions by (1) investigating novel predictors, accruals, and real activities manipulation variables, in the context of bankruptcy prediction modeling; (2) enabling analyses of the contribution of earnings management–based variables, in the form of static and dynamic indicators, to model performance; (3) revealing the influence of these variables on the forecasting horizon of bankruptcy prediction models (one- to three-year horizon); and (4) establishing that earnings management information provides a complementary explanatory variable for enhancing model performance.

本研究调查两种形式的盈余管理（应计和实际活动操纵）是否可以改进破产预测模型。特别是，它检查了从盈余管理中提取的特殊信息，包括对财务报表中报告盈余的潜在操纵，是否可以提高破产预测模型的准确性。它将基于财务比率和盈余管理变量的基于盈余管理的模型应用于 6,000 家法国中小企业的样本，然后将这些模型获得的分类率与仅基于财务比率的模型进行比较。因此，本研究通过 (1) 调查新的预测变量、应计项目和实际活动操作变量做出了多项贡献，在破产预测建模的背景下；(2) 能够以静态和动态指标的形式分析基于盈余管理的变量对模型绩效的贡献；(3) 揭示这些变量对破产预测模型预测范围（1-3 年范围）的影响；(4) 确定盈余管理信息为提高模型性能提供了补充解释变量。