Towards forecasting in management studies: insights from operations research and empirical evidence on gender and motherhood income gap View supplementary material

Management theories and models aim to predict future states and outcomes. Yet, as management scholars, we often tend to prioritise model fit metrics over prediction and forecasting, assuming that strong model fit inherently leads to accurate predictions. We challenge this assumption, arguing that an exclusive focus on model fit can yield theories that fail to generalise to new datasets, thereby limiting their forecasting accuracy and practical relevance. In a systematic review of 6,514 studies, we find a pronounced dominance of model fit approaches. Model fit metrics are susceptible to overfitting, where models capture noise rather than patterns, and underfitting, where key relationships are overlooked. Both problems undermine predictive performance. Drawing on insights from operations research, we apply newly developed forecasting metrics to address these limitations. Empirically examining the gender gap and motherhood penalty in returns from employment and entrepreneurship, we demonstrate how these metrics can complement traditional fit measures. By integrating multiple assessment metrics, we offer a comprehensive framework for improving both predictive accuracy and theoretical development in management research. We provide the Stata syntax that scholars can download and use to assess the forecasting ability of their models.