Background <p>A recent review of research on the association between school-related stress and internalizing problems found growing scholarly interest in the topic, but also raised questions concerning the quality and reliability of the existing knowledge base. The aim of this study was to investigate how robust the association between school-related stress and internalizing problems is to differences in model specifications.</p> Methods <p>Longitudinal survey data from between 2,991 and 4,845 Swedish adolescent students aged 13–16 years were used. A total of 57,322 different models were estimated, varying the choice of sample, measure of internalizing problems, functional form, statistical method, and combinations of included control variables. The results were summarized using specification curve analysis.</p> Results <p>Most estimates of the association between school-related stress and internalizing problems were statistically significant at the 5% level. The choice of sample, outcome, functional form, and control variables had a limited influence on the size and significance of the estimates, but the estimates were markedly smaller and mostly non-significant in models investigating lagged effects.</p> Conclusions <p>This study showed that school-related stress is a robust predictor of internalizing problems as long as the association is assumed to be contemporaneous, while evidence for lagged effects was weaker. A key conclusion is that the choice of whether to estimate lagged or contemporaneous effects may be the most consequential in studies on school-related stress and internalizing problems or similar topics.</p>