BACKGROUND: The COVID-19 pandemic could increase violence against children at home. However, collecting empirical data on violence is challenging due to ethical, safety, and data quality concerns. OBJECTIVE: This study estimated the anticipated effect of COVID-19 on violent discipline at home using multivariable predictive regression models. PARTICIPANTS: Children aged 1-14 years and household members from the Multiple Indicator Cluster Surveys (MICS) conducted in Nigeria, Mongolia, and Suriname before the COVID-19 pandemic were included. METHODS: A conceptual model of how the COVID-19 pandemic could affect risk factors for violent discipline was developed. Country specific multivariable linear models were used to estimate the association between selected variables from MICS and a violent discipline score which captured the average combination of violent disciplinary methods used in the home. A review of the literature informed the development of quantitative assumptions about how COVID-19 would impact the selected variables under a "high restrictions" pandemic scenario, approximating conditions expected during a period of intense response measures, and a "lower restrictions" scenario with easing of COVID-19 restrictions but with sustained economic impacts. These assumptions were used to estimate changes in violent discipline scores. RESULTS: Under a "high restrictions" scenario there would be a 35%-46% increase in violent discipline scores in Nigeria, Mongolia and Suriname, and under a "lower restrictions" scenario there would be between a 4%-6% increase in violent discipline scores in these countries. CONCLUSION: Policy makers need to plan for increases in violent discipline during successive waves of lockdowns.