The article "Back to the Future: Modeling Time Dependence in Binary Data" by David B. Carter and Curtis S. Signorino discusses the methods for modeling time dependence in binary data, focusing on the use of time dummies and splines. The authors argue that while both methods are widely used, they have significant limitations. Time dummies can lead to estimation problems due to separation, especially when the data has a decreasing hazard, while splines can be complex and may not be interpreted correctly by substantive researchers. They propose a simpler alternative: including \( t \), \( t^2 \), and \( t^3 \) in the regression model, which is a cubic polynomial approximation. This method avoids the issues of time dummies and splines, is easier to implement and interpret, and performs well in Monte Carlo simulations. The authors also demonstrate how this method can accommodate nonproportional hazards and provide new empirical support for historical-institutional perspectives in political science research.The article "Back to the Future: Modeling Time Dependence in Binary Data" by David B. Carter and Curtis S. Signorino discusses the methods for modeling time dependence in binary data, focusing on the use of time dummies and splines. The authors argue that while both methods are widely used, they have significant limitations. Time dummies can lead to estimation problems due to separation, especially when the data has a decreasing hazard, while splines can be complex and may not be interpreted correctly by substantive researchers. They propose a simpler alternative: including \( t \), \( t^2 \), and \( t^3 \) in the regression model, which is a cubic polynomial approximation. This method avoids the issues of time dummies and splines, is easier to implement and interpret, and performs well in Monte Carlo simulations. The authors also demonstrate how this method can accommodate nonproportional hazards and provide new empirical support for historical-institutional perspectives in political science research.