I have a conflict dataset across the country and time. The authors of the dataset (say A) collects conflict information from different articles. Those articles list countries when the conflict started, when it ended or at the time of publication of that article, whether the conflict was ongoing.
I want to see the impact of conflict on outcome variable y. The problem arises when I try to assign the binary treatment because I don’t have information on whether the conflict was ongoing until 2019. For example, the authors "A" collected conflict information from an article which stated: conflict occurred in 2005 and it was still ongoing when the article was published, say the article got published at the end of 2005. So I can confidently assign 1 to 2005 but what about 2006-2020? I do not know whether conflict was ongoing at these periods, so I cant assign 1. But I am thinking whether I should just assign 1 for 2005 and then 0 for the upcoming years of 2006-2020 (also assign 0 for pre-2015) and mention this as the limitation of my study? I am thinking this way because we have advanced DID design which can take account of dynamic effects (present outcome depends on past as well as present treatment). Because here treatment switches on and off in multiple countries and also we need to take care of dynamic effects, so I am thinking of Liu et al. (2021): A Practical Guide to Counterfactual Estimators for Causal Inference with Time-Series Cross-Sectional. But any other suggestions would be greatly appreciated.
Or should I leave the design because there is no way around it to make it a counterfactual analysis?
I want to see the impact of conflict on outcome variable y. The problem arises when I try to assign the binary treatment because I don’t have information on whether the conflict was ongoing until 2019. For example, the authors "A" collected conflict information from an article which stated: conflict occurred in 2005 and it was still ongoing when the article was published, say the article got published at the end of 2005. So I can confidently assign 1 to 2005 but what about 2006-2020? I do not know whether conflict was ongoing at these periods, so I cant assign 1. But I am thinking whether I should just assign 1 for 2005 and then 0 for the upcoming years of 2006-2020 (also assign 0 for pre-2015) and mention this as the limitation of my study? I am thinking this way because we have advanced DID design which can take account of dynamic effects (present outcome depends on past as well as present treatment). Because here treatment switches on and off in multiple countries and also we need to take care of dynamic effects, so I am thinking of Liu et al. (2021): A Practical Guide to Counterfactual Estimators for Causal Inference with Time-Series Cross-Sectional. But any other suggestions would be greatly appreciated.
Or should I leave the design because there is no way around it to make it a counterfactual analysis?
