I am writing my dissertation and will use a binary outcome variable (decision of a country to sanction other states or not). I was planning to use a logit model, but I am not sure which countries should be included in the sample.

Should the sample comprise all countries where there has been a sanction episode at some point, for all years (regardless of sanction event of that year or not, 0/1)?
Or should the sample also include countries where there has been no sanctions at all ever? (that is, Y=0 always). I hope somebody can help me, bc I am totally lost...

Thanks a lot in advance!!