I'm working on my thesis project and wondering which fixed effects to include in my panel data analysis. My dataset is of dyadic structure (one column indicates donor country and one column indicates recipient country of official development aid). Naturally, the dyads are somewhat codependent: In year 2003, France will commit aid to Kenya, but also to Egypt, and since France is the donor in both cases, many of the values of explanatory variables (GDP of donor, certain policies of donor) will be identical. My research question focuses on the impact of a certain policy on donor behaviour. For example, in the case of France, this policy was adopted in 2008 and within my range 2002-2019, I expect this adoption to have an impact on France's aid allocation. Some of the explanatory variables do not change a lot over time (e.g. political constitution). I could include year fixed effects, dyad fixed effect, or add either donor or recipient dummies. Since some variables show limited variation for one dyad, I risk multi collinearity with the fixed effects. Since my aim is to investigate change over time, I'm not sure it makes sense to include year fixed effects - at the same time, I'm only interested in change in one specific variable (the policy change), that might not be an issue. I'm happy to hear your ideas on this. Thanks a lot!
