Dear Statalist,
I wonder if I could get advice on the following question.
I'm currently examining whether Ministers who have discretionary power over capital investment in sports facilities, target neighbouring constituencies. To date studies have found constituencies which Ireland's Minister of Sport and Finance reside in, perform particularly well in regards to investment in sports facilities - as they control the purse strings.
However, what I wish to examine is whether neighbouring constituencies also perform well i.e. are there spatial spillovers?
My data is currently in panel form, with my dependent variable alternating between a count of projects awarded to a constituency in a particular year as well as the combined monetary total each constituency received in any given year. My two main indo variables are the distance between the centre of the Minister of Sport's constituency, to each other constituency. This is the same for the Minister of Finance. I also have a number of controls, relating to population etc.
What my issue is relating to is how would I control for the two Ministerial constituencies? When I generate my distances, they appear as 0 so is it best to leave them as such or convert them to 0.1?
Through Moran's I I note the presence of spatial autocorrelation for both my dependent variables, however I'm only particularly interested in spillovers near Ministerial constituencies. Would I be better to generate a LISA cluster map also? (Note - my matrix is a binary contiguity one)
Any advice would be most welcome.
Best,
Sean
I wonder if I could get advice on the following question.
I'm currently examining whether Ministers who have discretionary power over capital investment in sports facilities, target neighbouring constituencies. To date studies have found constituencies which Ireland's Minister of Sport and Finance reside in, perform particularly well in regards to investment in sports facilities - as they control the purse strings.
However, what I wish to examine is whether neighbouring constituencies also perform well i.e. are there spatial spillovers?
My data is currently in panel form, with my dependent variable alternating between a count of projects awarded to a constituency in a particular year as well as the combined monetary total each constituency received in any given year. My two main indo variables are the distance between the centre of the Minister of Sport's constituency, to each other constituency. This is the same for the Minister of Finance. I also have a number of controls, relating to population etc.
What my issue is relating to is how would I control for the two Ministerial constituencies? When I generate my distances, they appear as 0 so is it best to leave them as such or convert them to 0.1?
Through Moran's I I note the presence of spatial autocorrelation for both my dependent variables, however I'm only particularly interested in spillovers near Ministerial constituencies. Would I be better to generate a LISA cluster map also? (Note - my matrix is a binary contiguity one)
Any advice would be most welcome.
Best,
Sean
