This may have to do with you unbalanced panel. You also have gaps in your time series, which you'll need to interpolate. I know for my SCM command at least, I ensure the users have a balanced panel of observations before they can do anything.
Beyond this though, now that I can see pretty much everything, I have more fundamental concerns. Consider the target time series... or rather, in fact, let's jut consider the more general time series of everything.
I don't like being the bearer of bad news, but, I don't think SCM is an appropriate design for this question. For any of these variables, I should say. The outcomes are super noisy. You have 9 preintervention time periods... the panels aren't balanced. The results you would get even if we got it to run would be extremely overfit to noisy fluctuations and not be representative of the underlying DGP. Even if I do it with insurance
or Bank Accounts
the results become even more grim. I really don't like saying this or suggesting it even, because I feel like people should be able to study whatever they want to study, but if I were advising you (by the way, no need to listed to me at all, I'm just a PhD student who happens to love SCM and its applications), I would get another topic and start fresh. You don't need to switch the intervention, you can still use the earthquake if you'd like, but whatever you do, don't use this particular dataset, find another outcome or dataset that has less problems associated with it.
I'm not trying to discourage you or be mean at all, but that's my view on it after having looked at it. Just as a suggestion, a simpler, less messy topic could be its effects on GDP in general. You could argue that since Ica's was among the biggest earthquakes in Peruvian history, that it could have widespread social impacts. And, we can look at GDP per Capita for this
Less messy dataset. A readily available outcome variable. Control variables easily collected... simpler. Again, you need not abandon your project if you don't want to, but it is what I would do, unfortunately.
Beyond this though, now that I can see pretty much everything, I have more fundamental concerns. Consider the target time series... or rather, in fact, let's jut consider the more general time series of everything.
Code:
line Elec month if inrange(region,1,25), connect(L) /// <-- the key option lcolor("gs12") || line Elec month if region ==0, lcol(black),, /// legend(order(1 "Donors" 2 "Target") ring(0) pos(11)) xli(`=tm(2068m5)')
Code:
line Ins month if inrange(region,1,25), connect(L) /// <-- the key option lcolor("gs12") || line Ins month if region ==0, lcol(red),, /// legend(order(1 "Donors" 2 "Target") ring(0) pos(11)) xli(`=tm(2068m5)' , lwidth(thick) lpat(solid))
Code:
line Ba month if inrange(region,1,25), connect(L) /// <-- the key option lcolor("gs12") || line Ba month if region ==0, lcol(red),, /// legend(order(1 "Donors" 2 "Target") ring(0) pos(11)) xli(`=tm(2068m5)' , lwidth(thick) lpat(solid))
I'm not trying to discourage you or be mean at all, but that's my view on it after having looked at it. Just as a suggestion, a simpler, less messy topic could be its effects on GDP in general. You could argue that since Ica's was among the biggest earthquakes in Peruvian history, that it could have widespread social impacts. And, we can look at GDP per Capita for this
Code:
import delim "https://raw.githubusercontent.com/lukes/ISO-3166-Countries-with-Regional-Codes/master/all/all.csv", clear drop countrycode rename alpha3 countrycode tempfile codes sa `codes', replace cls u "https://www.rug.nl/ggdc/historicaldevelopment/maddison/data/mpd2020.dta", clear //replace country = "Tanzania" if strpos(country,"U.R.") keep if inrange(year,1990,2015) cls merge m:1 countrycode using `codes', keepusing(region) keep(3) nogen keep if region=="Americas" egen id = group(countrycode) line gdp year if id != 21, connect(L) /// <-- the key option lcolor("gs12") || line gdp year if id ==21, lcol(red),, /// legend(order(1 "Donors" 2 "Peru") ring(0) pos(11)) xli(2007 , lwidth(thick) lpat(solid))
Comment