Comparative politics and the synthetic control method revisited: a note on Abadie et al. (2015)

Swiss Journal of Economics and Statistics

Table 1 Results (predictor weights V, donor weights W for main application and in training period, cross-validation criterion) obtained in different ways

		ADH	Orig.	Alph.	Min.	Max.
V weights	GDP per capita	44.2	71.0	84.5	38.5	87.7
	Trade openness	13.4	7.9	5.1	4.4	14.6
	Inflation rate	7.2	2.7	0.5	0.0	8.1
	Industry share	0.1	0.9	1.2	0.0	1.3
	Schooling	10.7	6.4	4.2	3.7	11.6
	Investment rate	24.5	11.1	4.5	2.9	27.2
W weights main	USA	21.9	22.1	22.0	20.1	22.1
	UK	0.1	0.0	0.0	0.0	1.8
	Austria	41.8	42.2	43.1	41.6	47.1
	Netherlands	9.0	9.2	8.5	0.6	9.4
	Norway	0.1	0.0	0.0	0.0	1.9
	Switzerland	11.1	10.9	10.8	10.8	13.7
	Japan	15.5	15.7	15.4	9.9	15.7
W weights training	USA	13.5	13.5	13.5	13.5	13.5
	Austria	50.7	50.7	50.7	50.7	50.7
	Switzerland	16.6	16.6	16.6	16.6	16.6
	Japan	14.6	14.6	14.6	14.6	14.6
	Australia	4.5	4.5	4.5	4.5	4.5
C-V criterion	RMSPE	67.7	67.7	67.7	67.7	67.7

Notes: “ADH” stands for the results of Abadie et al. (2015), “Orig.” are results from Stata with the same ordering of donors as in the code of Abadie et al. (2015), “Alph.” denotes results from Stata with donors sorted alphabetically, “Min.” and “Max.” denote minimal and maximal values, respectively, found under the condition that the corresponding predictor weights V lead to identical donor weights W in the training period. All numbers are given in percent, suppressing donors with weight less than 1%