@ARTICLE{26543120_547235251_2021, 
	author = {L. Hoang and M. Chatterjee and K. Nguyen and D. Nguyen}, 
	keywords = {, country classification, multidimensional clustering technique, public expenditure, Scree PlotSilhouette analysis},
	title = {Classifying Countries in Terms of Government Expenditure: A Multi-criteria Approach},
	journal = {Экономический журнал ВШЭ},
	year = {2021},
	volume = {25},
	number = {4},
	pages = {610-627},
	url = {https://ej.hse.ru/2021-25-4/547235251.html},
	publisher = {},
	abstract = {This present article contributes to the literature on country classification by investigating the pattern of government expenditure across countries. Here we adopt a k-means clustering technique on indicators reflecting the actual situation of government expenditures. The variables used for the classification include GNI per capita, government effectiveness, subsidies, and other transfers, compensation of employees, goods and services expense. After implementing the unsupervised classification technique, four clusters of countries were selected. We have identified variables that vary less within a particular cluster, though the variation is higher between clusters. Our algorithm allows dropping those variables which are less relevant for identifying the clusters. By allowing for multi-criteria simi­larities in the classification, the possibility of sourcing differences in government expenditure and economic growth across countries could be enhanced. Also, our proposed method helps to study the nature and characteristics of each cluster and thereby describe the clusters in terms of the variables involved. The homo­geneous clusters of countries that are being observed are likely to help future public expenditure-related research. Our study can help establish a relationship between govern­ment expenditure and economic growth in the future.},
	annote = {This present article contributes to the literature on country classification by investigating the pattern of government expenditure across countries. Here we adopt a k-means clustering technique on indicators reflecting the actual situation of government expenditures. The variables used for the classification include GNI per capita, government effectiveness, subsidies, and other transfers, compensation of employees, goods and services expense. After implementing the unsupervised classification technique, four clusters of countries were selected. We have identified variables that vary less within a particular cluster, though the variation is higher between clusters. Our algorithm allows dropping those variables which are less relevant for identifying the clusters. By allowing for multi-criteria simi­larities in the classification, the possibility of sourcing differences in government expenditure and economic growth across countries could be enhanced. Also, our proposed method helps to study the nature and characteristics of each cluster and thereby describe the clusters in terms of the variables involved. The homo­geneous clusters of countries that are being observed are likely to help future public expenditure-related research. Our study can help establish a relationship between govern­ment expenditure and economic growth in the future.}
}