engleski

Regression analysis of e-commerce: focusing the selected EU candidates and the EU countries

Development level of a country may be measured not only in a traditional way by using the Gross Domestic Product (GDP) per capita, but by introducing some modern measures. One of such measures is the percentage of individuals who had the last online purchase recently, i.e. the e-commerce adoption level, which shows what share of the people in a country use an option to do a transaction of buying or selling online during a particular time interval. The aim of the paper is to investigate the impact of several variables on the e-commerce adoption level in selected European Union (EU) candidates (the Former Yugoslav Republic of Macedonia, Serbia and Turkey) and the EU-28 member states, as well. In the paper the e- commerce adoption level, as given in the Eurostat methodology, is estimated by the variable Internet purchases by individuals, aged 16 to 74, who had last online purchase in previous 12 months. The impact is observed by using regression modelling where the All Possible Regression approach was applied. As explanatory variables in the regression modelling, the four variables have been used: GDP per capita in Purchasing Power Standards (PPS), government expenditure on education, Internet penetration rate and Internet skills level of individuals. Initially, 15 Ordinary Least Squares linear regression models were developed. However, only three regression models were focused and built in details. In the process of model selection, three regression models satisfied two conditions, which were pre-set by the researchers, and these were as follows: all explanatory variables are found to be statistically significant at the 5% significance level, and they have the lowest Mallow's Cp indicator value among the regression models with the same number of explanatory variables. The regression assumptions were satisfied by all three selected linear regression models. All four explanatory variables appeared at least at one of selected models. Furthermore, in all three observed regression models the signs of the parameter estimates at explanatory variables have been positive. On that way, it can be concluded that the increase in the explanatory variable should also have a positive effect on the e-commerce level or on the development level of a country, respectively. In the future research the development of the e-commerce levels in the EU member states and the EU candidates should be observed in different time points. Unfortunately, some data were not available and, because of that, some estimates have been produced, which limits the possible conclusions.

all possible regression approach ; linear regression model ; Mallow's Cp ; Ordinary Least Squares estimation ; regression diagnostics ; regression predicting

This work has been in part supported by the Croatian Science Foundation within the project STRENGTHS (project no. IP-2013-9402). Abstract je objavljen u Book of Abstracts, ISBN: 978-961- 6997-03-4