This chapter presents the dataset and the sample used for the estimation. The thesis aims to analyze the impact of joining the EU on the V4 countries and therefore, the macroeconomic dataset is composed of standard economic growth predictors. Two data sources were used, World Bank’s Databank and United Nations Development Programme’s Human Development Report. Further explanation of the variables and the sources of the variables is attached in Appendix A: Data description.
The selection of the outcome variables and predictors was based on Abadie, Diamond and Hainmueller (2015) complemented by several new predictors. The country-level data are collected for the period 1991-2017, depending on the country’s data availability14. The approximately decade long pre-treatment period should retrieve satisfactory results. In the previous research were used two possible years as a treatment.
The year of accession, 2004, or the year of signing the accession documents, 1998 or 1999. The latter option provides too short pre-treatment period, which may lower the credibility of the results, therefore year 2004 was used as a treatment. The countries share the same donor pool and predictors to maintain the same conditions and environment for each of the countries15.
The credibility of the results is checked by three robustness tests16. First robustness check is conducted by adding Iceland, Norway and Switzerland to the dataset. Those countries were firstly omitted from the donor pool as having signed trade agreements with the EU. The second check was conducted by leaving-one-out re-analysis of the sample, based on Abadie (2019). For each of the SCM estimates, the country with highest value assigned was dropped from the dataset. The third robustness check tests the robustness of the estimates on larger donor pool that consists of 37 countries17. The results of the robustness checks can be found at the end of each country’s results section18.
14 For the Czech Republic and Hungary, the data are retrieved from 1991, for Slovakia from 1992 and for Poland from 1994. This gives 13 years, 12 years and 9 years long pre-treatment panel, respectively.
15 The pre-intervention period length differs for the treated countries due to different data availability for each of the countries.
16 A robustness check proposed by Abadie (2019) with backdating the treatment was not used as the pre-treatment period for this check is not long enough.
17 The added countries for this robustness check are United States, Guatemala, Indonesia, Mexico, Mongolia, South Africa, Iceland, Norway and Switzerland.
18 In the graphs, robustness check of leaving one country out of the dataset is called "leave-one-out", for adding the three countries with EU trade agreements, the variable is called "add" and for larger donor pool, the robustness check is called "different donor pool".
‧
Out of the world-level data were selected for further analysis only those with enough data and no missing observations for the outcome variables. Those that joined the EU in the pre-treatment or post-treatment period had to be omitted. Following previous methodology introduced by Böwer and Turrini (2010), those countries with large dependency on oil production and least-developed countries were omitted from the dataset. Furthermore, those countries for which the total sum of the GDP components did not match the GDP level, were omitted. Switzerland, Iceland and Norway were omitted from the donor pool sample to avoid including countries that are economically similar to those in EU. The final donor pool consists of the below 28 countries.
Table 1: List of donor pool countries
Albania Korea
‧
based on Abadie et al. (2015), complemented by several new variables to achieve the best fit of the synthetic predictors of the outcome variables in the pre-intervention period. Moreover, the same set of covariates and outcome variables was used for all four V4 countries.Outcome Variables
The primary objective of this thesis is to estimate the effect of joining the EU on the GDP per capita levels of the V4 countries. Therefore, the main outcome variable is GDP per capita in Purchasing Power Parity (PPP) in constant 2011 international dollars (USD)19. The choice of the outcome variable follows the previous literature (Abadie &
Gardeazabal, 2003; Abadie et al., 2015).
Furthermore, to determine the effects on the GDP components, SCM was also build for each of the GDP components. The components were calculated from the above-mentioned GDP per capita and the component share in GDP. Therefore, the additional outcome variables are import per capita, export per capita, government spending per capita, investment per capita, private consumption per capita and lastly, net export per capita.
Treatment
The chosen treatment is joining the EU. The V4 countries joined the EU in 2004 and this year is also selected as the treatment. Furthermore, the V4 countries are the only in the dataset to receive the treatment over the studied period. Due to this, countries with relatively similar economic environment, such as Romania or Croatia, had to be omitted from the donor pool as they joined the EU during the post-treatment period and thus, would provide inaccurate results. Countries that have not joined the EU, however, are strongly connected with the EU via trade agreements, were also excluded from the dataset20.
19 In the following text, USD refers to constant 2011 international dollars in PPP.
20 Those countries are Iceland, Norway and Switzerland. Nonetheless, they were used for the robustness check.
‧
Gardeazabal, 2003; Abadie et al., 2015), complemented by new variables to provide a better match of the model21. Some variables used in the above-mentioned sources were omitted as they are later used as an outcome variable22. Below is list of the variables used in the model. The same set of predictors was used for GDP and its components to assure the same sources of changes in the outcome.
Table 2: List of predictors
Age dependency ratio as % of working-age population Agriculture, value added as % of GDP
Human Development Index23 Industry, value added as % of GDP Inflation, consumer prices, annual %24
Labor force participation rate, % of total population ages 15-64 Population growth, annual %
Source: Created by author