Longitudinal dataset of political issue-positions of 411 parties across 28 European countries (2009–2019) from voting advice applications EU profiler and euandi

Abstract

This data article provides a descriptive overview of the “EU Profiler/euandi trend file (2009–2019)“ dataset and the data collection methods. The dataset compiles party position data from three consecutive pan-European Voting Advice Applications (VAAs), developed by the European University Institute for the European Parliament elections in 2009, 2014 and 2019. It includes the positions of 411 parties from 28 European countries on a wide range of salient political issues. Altogether, the dataset contains more than 20 000 unique party positions. To place the parties on the political issues, all three editions of the VAA have used the same iterative method that combines party self-placement and expert judgement. The data collection has been a collective effort of several hundreds of highly trained social scientists, involving experts from each EU member state. The political statements that the parties were placed on, were identical across all the countries and 15 of the statements remained the same throughout all three waves (2009, 2014, 2019) of data collection. Because of the unique methodology and the large volume of data, the dataset offers a significant contribution to the research on European party systems and on party positioning methodologies.

Specifications table

Subject	Sociology and Political Science
Specific subject area	Comparative Politics; Political Issues; Party Placement
Type of data	Table
How data were acquired	Party positions on the issue statements were determined via an iterative procedure, combining party self-placement and expert judgement. Data were acquired prior to 2009, 2014 and 2019 European Parliament elections. The identical list of statements was translated to each European Union country's official language (23 in total).
Data format	Raw
	analysed
	Filtered
Parameters for data collection	The parties selected into the sample had to be relevant in the respective country, indicated by seat(s) in the national and/or European Parliament or being credited with at least 1% of the popular vote in recent polls. These parties were placed on the list of political statements on a 5-point Likert scale, ranging from complete agreement to a complete disagreement with the statement (and a “no opinion option”).
Description of data collection	Prior to the European Parliament elections in 2009, 2014 and 2019, the selected parties were contacted by e-mail (in their native language) and invited to self-place their party on the attached list of political statements. The parties were also placed on the same list of statements by a group of experts from each country. The placements were then compared and, if necessary, calibrated in co-operation with the party representatives. If the party did not participate in the self-placement procedure, the positions were determined solely by the country expert teams.
Data source location	27 European Union member states and the United Kingdom.
Data accessibility	The dataset with the accompanying codebook is deposited at Cadmus (The European University Institute Research Repository), where it is freely available to download.
	Direct URL to data:https://cadmus.eui.eu/handle/1814/65944

Value of the data

• •The data can be used for cross-national comparisons of European parties and party systems, as it uses an identical set of survey items across 28 countries
• •The data can be used for longitudinal comparisons of European parties and party systems, as it provides data from three different time points over a ten-year period (2009, 2014, 2019).
• •The data is highly valuable to the research on party positioning methodologies. It can be studied from the perspective of its internal reliability, but also for cross-validation with other party position datasets that have used different methodologies to place the parties. To facilitate cross-validation, the dataset contains a variable that allows it to be easily merged with the Chapel Hill Expert Survey (CHES).

Data description

The dataset “EU Profiler/euandi trend file (2009–2019)“ consists of one data file (provided in both Excel and Stata format) and a Codebook that provides information about each variable and lists all the parties across countries. The dataset compiles party position data from three subsequent pan-European Voting Advice Application (VAA) projects, each corresponding to one European Parliament election (2009, 2014, 2019). As all the countries that were members of the European Union at the time of the respective election were included in the VAA, it contains party positions from 28 countries in three different time points.1 The unit of analysis in the dataset is party-year, i.e. one party at one election (e.g. Austrian Green Party at 2009 election). The number of party-year observations across the three waves is 768 and the whole dataset contains 20,872 unique party positions (i.e. the position of one party in one year on one political issue-statement). That makes it - to the best of our knowledge - the largest dataset of VAA-based party placements. Also, it is the first VAA dataset that allows for both cross-national and longitudinal comparisons of party positions (cf. Gemenis et al. 2019 [1]).

The dataset includes all the relevant parties in each of the 28 countries. The final selection of the parties from each country was at the discretion of respective country expert teams that were instructed not to exclude any parties that bared at least some chance of achieving representation in the upcoming EP election. In any case, all the parties that were credited with at least 1% of the popular vote in the polls preceding the EP election were considered for data collection. The number of parties per country (across all three waves) ranges from 5 (Malta) to 23 (Slovenia). Altogether, the dataset contains 411 parties, each wave separately covering around 240–270. Some of the parties were present in all three waves (141), some that were founded or disappeared from the political arena after 2009 were included only once (195) or twice (75). The name of each party at a given election is presented in both English and the original language. We have also provided a party identifier variable (“PUI“ – Party Unique Identifier) that takes a different value for each unique party. This is especially important regarding the parties that have changed their name over time.2

In each data collection wave, the parties across all the countries were positioned on a set of identical political statements. In 2009 and 2014, the number of statements was 28, while in 2019 it was 22. Similarly to the parties, some of the political statements remained unchanged throughout all three waves, whereas some were replaced. Altogether, the dataset contains party positions on 42 different statement; 15 of these were present in all three waves, allowing for a direct comparison over a time span of ten years (these 15 statements are displayed in Table 2 in the next section). Each statement in each year is represented by a separate variable, e.g. statement number 18 in year 2009 is in the dataset captured by variable S18_09. In 2009 and 2014, also two country-specific statements (selected by each country's expert team separately) were included in every country. In the dataset, the variables corresponding to these statements are S29_09, S30_09, S29_14 and S30_14. Thus, these variables cannot be used for cross-national comparison, as they signify different political issues across countries. The exact wording of all the cross-national and country-specific statements is documented in the Codebook.

Table 2.

15 continuous issue-statements across the three waves of data collection (2009–2019).

Statement
Social programs should be maintained even at the cost of higher taxes.
Government spending should be reduced in order to lower taxes.
Immigration into the country should be made more restrictive.
Immigrants from outside Europe should be required to accept our culture and values.
The legalisation of same sex marriages is a good thing.
The legalisation/decriminalisation of the personal use of soft drugs is to be welcomed.
Euthanasia should be legalised.
Criminals should be punished more severely.
Renewable sources of energy (e.g., solar or wind energy) should be supported even if this means higher energy costs.
The promotion of public transport should be fostered through green taxes (e.g., road taxing).
The EU should acquire its own tax raising powers.
On foreign policy issues, [such as the relationship with Russia], the EU should speak with one voice.1
The European Union should strengthen its security and defence policy.
European integration is a good thing.
Individual member states of the EU should have less veto power.

¹In 2009, this statement was “On foreign policy issues, such as the relationship with Russia, the EU should speak with one voice”. In 2014 and 2019, it was “On foreign policy issues, the EU should speak with one voice”.

On each issue, the parties were placed on a 5-point Likert scale, based on their degree of (dis)agreement with the respective statement. The answer categories and the corresponding numerical values in the dataset are: Completely disagree = 0; Tend to disagree = 25; Neutral = 50; Tend to agree = 75; Completely agree = 100. In case the party had no discernible position on the statement, it was coded as having “No opinion“.

The main parameters of the dataset are summarized in Table 1.

Table 1.

Summary of the dataset.

N of data collection waves1	N of countries	N of political parties	N of political statements2	N of parties present in all 3 waves	N of statements present in all 3 waves	Statement answering scale
3	28	411	42	141	15	5-point Likert

¹Each wave corresponds to one European Parliament election (2009, 2014, 2019).

²This signifies only the number of cross-national statements.

A number of additional variables complete the dataset. “SELFPLACEMENT” is a dummy variable indicating whether the party participated in the self-placement process (explained in the subsequent section). Variable “EPVOTE” indicates the vote share of the party in the respective European Parliament election. Variable “EPVOTE_COALITION” indicates the vote share of a party coalition that the respective party was part of, in case the separate party vote share is not discernible. Finally, variable “CHESS” indicates Chapel Hill Expert Survey (CHES) code of the party, which allows to merge these datasets or individual parties therein.

Experimental design and methods

The parties in the dataset were positioned using an iterative method (also known as the Kieskompas method) that combines party self-placement and expert judgement [2,3]. This approach is clearly distinct from other prominent party placement methodologies, such as expert surveys, manifesto coding, party elite surveys, or public opinion and roll-call votes aggregation [4,5].

Although the VAA project changed its name from the original EU Profiler (2009) to euandi (read: “EU and I”, 2014 and 2019), the party placement methodology has remained the same, allowing for direct cross-national and temporal comparisons. As such, the dataset offers a significant contribution to the study of measuring the positions of political parties, which is a highly important endeavour in political science. In this section, we give an overview of the three central parts of this design: (1) Statement selection, (2) Party positioning methodology, and (3) Party positioning process.

Statement selection

The foundations of the data collection project rest on the political statements that the parties are placed on. Selecting the statements is one of the first and most crucial steps when developing a VAA [6,7]. Most importantly, the statements must be sufficiently polarizing, i.e. there are parties that agree with the statement as well as parties that disagree, and salient, i.e. the statements must address the political issues that are important in the respective electoral context. Composing a list of divisive and salient statements in pan-European context is undoubtedly a challenging task [8]. A large number of sources – such as opinion polls, earlier party manifestos, information from experts, academics and journalists – has been utilized to create and revise this list of statements [2].

In the 2014 and 2019 editions of the VAA, an aim to maximize the amount of longitudinal data has also been taken into consideration. Thus, the statements were retained as they were, unless they had lost their salience during the 5-year period between the EP elections or failed to discriminate between the parties. 17 out of 28 statements from 2009 VAA were kept in the survey also for the 2014 data collection wave, while the remaining 11 were replaced. The 2019 project continued with 19 statements from 2014 and introduced 3 new ones, reducing the overall number of statements to 22. Fifteen of the statements remained the same across all three waves from 2009 to 2019. These statements are listed in Table 2.

In addition to capturing parties’ stances on the specific issues (e.g. whether euthanasia should be legalized), most statements were also aimed at measuring broader political dimensions. With very few exceptions, the issue-statements in the dataset were attached either to economic left-right, cultural liberal-conservative or pro-/anti-EU dimension. Factor analyses conducted after the data collection waves have confirmed the described three-dimensional structure of the data and the a priori decisions regarding which statement should be assigned to which dimension have in most cases proven to be valid [8] [9] [10]. The three-dimensional political space visualization that was displayed to the euandi users in 2014 is shown on Fig. 1.

Fig. 1.

The three-dimensional political space visualization in euandi (2014)

Source: Garzia et al. 2015 [9].

Party positioning methodology

All three waves of the project have used the so-called iterative method to position the parties on the given list of statements. This method - first introduced by the Dutch Kieskompas in 2006 [11] – aims to maximize the strengths and alleviate the weaknesses of other commonly used party positioning techniques. The core of the iterative method consists in combining expert judgments with party self-placement. Party representatives are asked to provide the placements of their party on the selected set of political statements. At the same time, a group of country experts positions the parties on the same statements, independently from party self-placement. Then, the placements are compared, and calibrated, ideally in co-operation with the parties (in case they want to engage in such co-operation). If disagreement between the party and the experts over some placements does not get solved during the calibration, the final decision on how to position the party is taken by the country expert teams. In case the party does not react to the self-placement invitation, the whole placement procedure is done solely by the expert coders [9]. The experts work together as a group when placing the parties. Each party is first positioned by at least 2 coders and the final placement is reached via an intragroup deliberative calibration procedure within the group. This differentiates this approach from the classic expert survey methodology of aggregating individual expert placements to reach the estimated party position (e.g. Chapel Hill Expert Survey). In sum, the iterative method is a hybrid of party self-placement and expert evaluations, mitigating the pitfalls related to relying solely on each of these methods [3].

The methodological guidelines also insist that every party position must be justified with supporting evidence from an authorized source [11]. In order to reduce the chance that a party cannot be placed on a given policy statement, the list of potential sources goes well beyond the current election manifesto. The expert coders are given a hierarchical list of sources, ranging from the election manifesto for the respective EP election and other official party documentation to interviews with the top figures of party and other media coverage. In case parties themselves do not provide supporting evidence for their self-placement, the experts are obliged to find a justification (an extracted text snippet from any of the sources) for each position. Thus, the iterative method constitutes an advancement compared to expert and elite surveys that in most cases do not require justifying the determined position with a citation,3 and compared to manifesto coding method by relying on a wider set of sources than just the current election manifesto. The hierarchy of data sources used for party placement is displayed in Table 3.

Table 3.

Hierarchy of data sources for party placement.

	Data Source
i.	EU Election Manifesto [respective election year] of national party
ii.	Party Election Platform
iii.	Current/latest national election manifesto
iv.	EU Election Manifesto [respective election year] of Europarties
v.	Other programmatic and official party documentation
vi.	Actions/statements of party representatives in government and parliament1
vii.	Interviews and other coverage in media outlets2
viii.	Older Election Manifestos, party documentation, statements and interviews
ix.	Other sources

¹This category was removed from the list for the 2019 data collection wave, as it proved to be too wide and partially overlapped with category viii below.

²In 2019, this category was rephrased as “Interviews, press releases and social media communication by party leader and leading candidates”.

As explained in the previous section, the attitudes of the parties are measured on a 5-point Likert scale, capturing not only whether the party (dis)agrees with the given statement, but also the intensity of (dis)agreement. The coders are instructed to position the party to be in complete (dis)agreement with the statement, in case they clearly emphasize their position and do not place any restrictions or doubts (e.g. the party clearly states in the manifesto that it supports/is against euthanasia being legal), whereas the placement “tend to (dis)agree“ is used when the party does not clearly emphasize their position, has doubts, or if it only focuses on one part or element of the statement. The neutral position is assigned when the party only addresses the issue indirectly/vaguely, or if a party makes a point about not taking sides, emphasizing both sides of the issue. It is important to highlight that neutral position is not the same as the “no opinion“ placement. The latter is only used when there is no information whatsoever available about the party's stance on a given statement, whereas the former is an argued position and requires a sourced justification.

Party positioning process

Implementing this rather demanding coding methodology simultaneously on hundreds of parties in different political and linguistic contexts, constitutes a very large-scale and highly complex data collection effort. The backbone of this process are the country expert teams (one per each country involved in the project). In each wave of data collection, more than 100 highly trained social scientists partake the project as country experts, a majority of them affiliated with the European University Institute (EUI) – the organization that has led the project since its initial launch in 2009. In particular, these experts are academics and academically qualified experts in the social sciences, with a specific knowledge of political parties and party systems. Differently from expert surveys, where an extensive number of scholars receive a link to fill out an questionnaire and participate to the process with this single contribution, these experts were recruited and offered a remunerated contract for the entire duration of the project, working together in country teams and liaising constantly with the scientific leadership. Country teams usually were composed of four experts, with some particularly complex and multi-party system countries requiring more experts and vice-versa (e.g. Germany's country team in 2019 was composed of six experts while Malta's two-party system in 2009 only required two experts). The total number of experts involved was 114 in 2009 (average of 4.1 experts per country), 121 in 2014 (average 4.3), and 133 in 2019 (average 4.7). Since the bulk of the recruitment of country expert comes from the pool of researchers affiliated to the EUI at the time of the European elections over the course of 10 years, only a relatively small number of them has participated to more than one wave. These country teams are responsible for selecting the parties that are included in the VAA from their country, translating the materials relating to the project into their native language and – most importantly – positioning the parties on the given list of political statements, in line with the previously described methodological framework.

Few months before each European Parliament election, the country teams contacted the parties that were selected into the VAA, sending them the official invitation letter and the list of statements (both translated into the native language of each country). In an ideal case scenario, the parties reacted and sent their documented self-placements on each statement. Then, the country teams compared their placements to the party self-placements and, in case of discrepancies, asked the party to provide more support for its declared position. After this calibration phase, the experts and the party reached a consensus and final positions on the statements were confirmed. Obviously, in reality, the process was not always that smooth. Some parties started co-operating only in the calibration phase, when the country team had sent them the expert placement. Many, however, did not react to the efforts to contact them at all, and the placement was determined solely by the expert team [9]. In the first edition of the VAA in 2009, the party co-operation rate remained under 40%, whereas in 2014 and 2019 it was consistently above 50%. The dataset, as explained in the first paragraph, includes a dummy variable, signifying whether the party took place in the self-placement procedure (or at least in the calibration phase) or not. Table 4 lists the party co-operation rates country-by-country across all three waves of data collection.

Table 4.

Party co-operation rates across countries and data collection waves.

Country	2009	2014	2019
	% of co-operation (total N of parties)	% of co-operation (total N of parties)	% of co-operation (total N of parties)
Austria	66.7% (6)	100.0% (6)	100.0% (6)
Belgium	76.9% (13)	91.7% (12)	57.1% (14)
Bulgaria	37.5% (8)	25.0% (8)	0.0% (9)
Croatia	14.3% (7)	57.1% (7)	50.0% (12)
Cyprus	100.0% (6)	62.5% (8)	100.0% (9)
Czech Republic	22.2% (9)	50.0% (10)	87.5% (8)
Denmark	66.7% (9)	50.0% (8)	90.0% (10)
Estonia	50.0% (8)	85.7% (7)	50.0% (8)
Finland	83.3% (12)	70.0% (10)	83.3% (12)
France	12.5% (16)	30.0% (10)	8.3% (12)
Germany	50.0% (10)	61.5% (13)	100.0% (15)
Greece	42.9% (7)	33.3% (12)	8.3% (12)
Hungary	66.7% (6)	83.3% (6)	14.3% (7)
Ireland	14.3% (7)	66.7% (6)	50.0% (10)
Italy	12.5% (8)	63.6% (11)	14.3% (7)
Latvia	0.0% (9)	14.3% (7)	90.0% (10)
Lithuania	0.0% (9)	57.1% (7)	14.3% (7)
Luxembourg	37.5% (8)	87.5% (8)	100.0% (10)
Malta	50.0% (4)	33.3% (3)	0.0% (3)
Netherlands	81.8% (11)	91.7% (12)	83.3% (12)
Poland	22.2% (9)	37.5% (8)	16.7% (6)
Portugal	8.3% (12)	12.5% (8)	25.0% (12)
Romania	0.0% (5)	0.0% (9)	14.3% (7)
Slovakia	0.0% (6)	30.0% (10)	30.0% (10)
Slovenia	44.4% (9)	66.7% (9)	73.3% (15)
Spain	63.6% (11)	75.0% (4)	25.0% (8)
Sweden	72.7% (11)	90.0% (10)	88.9% (9)
United Kingdom	10.5% (19)	23.1% (13)	N/A (14)1
Total	39.5%	55.0%	54.3%

¹Due to ongoing Brexit negotiations, it was uncertain whether the British parties would participate in the 2019 EP elections and the euandi2019 British team could only start coding late in the process. Thus, they were not able to contact parties for self-placement. British parties are therefore excluded from calculation of party response rates [10].

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have influenced in any way the work reported in this article.