The CSES Dataset

The Comparative Study of Electoral Systems (CSES) project provides a comparative dataset of election studies and is created by an international team of researchers from around the world. The national research teams include a common set of questions in the national post-elections surveys, such as the British Election Study. This data is combined with electoral variables, demographics, district and national level variables and information on the respective political system. One key advantage of the dataset is that it combines the election studies of a wide range of countries into one single data file and ensures that a considerable number of important questions are asked across these countries in a consistent manner.

A core set of questions remains the same in every survey, but the CSES also incorporates new research topics in its questionnaire, with every of the five-year waves of data collection focusing on a different topic. For instance, the fifth iteration of the project, runs from 2016 to 2021 and is called “Democracy Divided? People, Politicians and the Politics of Populism.”. The subsequent module has specific questions on the topic of “Representative Democracy under Pressure”. Each module comprises a different set of countries, driven, among other aspects by the cooperation of a local team to ensure that a post-election survey is run and the CSES questions are included. Some countries, such as the USA, Germany and Sweden are included in all five modules. The dataset can be accessed here. We present the dataset below. While reading, please keep in mind the questions you see below and answer them once you reached the end. In the end, we will provide a link to a platform with an interactive version of the dataset and additional tasks.

Table 1: General Tasks for the Dataset

Tasks
In which countries are voters more likely to report that they voted in the election?
In which countries do voters overreport that they went to vote compared to official statistics?

Dataset Description

Micro-Level Data

The CSES dataset provides data for respondents on three different levels: micro-level (demographic) data, micro-level (survey) data, district level data and macro-level data. Table 2 shows ten randomly drawn respodents from the dataset. Besides the variables shown here, CSES provides data on occupation, race, income, immigration and region/electoral district.

Table 2: Example from the CSES dataset

	Year of Birth	Gender	Education	Religion	Country
106152	1964	2	7	1208	United States of America
58461	1946	1	5	1208	Iceland
97293	1996	1	4	3000	Tunisia
94885	1940	9	4	8300	Sweden
6625	1994	2	4	1101	Brazil
65727	1943	2	3	1101	Lithuania
28260	1987	1	7	8300	Denmark
96590	1954	2	96	3000	Tunisia
40076	1954	1	3	9999	Hungary
64259	1994	1	4	8300	Republic of Korea

Table 3 shows an example of the voting data in the survey section of the dataset for the 2021 election in Germany. For each of the major parties, the survey asks for their left-right self-placement, their perceived placement of the parties on the left-right scale, and how much they like or dislike the parties. Then, it asks for voting decisions in the most recent parliamentary and presidential election, both on the national and the district level.

Table 3: Example of voting questions from the German dataset

Vote Choice - Party	Left-Right Self-placement	Left-Right CDU	Left-Right SPD	Left-Right Linke	Left-Right Gruene	Left-Right CSU	Left-Right FDP	Left-Right AfD	Like-Dislike CDU	Like-Dislike SPD	Like-Dislike Linke	Like-Dislike Gruene	Like-Dislike CSU	Like-Dislike FDP	Like-Dislike AfD
SPD	5	5	5	10	5	0	5	5	8	8	0	7	3	7	7
Gruene	3	3	7	10	8	2	5	3	6	2	0	1	8	4	9
Gruene	2	2	8	10	9	0	8	3	8	2	0	1	4	1	8
CDU/CSU	0	1	0	10	0	3	2	1	7	8	0	8	0	0	6
CDU/CSU	8	4	6	9	7	1	5	2	4	7	2	8	0	3	1
Gruene	1	1	10	8	10	0	7	8	10	4	2	4	3	4	8
Gruene	3	3	5	9	6	1	5	3	6	5	0	4	1	3	9
CDU/CSU	5	5	5	10	8	0	8	5	3	7	0	3	0	2	9
CDU/CSU	6	2	5	10	7	0	6	5	4	9	0	6	0	5	8
Gruene	3	4	6	10	7	2	6	3	7	5	0	3	7	3	9

In addition to the qustions on voting, the CSES includes a battery of questions in all national post-election surveys that follow a certain theme. In wave 5, this theme is centered around populism. Consequently, Table 4 shows some of the questions on populism from the survey: The respondents were asked to strongly agree, somewhat agree, neither agree not disagree, somewhat disagree or strongly disagree with statements on political elites. In the table, higher values indicate more disagreement with the respective statement.

Table 4: Example of substantive questions from the German dataset

Vote Choice - Party	Elites: Compromise Selling Out	Elites: Do Not Care	Elites: Trustworthy	Elites: Main Problem	Elites: Strong Leader	Elites: People Make Decisions	Elites: Rich and Powerful
SPD	2	3	3	3	4	2	3
Gruene	4	3	2	4	5	4	3
Gruene	4	4	2	4	5	3	3
CDU/CSU	3	3	3	4	1	3	4
CDU/CSU	4	4	2	4	4	5	5
Gruene	1	2	4	2	5	1	2
Gruene	5	2	2	5	5	5	5
CDU/CSU	5	3	4	4	5	5	3
CDU/CSU	4	4	2	5	5	5	5
Gruene	3	2	3	2	4	3	2

District-Level Data

On the district level, the CSES dataset matches district-level results to the place of residence of the respondent. Table 5 shows an example from the 2021 Bundestag election in Germany. In mixed-systems such as Germany, the CSES always reports the district level vote share in this section, which corresponds to the vote for the district candidate in Germany.

Table 5: Example of district level data from the German dataset

	Electoral District	District Vote Share CDU	E5001_A	District Vote Share SPD	District Vote Share AfD	District Vote Share FDP	District Vote Share Linke	District Vote Share Gruene	District Vote Share CSU
2387	63	28.02	25.74	16.29	6.31	7.10	21.49	0.00	11.17
763	89	25.92	25.74	39.73	10.69	7.44	7.63	0.00	2.51
2011	249	21.49	25.74	0.00	10.09	6.07	8.02	38.58	2.60
213	49	38.62	25.74	27.19	10.56	6.56	9.85	0.00	3.90
1187	170	39.35	25.74	24.02	8.88	8.79	10.33	0.00	2.70
1666	220	19.86	25.74	0.00	26.95	9.71	3.85	27.02	3.53
117	20	29.60	25.74	17.16	29.83	8.10	3.41	0.00	7.09
116	19	28.59	25.74	16.75	29.67	8.15	3.42	0.00	9.69
1122	172	33.05	25.74	30.08	9.11	7.98	10.47	0.00	3.15
1084	177	29.66	25.74	28.34	13.87	10.95	8.27	0.00	2.85

Macro-Level Data

On a macro-level, the dataset provides information on electoral results on the national level (votes and seats) for lower and upper houses as well as presidential elections. In the dataset, some countries had parliamentary elections included in the dataset, some presidential, and some both. Table 6 shows 10 individual respondents from the main dataset and the elecotral results that were matched based on the coutnry they live in.

Table 6: Example of macro level electoral data from the CSES dataset

	Country	Election Year	Vote Party A	Vote Party B	Seats Party A	Seats Party B	Vote President A	Vote President B	Turnout
60511	Israel	2020	29.46	26.59	30.00	27.50	997.00	997.00	71.52
60827	Italy	2018	32.68	18.76	36.03	18.89	997.00	997.00	72.94
79597	New Zealand	2017	44.45	36.89	46.67	38.33	996.00	996.00	78.92
10293	Canada	2019	34.34	33.12	35.80	46.45	996.00	996.00	67.04
4719	Belgium	2019	25.46	18.65	28.74	20.69	996.00	996.00	89.72
86080	Poland	2019	35.30	19.43	40.65	22.17	997.00	997.00	61.74
6888	Brazil	2018	11.64	10.29	10.14	10.53	46.03	29.28	79.80
7653	Brazil	2018	11.64	10.29	10.14	10.53	46.03	29.28	79.80
11303	Canada	2019	34.34	33.12	35.80	46.45	996.00	996.00	67.04
155	Albania	2017	48.34	28.85	52.86	30.71	997.00	997.00	46.75

Table 7 shows some of the data that is available on the party level: Parties are coded by party experts into party families, on the left-right scale and in relation to populism. Additionally, the dataset includes information on whether parties were in parliament, in government and what share of government offices they held.

Table 7: Example of macro level party data from the CSES dataset

	Country	Election Year	Portfolios Before Election Party A	Portfolios Before Election Party B	Portfolios After Election Party A	Portfolios After Election Party B	Ideological Family Party A	Ideological Family Party B
104211	United States of America	2016	999	999	999	999	6	9
14690	Switzerland	2019	2	2	2	2	10	4
82360	New Zealand	2020	16	0	20	0	4	9
51043	India	2019	37	0	29	0	9	6
8648	Brazil	2018	5	0	2	0	9	4
94777	Sweden	2018	17	0	18	0	4	9
40158	Hungary	2018	14	0	14	0	10	10
88939	El Salvador	2019	999	999	999	999	9	7
111758	United States of America	2020	999	999	999	999	6	9
84791	Poland	2019	15	0	14	0	9	7

Example: Predictors of Turnout

One major example of the CSES is that the dataset contains a multitude of variables describing the electoral system, such as whether there is compulsory voting, whether there are party thresholds or which voting system (e.g. party lists, single member districts,…) is used. All data is available in one dataset so researchers can combine individual-level survey data and macro-level information in a coherent manner. As an example, we can look at self-reported turnout in national elections (Figure 1) and plot it in relation to actual turnout from official figures (Figure 2).

##               
##                Albania Australia Austria Belgium Brazil Canada Chile Costa Rica
##   Did not vote     223        69      75     142    479    183   690        316
##   Voted           1031      1765    1127    1595   2014   2700  1294       1128
##               
##                Czechia Denmark El Salvador Finland France Germany Great Britain
##   Did not vote     971      38         397     181    274     343           564
##   Voted           2044    1363        1089    1285   1555    4818          2922
##               
##                Greece Hong Kong Hungary Iceland India Ireland Israel Italy
##   Did not vote    262       128     282     175  1136      95    129   176
##   Voted          2122       888     894    2952 12229     898   1077  1813
##               
##                Japan Latvia Lithuania Mexico Montenegro Netherlands New Zealand
##   Did not vote   440    242       839    233        181         338         162
##   Voted         1215    766      2384    936       1003        4977        3315
##               
##                Norway  Peru Poland Portugal Republic of Korea Romania Slovakia
##   Did not vote    108   160    530      497               479     328      225
##   Voted          1681  1024   1427      992               720     742      763
##               
##                Sweden Switzerland Taiwan Thailand Tunisia Turkey
##   Did not vote     67        1279    432       21     909     71
##   Voted          3587        3328   2933     1453     497    998
##               
##                United States of America Uruguay
##   Did not vote                     1222      23
##   Voted                            9130    1088