search | contact us

CSES Module 3 Data Set Errata
Posted: December 15, 2015

Updates made to the Full Release of December 15, 2015, since the Full Release of March 27, 2013

The files released on December 15, 2015 are an update to the prior release of March 27. 2013. Please do not use files from the previous release with files from the current release. Variable names and attributes, as well as the data, have sometimes been changed. However, no records were removed nor added. The respondent, country, and election study count remain the same.

With this release, the data and syntax file formats, and the codebook, have been improved to be more consistent with other CSES data releases. Furthermore, newly accompanying the release are datasets already prepared and ready to be immediately loaded into R, SAS, SPSS, and STATA. The dataset has also been made available in CSV format (albeit without metadata beyond variable names in the column headers).

Last, a number of corrections were made to known issues in the data and documentation. The CSES user community originally made us aware of many of these issues, and we thank you for that feedback. A list of some of the most important changes made is as follows, though this is not a comprehensive list of all changes:

  1. For Poland (2007), the coding of variable C3032_LH_PL (PREVIOUS ELECTION: VOTE CHOICE LOWER HOUSE - PARTY LIST) was corrected to now correspond to the Party table for Poland (2007).

  2. For the Ireland (2007) election study, the value labels of the four variables concerning 'Most important problem' (C3001_1-C3002_2) were added to the codebook. Furthermore, the missing codes of these variables have been adjusted to the CSES standard. An election study note specifying the value labels has been included in the codebook.

  3. For Czech Republic (2010), an error in the number of portfolios after election assigned to parties A, B and C (variables C5013_A, C5013_B and C5013_C - NUMBER OF PORTFOLIOS AFTER) has been corrected.

  4. For Poland (2005), the value labels of the two variables C3002_1 and C3002_2 (MOST IMPORTANT PROBLEM-SOCIOTROPIC) have been added to the codebook.

  5. Variable C5037 (MULTI-PARTY ENDORSEMENTS ON BALLOT) was recoded from '3. YES, CANDIDATE'S NAME APPEARS ONCE, TOGETHER WITH THE NAMES OF ALL SUPPORTING PARTIES' to '1. NO' for the Polish election studies of both 2005 and 2007.

  6. For Ireland (2007), variable C5065 (ELECTORAL FORMULA - LOWEST SEGMENT (TIER) - LOWER HOUSE has been recoded to '32. PR - LARGEST REMAINDER - DROOP' instead of '33. PR - LARGEST REMAINDER - HARE'.

  7. For Japan (2007), variable C5067 (AVERAGE DISTRICT MAGNITUDE - SECOND TIER - LOWER HOUSE) has been recoded to 16.36.

  8. The variable C5054 (REGIME: TYPE OF EXECUTIVE) was recoded from '2. MIXED DEMOCRACY' to '1. PARLIAMENTARY DEMOCRACY' for the Icelandic election studies of 2007 and 2009.

  9. For Taiwan (2008), variable C5041_1 (PARTY LISTS - LOWER - 1ST SEGMENT (TIER) was recoded from code 1. CLOSED to 7. NA, as this concerns a FPTP System. Furthermore, variable C5047_2 (UNIT FOR THE THRESHOLD - LOWER - 2ND SEGMENT (TIER) is not missing anymore but now contains a value for this election study.

  10. The following note was added to the variable notes of C5093-C5096 (concerning the effective number of electoral parties) to mention that data can refer to upper house elections as well, in order to accommodate the Japanese election study: The electoral data employed to calculate this index comes from lower house elections, unless the study is focused on upper house election exclusively.

  11. The following party codes have been added to the party table for Slovenia (2008): 09. United List of Social Democrats (ZLSD), 10. Slovenian People's Party (SLS), 11. Youth Party of Slovenia (SMS).

  12. All Irish cases were coded '000. NOT APPLICABLE' for variable C4003 (NUMBER OF PARTY LISTS IN DISTRICT), as Ireland does not employ party lists.

  13. The alphabetical country codes of Romania and Hong-Kong in Module 3 were corrected. They now read ROU_2009 and HKG_2008.

  14. For Sweden (2006), variable C5055 (NUMBER OF MONTHS SINCE LAST LOWER HOUSE ELECTION) was recoded 48, as Sweden switched to having elections every four years from 1998 onwards.

  15. For both the German elections of 2005 and 2009, variables C4004_A-C4004_G (PERCENT VOTE IN DISTRICT - PARTY X) had been coded according to election outcomes of the 'second vote'. This has been corrected to refer to the 'first vote' instead, as these votes are actually counted on the district level. As a consequence, also the variable C4005 (TURNOUT IN DISTRICT) has been recoded for both Germany (2005) and Germany (2009) as the turnouts for the 'first vote' and 'second vote' differ slightly. In accordance with these changes, C4003 was coded '000. NOT APPLICABLE' for Germany (2009) and data was added for C4002 (NUMBER OF CANDIDATES IN DISTRICT) for Germany (2005).

  16. The Latvian collaborator asked only one Party Identification question which was decided to be coded into variable C3020_1 (ARE YOU CLOSE TO ANY POLITICAL PARTY) instead of C3020_2 (DO YOU FEEL CLOSER TO ONE PARTY) where it was coded before. C3020_2 was set to missing.

  17. We have inserted values for variable C4002 for the election studies Brazil (2006) and Brazil (2010). As the Brazilian electoral system employs open lists, both variable C4003 (NUMBER OF PARTY LISTS IN DISTRICT) and C4002 (NUMBER OF CANDIDATES IN DISTRICT) apply and should contain data.

  18. Variables C5002_A to C5002_D (PERCENT SEATS - LOWER HOUSE) were corrected for Australia (2007).

  19. The election study note for Austria (2008) for variables C3002_1-C3002_2 (MOST IMPORTANT PROBLEM-SOCIOTROPIC) now also mentions the code '010. Unemployment'.

  20. The variable C2028 (RACE) has been recoded for the Mexico (2006) election study to now be consistent with Mexico (2009). An election study note has been added.

  21. Due to a processing error, the macro variables were wrongly coded for Estonia (2011). They were recoded to now contain the correct values, corresponding to the Estonian macro report.

  22. Peru (2011), Chile (2009), and Turkey (2011) had invalid missing codes (999) for macro variables C5016 (IDEOLOGICAL FAMILY) and C5017 (LEFT-RIGHT - PARTY X). These codes were recoded to the valid missing code (99) for these variables.

  23. The coding for the variable C5074 (FUSED VOTE) for the Slovenia (2008) study has been changed, from '5. NO' into '7. NOT APPLICABLE'.

  24. For variable C5065 (ELECTORAL FORMULA - LOWEST SEGMENT (TIER)), both Brazilian election studies (2006 and 2010) used a non-labelled, invalid code 14. This has been corrected and is now coded '31. PR - D'HONDT'.

If you have questions or concerns, e-mail us at and we would be pleased to help.