CSES Integrated Module Dataset (IMD) Errata

Updates made to Phase 4 Release of February 27, 2024 since Phase 3 Release of December 8, 2020

Phase 4 of the CSES Integrated Module Dataset (IMD) of February 27, 2024, contains additions, updates, and improvements on the Phase 3 release of December 8, 2020. We advise users not to use files from previous releases with files from the current release. Variable names and attributes, as well as the data, have sometimes been changed.

Overall, CSES IMD Phase 4 sees 114,714 respondents from CSES Module 5 and 17 new variables being added to the dataset, such that the dataset now encompasses 395,797 respondents and 400+ variables overall. A total of 56 election studies from Module 5 have thus been added to IMD Phase 4, bringing the number of election studies to 230 and the number of polities included to 59. All new studies and variables added to IMD Phase 4 from IMD Phase 3 are listed below in the data changes section. Naturally, these newly added 114,714 observations come with appropriate documentation, but users are advised that not all resulting Codebook changes arising from these additions are noted in the Codebook changes list below.

In addition to the set of new variables and election studies, we provide users with a list of other changes and corrections made to CSES IMD Phase 4 compared to CSES IMD Phase 3. We have divided this into data changes, codebook changes, and explanations of deviations between the CSES IMD Phase 4 and the Standalone CSES Modules. Users are advised that all these errata lists, which are not comprehensive but instead highlight the main changes incorporated in CSES IMD since the previous release of 2020.

CSES IMD Phase 4 Errata Lists [PDF, 23 pages]

Updates made to Phase 3 Release of December 8, 2020 since Phase 2 Release of October 17, 2019

Phase 3 of the CSES Integrated Module Dataset (IMD) of December 8, 2020, contains additions, updates, and improvements on the Phase 2 release of October 17, 2019. We advise users not to use files from previous releases with files from the current release. Variable names and attributes, as well as the data, have sometimes been changed.

Overall, CSES IMD sees a set of 60+ variables being added to the dataset. The new variables added to IMD Phase 3 from IMD Phase 2 are listed below in the data changes section.

In addition to the extensive set of new variables, we provide users with a list of other changes and corrections made to CSES IMD Phase 3 compared to CSES IMD Phase 2. We have divided this into data changes, codebook changes, and explanations of deviations between the CSES IMD Phase 3 and the Standalone CSES Modules. Users are advised that this list is not comprehensive but instead highlights the main changes incorporated since CSES IMD Phase 2, which was released on October 17, 2019.

DATA CHANGES

The following administrative variables have been added to the dataset:

IMD1006_UNAlpha2 ID COMPONENT – POLITY UN ISO_3166 ALPHABETIC TWO LETTER CODE: Variable classifying a polity with two alphabetical code (e.g., DE for Germany). Can facilitate data bridging with other common datasets in the political and social sciences.
IMD1006_REG ID COMPONENT – POLITY UN GEOGRAPHIC REGIONS NUMERIC CODES: Variable classifying geographical regions based on the United Nations Statistics Division classification. Can facilitate data bridging with other common datasets in the political and social sciences.
IMD1006_OECD ID COMPONENT – POLITY MEMBER OF OECD: Dichotomous variable classifying if a polity is or is not a member of the Organization for the Economic Cooperation and Development at the time of the election.
IMD1006_EU ID COMPONENT – POLITY MEMBER OF EUROPEAN UNION: Dichotomous variable classifying if a polity is or is not a member of the European Union at the time of the election.
IMD1006_VDem ID COMPONENT – V-Dem POLITY IDENTIFIER: Polity numerical identifier applied to states included in the Varieties of Democracy (V-Dem) project to facilitate data bridging with V-Dem data.
IMD1011_1 DATE 1ST ROUND ELECTION BEGAN – YYYY-MM-DD: Additional date variable complementing the original date variables in CSES IMD (IMD1011_M, IMD1011_D, and IMD1011_Y) which brings together year, month, and day of the first round election in the same variable to facilitate data bridging with other common datasets in the political and social sciences.
IMD1011_2 DATE 1ST ROUND ELECTION BEGAN – YYYYMM: Additional date variable complementing the original date variables in CSES IMD (IMD1011_M, IMD1011_D, and IMD1011_Y) which brings together year and month of the first round election in the same variable to facilitate data bridging with other common datasets in the political and social sciences.
IMD1012_1 DATE 2ND ROUND ELECTION BEGAN – YYYY-MM-DD: Additional date variable complementing the original date variables in CSES IMD (IMD1012_M, IMD1012_D, and IMD1012_Y) which brings together year, month, and day of the second round election (when applicable) in the same variable to facilitate data bridging with other common datasets in the political and social sciences.
IMD1012_2 DATE 2ND ROUND ELECTION BEGAN – YYYYMM: Additional date variable complementing the original date variables in CSES IMD (IMD1012_M, IMD1012_D, and IMD1012_Y) which brings together year and month of the second round election (when applicable) in the same variable to facilitate data bridging with other common datasets in the political and social sciences.

The following demographic variables have been added to the dataset:

IMD2001_G_ BIRTH GENERATION: Six dichotomous variables classifying respondents in CSES IMD into generational cohorts based on their year of birth.
IMD2005_1 RELIGIOUS SERVICES ATTENDANCE
IMD2005_2 RELIGIOSITY
IMD2008 REGION OF RESIDENCE: Data were harmonized across four Standalone Modules on the polity level, based on a scale newly created for CSES IMD.
IMD2010 RACE: Data were harmonized based on a scale newly created for CSES IMD.
IMD2011 ETHNICITY: Data were harmonized based on a scale newly created for CSES IMD, namely the European Standard Classification of Cultural and Ethnic Group (ESCEG), developed for the European Social Survey.
IMD2012_1 NUMBER IN HOUSEHOLD IN TOTAL
IMD2012_1 NUMBER IN HOUSEHOLD UNDER AGE 18
IMD2013 LANGUAGE USUALLY SPOKEN AT HOME
IMD2014 CURRENT EMPLOYMENT STATUS
IMD2015_ISCO_88 MAIN OCCUPATION (CSES MODULES 1-3): This variable is available for election studies included in CSES Modules 1-3.
IMD2015_ISCO_08 MAIN OCCUPATION (CSES MODULE 4): This variable is available for election studies included in CSES Module 4.
IMD2016 SOCIO-ECONOMIC STATUS: This variable is available for election studies included in CSES Modules 2-4.
IMD2017 EMPLOYMENT TYPE – PUBLIC OR PRIVATE
IMD2018 INDUSTRIAL SECTOR
IMD2019_1 UNION MEMBERSHIP OF RESPONDENT
IMD2019_2 UNION MEMBERSHIP OF OTHERS IN HOUSEHOLD
IMD2020 SPOUSE: CURRENT EMPLOYMENT STATUS: This variable is available for election studies included in CSES Modules 2-4.
IMD2021_ISCO_88 SPOUSE: MAIN OCCUPATION (CSES MODULES 1-3): This variable is available for election studies included in CSES Modules 1-3.
IMD2021_ISCO_08 SPOUSE: MAIN OCCUPATION (CSES MODULE 4): This variable is available for election studies included in CSES Module 4.
IMD2022 SPOUSE: SOCIO ECONOMIC STATUS: This variable is available for election studies included in CSES Modules 2-4.
IMD2023 SPOUSE: EMPLOYMENT TYPE – PUBLIC OR PRIVATE: This variable is available for election studies included in CSES Modules 2-4.
IMD2024 SPOUSE: INDUSTRIAL SECTOR: This variable is available for election studies included in CSES Modules 2-4.
IMD2025 BUSINESS OR EMPLOYERS’ ASSOCIATION MEMBERSHIP: This variable is available for election studies included in CSES Modules 2-4.
IMD2026 FARMERS’ ASSOCIATION MEMBERSHIP: This variable is available for election studies included in CSES Modules 2-4.
IMD2027 PROFESSIONAL ASSOCIATION MEMBERSHIP: This variable is available for election studies included in CSES Modules 2-4.

The following micro-level variables have been added to the dataset:

IMD3001_TS TURNOUT: TURNOUT SWITCHER BETWEEN CURRENT ELECTION AND PREVIOUS ELECTION: A derivative variable created by the CSES Secretariat based on respondent’s reported turnout in the current and previous elections classifying if a respondent voted in both elections, voted in only one, or abstained in both.
IMD3002_VS_1 VOTE CHOICE: VOTE SWITCHER BETWEEN CURRENT ELECTION AND PREVIOUS ELECTION: A derivative variable created by the CSES Secretariat based on respondent’s reported vote choice in current and previous main election, classifying if a respondent voted in both elections for the same party/coalition or if they switched and voted for different party/coalition across the two contests.
IMD3002_LR_CSES VOTE CHOICE: CURRENT MAIN ELECTION – VOTE FOR LEFTIST/CENTER/RIGHTIST – CSES: A derivative variable created by the CSES Secretariat based on respondent’s reported vote choice in the main election and CSES collaborators’ expert judgment of party A-I left-right position, classifying if a respondent voted for leftist, center or rightist party/coalition.
IMD3002_LR_MARPOR VOTE CHOICE: CURRENT MAIN ELECTION – VOTE FOR LEFT/RIGHT (RILE) – MARPOR/CMP: A derivative variable created by the CSES Secretariat based on respondent’s reported vote choice in the main election and MARPOR/CMP values on “RILE” scale for party A-I, classifying if a respondent voted for leftist or rightist party/coalition.
IMD3002_IF_CSES VOTE CHOICE: CURRENT MAIN ELECTION – VOTE CHOICE BY IDEOLOGICAL FAMILY CLASSIFICATION – CSES: A derivative variable created by the CSES Secretariat based on respondent’s reported vote choice in the main election and CSES collaborators’ expert judgment of party A-I ideological family, classifying party family respondent voted for in the main election.
IMD3100_LR_CSES VOTE CHOICE: CURRENT MAIN ELECTION – VOTE CHOICE LINKED WITH CSES COLLABORATOR EXPERT JUDGMENT L-R: A derivative variable created by the CSES Secretariat linking the respondent’s reported vote choice in the main election and CSES collaborators’ expert judgment of party A-I left-right position, assigning L-R value on a scale from 0 (left) to 10 (right)
IMD3100_LR_MARPOR VOTE CHOICE: CURRENT MAIN ELECTION – VOTE CHOICE LINKED WITH MARPOR/CMP RILE: A derivative variable created by the CSES Secretariat linking the respondent’s reported vote choice in the main election and MARPOR/CMP values on “RILE” scale for party A-I, assigning RILE value on a scale from -100 (left) to +100 (right).
IMD3100_IF_CSES VOTE CHOICE: CURRENT MAIN ELECTION – VOTE CHOICE LINKED WITH CSES COLLABORATOR EXPERT JUDGMENT IDEOLOGICAL FAMILY: A derivative variable created by the CSES Secretariat linking the respondent’s reported vote choice in the main election and CSES collaborators’ expert judgment of party A-I ideological family, assigning original ideological family classification.

The following macro-level variables have been added to the dataset:

IMD5032_1 ELECTION VIOLENCE: This variable is available for election studies included in CSES Modules 3-4.
IMD5032_2 GEOGRAPHIC CONCENTRATION OF VIOLENCE: This variable is available for election studies included in CSES Modules 3-4.
IMD5032_3 POST-ELECTION VIOLENCE: This variable is available for election studies included in CSES Modules 3-4.
IMD5032_4 POST-ELECTION PROTEST: This variable is available for election studies included in CSES Modules 3-4.
IMD5033 FAIRNESS OF THE ELECTION: This variable is available for election studies included in CSES Modules 3-4.
IMD5034_1 FORMAL COMPLAINTS AGAINST NATIONAL LEVEL RESULTS: This variable is available for election studies included in CSES Modules 3-4.
IMD5034_2 ELECTION IRREGULARITIES REPORTED: This variable is available for election studies included in CSES Modules 3-4.
IMD5034_3 ELECTION DATE IRREGULARITIES: This variable is available for election studies included in CSES Modules 3-4.
IMD5035 NUMBER OF PARTIES PARTICIPATING IN ELECTION: This variable is available for election studies included in CSES Modules 3-4.
IMD5036_1 ELECTORAL ALLIANCES PERMITTED IN AN ELECTION
IMD5036_2 ELECTORAL ALLIANCES IN PRACTICE: This variable is available for election studies included in CSES Modules 3-4.
IMD5036_3 DID ANY ELECTORAL ALLIANCE FORM?
IMD5038 REQUIREMENTS FOR JOINT PARTY LISTS
IMD5040_1-2 MULTI-PARTY ENDORSEMENTS
IMD5041_1 NUMBER OF ELECTORAL SEGMENTS (TIERS): For CSES Modules 1 and 2, values were separately provided for lower house and upper house. In IMD, the variable is harmonized based on lower house unless otherwise stated in a Polity or Election Study Note in Part 2 of the IMD Codebook.
IMD5041_2 LINKED ELECTORAL SEGMENTS (TIERS): This variable is available for election studies included in CSES Modules 3-4.
IMD5042_1 DEPENDENT FORMULA IN MIXED SYSTEMS: This variable is available for election studies included in CSES Modules 3-4.
IMD5042_2 SUBTYPES OF MIXED ELECTORAL SYSTEMS: This variable is available for election studies included in CSES Modules 3-4.
IMD5044 FUSED VOTE: This variable is available for election studies included in CSES Modules 3-4.
IMD5045_1-3 AVERAGE DISTRICT MAGNITUDE: across multiple tiers for the lower house elections.
IMD5046_1-3 TRANSFERABLE VOTES: across multiple tiers for the lower house elections.
IMD5048 REGIME: TYPE OF EXECUTIVE
IMD5061 CONSTITUTIONAL FEDERAL STRUCTURE

The following data bridging variables facilitating bridging party/coalition data have been added to the dataset:

IMD5100_A-I MANIFESTO RESEARCH ON POLITICAL REPRESENTATION (MARPOR/CMP) IDENTIFIER – PARTY A-I: Party identification codes from the Manifesto Research on Political Representation (MARPOR/CMP) project, provided for parties that are assigned an alphabetical code (A-I) to facilitate data bridging between CSES and MARPOR/CMP.
IMD5101_A-I PARLIAMENTS AND GOVERNMENT DATABASE (PARLGOV) IDENTIFIER – PARTY A-I: Party identification codes from the Parliaments and Governments Database (ParlGov) project, provided for parties that are assigned an alphabetical code (A-I) to facilitate data bridging between CSES and ParlGov.
IMD5102_A-I CHAPEL HILL EXPERT SURVEY (CHES) IDENTIFIER -PARTY A-I: Party identification codes from the Chapel Hill Expert Survey (CHES) project, provided for parties that are assigned an alphabetical code (A-I) to facilitate data bridging between CSES and CHES.
IMD5103_A-I PARTY FACTS IDENTIFIER – PARTY A-I: Party identification codes from the Party Facts project, provided for parties that are assigned an alphabetical code (A-I) to facilitate data bridging between CSES and other data sources linked with Party Facts.

IMD3005_1 (PARTY IDENTIFICATION: ARE YOU CLOSE TO ANY POLITICAL PARTY): Category “No” has been recoded to value 0. In previous versions of CSES IMD, the value for category “No” was 5.

IMD3005_2 (PARTY IDENTIFICATION: DO YOU FEEL CLOSER TO ONE PARTY): Category “No” has been recoded to value 0. In previous versions of CSES IMD, the value for “No” was 5.

IMD5021_ (IS THERE A PARTY THRESHOLD): Category “No” has been recoded to value 0. In previous versions of CSES IMD, the value for “No” was 5 for all four IMD5021_ variables.

IMD3004_PR_1 (VOTE CHOICE: PREVIOUS PRESIDENTIAL ELECTION – ROUND 1): For Chile (2005) this variable is set to “999999. Missing.” The variable concerning the second round of the previous presidential election. Thus, data is coded into IMD3004_PR_2 (VOTE CHOICE: PREVIOUS PRESIDENTIAL ELECTION – ROUND 2), which was coded as “9999999. Missing” in previous versions of CSES IMD.

IMD3003_LH (TURNOUT: PREVIOUS LOWER HOUSE ELECTION): For Australia (2007) and Denmark (2001) data was added for IMD3003_LH. Even though data exists in respective Standalone CSES Modules, in previous versions of CSES IMD it was incorrectly classified as missing. In Phase 3 of CSES IMD, this has been corrected.

IMD3009_E (LIKE-DISLIKE – LEADER E): For Turkey (2015) this variable is set to “99. Missing” due to a coding error. This issue will be addressed in future release of CSES IMD.

CODEBOOK CHANGES

The CSES IMD Codebook now consists of five separate sections (previously four). Part 5 documents party/coalition level data bridging identifiers included in CSES IMD Phase 3 for four different projects, namely:

MANIFESTO RESEARCH ON POLITICAL REPRESENTATION (MARPOR/CMP)
PARLIAMENTS AND GOVERNMENT DATABASE (PARLGOV)
CHAPEL HILL EXPERT SURVEY (CHES)
PARTY FACTS

The instructions regarding navigation of the CSES IMD Codebook have been updated. The update includes information on how users can quickly find Derivative Variables and Bridging Identifiers. Users can find this information in all five parts of the CSES IMD Codebook by searching for “))) HOW TO NAVIGATE THE CSES IMD CODEBOOK.”

Section has been added to the CSES IMD Codebook detailing Data Bridging. Users can find this information in Parts 1, 2, and 5 of the CSES IMD Codebook by searching “DATA BRIDGING: NEW FRONTIERS”

Section has been added to the CSES IMD Codebook detailing Derivative Variables. Users can find this information in Parts 1 and 2 of the CSES IMD Codebook by searching “DERIVATIVE VARIABLES”

The CSES IMD Codebook Bibliography section (Part 1 of CSES IMD Codebook) detailing data sources have been updated.

IMD5013 (ELECTORAL FORMULA IN ALL SEGMENTS: LOWER HOUSE): The Codebook entry has been replaced with a new entry detailing IMD5013. Previous versions of the CSES IMD Codebook had incorrect information.

IMD5014 (ELECTORAL FORMULA: PRESIDENTIAL ELECTION): Category 5 is now labelled “5. Alternative Vote” as it is labeled in the data. Previous versions of the CSES IMD Codebook had listed this as “5. Single Transferable Vote.” – as only one position is contested, the Alternative Vote is the more appropriate classification.

IMD1010_3 (ELECTION STUDY WEIGHT: POLITICAL): For Portugal (2005) an Election Study Note is added explaining the zero values.

IMD2001_1 (AGE OF RESPONDENT (IN YEARS)): Several Election Study Notes added explaining the reasons why some respondents, younger than eligible voting-age, are included in the sample. These notes refer to the following election studies (listed in alphabetical order):

Denmark (2001)
Germany (2002)
Montenegro (2012)
Netherlands (2010)
Norway (2001 and 2005)
USA (2008).

DEVIATIONS BETWEEN CSES IMD AND STANDALONE CSES MODULES

Despite question wordings being virtually identical across modules, CSES IMD harmonizes variables to be consistent over time. Consequently, variables in CSES IMD sometimes deviate in their coding from how they appear in Standalone CSES Modules. These deviations can occur because of coding errors in the Standalone CSES dataset or codebook. In these circumstances, appropriate errata are listed on the Standalone CSES Module page. Deviations may also legitimately arise because of different coding schemes applied in CSES, derivative variables created that are unique to CSES IMD, lack of data at the time of processing of Standalone CSES Modules which has since become available, or that data for certain variables were not collected in original Standalone CSES Modules but have become eligible for inclusion in the CSES IMD as it meets the “3 and 1” eligibility criteria (see CSES IMD Stimulus Paper).

Below, we detail the primary deviations between Standalone CSES Modules and CSES IMD Phase 3. CSES makes every effort to ensure that these deviations are minimal and that where they occur that details concerning them are listed in the CSES IMD Codebook and also here.

The following variables are not available in some or all Standalone CSES Modules (listed in the brackets) but are added to CSES IMD in Phase 3. They are:

IMD1006_UNALPHA2 ID COMPONENT – POLITY UN ISO_3166 ALPHABETIC TWO LETTER CODE: Not available in Standalone CSES Modules 1-4.
IMD1006_REG ID COMPONENT – POLITY UN GEOGRAPHIC REGIONS NUMERIC CODES: Not available in Standalone CSES Modules 1-4.
IMD1006_OECD ID COMPONENT – POLITY MEMBER OF OECD: Not available in Standalone CSES Modules 1-4.
IMD1006_EU ID COMPONENT – POLITY MEMBER OF EUROPEAN UNION: Not available in Standalone CSES Modules 1-4.
IMD1006_VDEM ID COMPONENT – V-DEM POLITY IDENTIFIER: Not available in Standalone CSES Modules 1-4.
IMD1011_1 DATE 1ST ROUND ELECTION BEGAN – YYYY-MM-DD: Not available in Standalone CSES Modules 1-4.
IMD1011_2 DATE 1ST ROUND ELECTION BEGAN – YYYYMM: Not available in Standalone CSES Modules 1-4.
IMD1012_1 DATE 2ND ROUND ELECTION BEGAN – YYYY-MM-DD: Not available in Standalone CSES Modules 1-4.
IMD1012_2 DATE 2ND ROUND ELECTION BEGAN – YYYYMM: Not available in Standalone CSES Modules 1-4.
IMD2001_G_ BIRTH GENERATION: Not available in Standalone CSES Modules 1-4.
IMD5048 REGIME: TYPE OF EXECUTIVE: Not available in Standalone CSES Modules 1-2.
IMD5061 CONSTITUTIONAL FEDERAL STRUCTURE: Not available in Standalone CSES Modules 1-2.
IMD5100_A-I MANIFESTO RESEARCH ON POLITICAL REPRESENTATION (MARPOR/CMP) IDENTIFIER – PARTY A-I: Not available in Standalone CSES Modules 1-3.
IMD5101_A-I PARLIAMENTS AND GOVERNMENT DATABASE (PARLGOV) IDENTIFIER – PARTY A-I: Not available in Standalone CSES Modules 1-3.
IMD5102_A-I CHAPEL HILL EXPERT SURVEY (CHES) IDENTIFIER – PARTY A-I: Not available in Standalone CSES Modules 1-4.
IMD5103_A-I PARTY FACTS IDENTIFIER – PARTY A-I: Not available in Standalone CSES Modules 1-4.

The following derivative variables are added to CSES IMD Phase 3 and are not available Standalone CSES Modules 1-4. These are:

IMD3001_TS TURNOUT: TURNOUT SWITCHER BETWEEN CURRENT ELECTION AND PREVIOUS ELECTION
IMD3002_VS_1 VOTE CHOICE: VOTE SWITCHER BETWEEN CURRENT ELECTION AND PREVIOUS ELECTION
IMD3002_LR_CSES VOTE CHOICE: CURRENT MAIN ELECTION – VOTE FOR LEFTIST/CENTER/RIGHTIST – CSES
IMD3002_LR_MARPOR VOTE CHOICE: CURRENT MAIN ELECTION – VOTE FOR LEFT/RIGHT (RILE) – MARPOR/CMP
IMD3002_IF_CSES VOTE CHOICE: CURRENT MAIN ELECTION – VOTE CHOICE BY IDEOLOGICAL FAMILY CLASSIFICATION – CSES
IMD3100_LR_CSES VOTE CHOICE: CURRENT MAIN ELECTION – VOTE CHOICE LINKED WITH CSES COLLABORATOR EXPERT JUDGMENT L-R
IMD3100_LR_MARPOR VOTE CHOICE: CURRENT MAIN ELECTION – VOTE CHOICE LINKED WITH MARPOR/CMP RILE
IMD3100_IF_CSES VOTE CHOICE: CURRENT MAIN ELECTION – VOTE CHOICE LINKED WITH CSES COLLABORATOR EXPERT JUDGMENT IDEOLOGICAL FAMILY

IMD2005_1 (RELIGIOUS SERVICE ATTENDANCE): For Australia (2007), respondents attending religious services “at least once a month” have been recoded to “4. Once a Month” in the CSES IMD Release of December 8, 2020, to harmonize data with other Australian studies. In the CSES Module 3 Full Release of December 15, 2015, this category was incorrectly classified as “5. Two or More Times a Month.”

IMD2005_1 (RELIGIOUS SERVICE ATTENDANCE): For Switzerland (2007), data have been recoded for the following categories:

“Once a year” recoded to “2. Once a year” from “3. Two to eleven times a year”
“Several times a year” recoded to “3. Two to eleven times a year” from “4. Once a month”
“Once or twice per month” recoded to “4. Once a month” from “5. Two or more times a month”
“Several times a week “ recoded to “6. Once a week / more than once a week” from “7. Volunteered: Refused”

Data remain unchanged in the CSES Module 3 Full Release of December 15, 2015.

IMD2005_2 (RELIGIOSITY): For Hungary (1998), respondents stating “I cannot tell whether I am religious or not” have been recoded to “2. Not very religious” in the CSES IMD Release of December 8, 2020 to harmonize data with the Hungarian (2002) study. In CSES Module 1 Full Release of December 15, 2015, that category was incorrectly coded as “8. Don’t know.”

IMD2010 (RACE): There was no unique coding scheme for a respondent’s race across Standalone Modules. Data were harmonized based on a scale newly created for IMD. Additionally, the following changes were applied:

For Canada (2008) this variable distinguishes respondent’s racial minorities from other respondents in the sample. As this approach is challenging to translate to the newly established coding scheme, IMD2010 is set to “999. Missing” for Canada (2008).
For Hungary (1998) the race variable included in the CSES Module 1 Full Release of December 15, 2015, distinguishes between respondents of Romani and other origin. The IMD provides this data in IMD2011 (Ethnicity), while IMD2010 (Race) has been recoded to “999. Missing” for Hungary (1998).
For New Zealand (1996) the race variable included in the CSES Module 1 Full Release of December 15, 2015 includes data for a respondent’s race and ethnicity. Harmonization revealed that data in both variables refer to the same scale. Therefore, the IMD includes the respective data in IMD2011 (Ethnicity), while IMD2010 (Race) has been recoded to “999. Missing” for New Zealand (1996).

IMD2011 (ETHNICITY): In the CSES Standalone Module datasets, data were coded according to national standards. For the CSES IMD Release of December 8, 2020, these individual scales were harmonized based on the European Standard Classification of Cultural and Ethnic Group (ESCEG), developed for the European Social Survey. Additionally, the following changes were applied:

For Great Britain (1997 and 2005) the ethnicity variable studies included in the CSES Modules 1 and 2 Full Release of December 15, 2015, includes data for a respondent’s race and ethnicity. Harmonization revealed that data in both variables refer to the same scale. Therefore, the IMD includes the respective data in IMD2010 (Race), while IMD2011 (ethnicity) has been recoded to “999999. Missing” for Great Britain (1997 and 2005).
For all Mexico studies included in the CSES IMD the ethnicity variable includes data for a respondent’s race and ethnicity. Harmonization revealed that data in both variables refer to the same scale. However, while the Standalone Modules 1 and 2 coded the respective data into ethnicity, later Modules provide the data as race. The IMD harmonizes coding by including all the available data in IMD2010 (Race). Hence, IMD2011 has been recoded to “999999. Missing” for Mexico (1997, 2000, 2003, 2006, 2009, 2012, 2015).
For Peru (2011 and 2016) the ethnicity variable distinguish between the following four ethnic groups: Indigenous peoples, Afro Peruvians, Whites and Mestizos. For the IMD, these groups were coded into IMD2010 (Race), while IMD2011 has been coded to “999999. Missing” for Peru (2011 and 2016).
For Romania (1996) this variable includes a wild code (i.e., codes where the meaning of the classification is not ascertainable) 5 in the data which is not documented in the CSES Module 1 Codebook. Data have been set to “999999. Missing” in the CSES IMD Release of December 8, 2020 for Romania (1996), as the meaning of the code could not be ascertained.
For Slovenia (2004) this variable, included in CSES Module 2 Full Release of December 15, 2015, the codes assigned for the ethnicity variable do not match the codes listed in the Codebook. Data have been set to “999999. Missing” in the CSES IMD Release of December 8, 2020 for Slovenia (2004), as the meaning of the codes could not be ascertained.

IMD2013 (LANGUAGE USUALLY SPOKEN AT HOME): The Standalone CSES Modules 1-4 included several auxiliary codes 227, 300 – 312 (Module 1) and 980 to 995 (Modules 2-4), for languages originally not envisaged by the original CSES coding scheme. For the CSES IMD, these codes have now been fully harmonized, thereby expanding the existing coding scheme by codes 300 to 340 and 992 to 995.

IMD2013 (LANGUAGE USUALLY SPOKEN AT HOME): The CSES Module 2 Full Release of December 15, 2015, includes several wild codes (i.e., codes where the meaning of the classification is not ascertainable) which are not documented in the CSES Module 2 Codebook. These are: Code 981 for Canada (2004), code 990 for Great Britain (2005) and code 991 for New Zealand (2002). The respective respondents have been recoded to “999. Missing” in the CSES IMD Release of December 8, 2020, as the meanings of the codes could not be ascertained.

IMD2014 (CURRENT EMPLOYMENT STATUS): The Standalone CSES Modules 2-4 included several auxiliary codes 11 and 12 for employment statuses originally not envisaged by the original CSES coding scheme. These codes have been harmonized for IMD, thereby expanding the existing coding scheme by codes “11. On temporary job leave” and “12. Civil/Military Service”.

IMD2014 (CURRENT EMPLOYMENT STATUS): Employed respondents from studies not differentiating employment by working hours were mostly coded as “1. Employed – Full Time” in the Standalone Modules, but have been recoded to “0. Employed – no working hours specified” in the CSES IMD Release of December 8, 2020.

IMD2014 (CURRENT EMPLOYMENT STATUS): The following recodes have been applied to individual election studies for this variable:

For Canada (2008) self-employed respondents and respondents working at two or more jobs were assigned codes 11 and 12 in the CSES Module 3 Full Release of December 15, 2015. These respondents been recoded to “1. Employed – full time” for Canada (2008) in the CSES IMD Release of December 8, 2020 to harmonize data with other Canadian studies.
For Canada (2011 and 2015) respondents being retired and working for pay or a student and working for pay were assigned codes 11 and 12 in CSES Module 4 Full Release of May 29, 2018. These respondents been recoded to “2. Employed – part time” for Canada (2011 and 2015) in the CSES IMD Release of December 8, 2020 to harmonize data with the Canadian 2004 study. For the same reason, working students coded as “3. Employed – less than part time” in Module 4 have been recoded to “2. Employed – part time” for Canada (2011 and 2015).
For France (2007) respondents refusing to answer IMD2014 were incorrectly coded “98. Volunteered: Don’t know” in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “97. Volunteered: Refused” for France (2007) in the CSES IMD Release of December 8, 2020.
For the Netherlands (2010) respondents working less than 12 hours a week were coded as “2. Employed – part time” in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “3. Employed – less than part time” in the CSES IMD Release of December 8, 2020.
For New Zealand (2011) respondents engaging in an unpaid work outside the home were classified as “4. Employed – less than part time” in the CSES Module 4 Full Release of May 29, 2018. These respondents have been recoded to “10. Others, not in labor force” for the CSES IMD Release of December 8, 2020, to harmonize data with the New Zealand (2002) study.
For Slovenia (2008) self-employed respondents were assigned code 11 in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “1. Employed – full time” in the CSES IMD Release of December 8, 2020, to harmonize data with the Slovenian (2004) study.
For Slovenia (2008) retired respondents with additional employment are coded 12 in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “7. Retired” in the CSES IMD Release of December 8, 2020.

IMD2015_ISCO_88 (MAIN OCCUPATION (CSES MODULES 1-3)): This variable reports a respondent’s main occupation based on ISCO-88 two-digit codes for election studies included in CSES Modules 1-3. The following recodes have been applied:

For CSES MODULE 1 Full Release of December 15, 2015, categories “0. Not Applicable” and “97. No Occupation” were recoded to “999. Missing” in the CSES IMD Release of December 8, 2020.
For Canada (1997) the following recodes have been made: Teachers were assigned code 28, a code not envisaged by ISCO-88. For CSES IMD, they have been recoded to “23. Teaching Professionals.” Further, forestry workers coded as “63. Subsistence farmers” in Module 1 were recoded to “61. Market-oriented skilled agricultural workers”. Finally, non-classifiable occupations coded 94 in the CSES Module 1 Full Release of December 15, 2015, were recoded to “996. Other non-classifiable occupations” for Canada (1997) in the CSES IMD Release of December 8, 2020.
For Canada (2004) mid-managers were coded “20. Professionals” in the CSES Module 2 Full Release of December 15, 2015. These respondents were recoded to “10. Legislators, senior officials and managers” for Canada (2004) in the CSES IMD Release of December 8, 2020.
For Canada (2004) special teaching employees are coded as “30. Technicians and associated professionals” in the CSES Module 2 Full Release of December 15, 2015. These respondents were recoded to “23. Teaching Professionals” for Canada (2004) in the CSES IMD Release of December 8, 2020.
For Japan (2004) respondents without occupation were assigned code 90 in the CSES Module 2 Full Release of December 15, 2015. These respondents were recoded to “999. Missing” for Japan (2004) in the CSES IMD Phase 3 Release of December 8, 2020.
For Latvia (2010) skilled workers were coded “60. Skilled agricultural and fishery workers” in the CSES Module 3 Full Release of December 15, 2015. These respondents were recoded to “996. Other non-classifiable occupations” for Latvia (2010) in the CSES IMD Phase 3 Release of December 8, 2020, as skilled workers are a generic group not matching a single ISCO-88 category.
For Norway (1997) wild codes (i.e., codes where the meaning of the classification is not ascertainable) 25, 63 and 64 are included in the data which are neither envisaged by ISCO-88, nor documented in the CSES Module 1 Codebook. These respondents were recoded to “999. Missing” for Norway (1997) in the CSES IMD Release of December 8, 2020, as the meaning of the codes could not be ascertained.
For Sweden (2006) respondents who have never been working were assigned code “97. Refused” in the CSES Module 3 Full Release of December 15, 2015. These respondents were recoded to “999. Missing” for Sweden (2006) in the CSES IMD Release of December 8, 2020.
For United States (2004) sales professionals for finance and business services were coded 50. “Service workers and shop and market sales workers” in the CSES Module 2 Full Release of December 15, 2015. These respondents were recoded to “34. Other associate professionals” for United States (2004) in the CSES IMD Release of December 8, 2020, as the ISCO-88 coding scheme employs the separate code 3410 for finance and sales associate professionals.

IMD2015_ISCO_08 (MAIN OCCUPATION (CSES MODULE 4)): This variable reports a respondent’s main occupation based on ISCO-08 two-digit codes for election studies included in CSES Module 4. The following recodes have been applied:

For Austria (2013) wild code (i.e., codes where the meaning of the classification is not ascertainable) 552 is included in the data which is neither envisaged by ISCO-08, nor documented in the CSES Module 4 Codebook. Data have been set to “999. Missing” for Austria (2013) in the CSES IMD Release of December 8, 2020, as the meaning of the code could not be ascertained.
For Czech Republic (2013) wild code (i.e., codes where the meaning of the classification is not ascertainable) 150 is included in the data which is not envisaged by the ISCO-08 coding scheme and refers to science and engineering professionals. Data have been recoded to “21. Science and engineering professionals” for Czech Republic (2013) in the CSES IMD Phase 3 of December 8.
For New Zealand (2011) wild code (i.e., codes where the meaning of the classification is not ascertainable) 942 is included in the data which is neither envisaged by ISCO-08, nor documented in the CSES Module 4 Codebook. Data have been set to “999. Missing” for New Zealand (2011) in the CSES IMD Phase 3 of December 8, as the meaning of the code could not be ascertained.

IMD2016 (SOCIO-ECONOMIC STATUS): Standalone Modules included several auxiliary codes 5 and 6 for groups not envisaged by the original CSES coding scheme. These codes have been harmonized for IMD as follows: Respondents out of labor force coded as 5 or 6 in the Standalone Modules have been set to “9. Missing” in the CSES IMD Release of December 8, 2020. The remaining non-classifiable groups were set to “5. Other.” Additionally, the following recodes have been applied to individual election studies:

For Austria (2008) public sector workers and officials are assigned code 5 in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “1. White Collar” for Austria (2008) in the CSES IMD Release of December 8, 2020.
For Germany (2009) academics are assigned code 5 while civil servants are assigned code 6 in the CSES Module 3 Full Release of December 15, 2015. These respondents were recoded to “1. White Collar” for Germany (2009) in the CSES IMD Release of December 8, 2020.
For Italy (2006) work-related questions refer to the “head of family” rather than the respondent in the CSES Module 2 Full Release of December 15, 2015. In CSES Module 2, respondents not being the head of family were assigned code 5. These respondents have been recoded to “9. Missing” for Italy (2006) in the CSES IMD Release of December 8, 2020.
For Slovenia (2008) freelancers are assigned code 5. These respondents have been recoded to “4. Self-Employed” for Slovenia (2008) in the CSES IMD Release of December 8, 2020.

IMD2017 (EMPLOYMENT TYPE – PUBLIC OR PRIVATE): The coding scheme for employment type applied in CSES IMD follows the classification used in the Standalone CSES Modules 2-4. Further, Standalone Modules included several auxiliary codes 5 and 6 for employment types not envisaged by the original CSES coding scheme. These codes have been harmonized for IMD as follows: Farmers, self-employed respondents and helping family members coded as 5 or 6 in the Standalone Modules have been recoded to 2. “Private Sector” in the CSES IMD Release of December 8, 2020. Respondents out of labor force coded as 5 or 6 in the Standalone Modules have been set to “9. Missing” in the CSES IMD Release of December 8, 2020. Additionally, the following recodes have been applied to individual election studies:

For Czech Republic (2006 and 2010) state administration and self-government administration employees are assigned code 5. These respondents have been recoded to “1. Public Sector” for Czech Republic (2006 and 2010) in the CSES IMD Release of December 8, 2020.
For Great Britain (2005) wild code (i.e., codes where the meaning of the classification is not ascertainable) 5 is included in the data not documented in the CSES Module 2 Codebook. This variable have been set to “9. Missing” for Great Britain (2005) in the CSES IMD Release of December 8, 2020, as the meaning of the code could not be ascertained.
For Hong Kong (1998 and 2000) wild code (i.e., codes where the meaning of the classification is not ascertainable) 7 is included the data not documented in the CSES Module 1 Codebook. This variable have been set to “9. Missing” for Hong Kong (1998 and 2000) in the CSES IMD Release of December 8, 2020, as the meaning of the code could not be ascertained.
For Italy (2006) work-related questions refer to the “head of family” rather than the respondent. In the CSES Module 2 Full Release of December 15, 2015, respondents not being the head of family were assigned code 5. These respondents have been recoded to “9. Missing” for Italy (2006) in the CSES IMD Release of December 8, 2020.
For Slovenia (2008) agricultural cooperatives and farmers were jointly assigned code 5 in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “3. Mixed” for Slovenia (2008) in the CSES IMD Release of December 8, 2020.

IMD2018 (INDUSTRIAL SECTOR): The following recodes have been applied:

CSES MODULE 1 category “0. Not Applicable” referring to respondents not in labor force was recoded to “9. Missing” in the CSES IMD Release of December 8, 2020.
For Brazil (2006) respondents working for public/state enterprises or NGOs were assigned code “4. Other” in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “3. Tertiary sector” for Brazil (2006) in the CSES IMD Release of December 8, 2020.
For Germany (2002 Telephone) respondents working for public services or public enterprises were assigned code “4. Other” in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “3. Tertiary sector” for Germany (2002 Telephone) in the CSES IMD Release of December 8, 2020.
For Hong Kong (1998) wild code (i.e., codes where the meaning of the classification is not ascertainable) 7 is included the data not documented in the CSES Module 1 Codebook. This value has been set to “9. Missing” for Hong Kong (1998) in the CSES IMD Release of December 8, 2020, as the meaning of the code could not be ascertained.
For Ireland (2002) respondents working for extra-territorial organizations were assigned code “4. Other” in the CSES Module 2 Full Release of December 15, 2015. These respondents have been recoded to “3. Tertiary sector” for Ireland (2002) in the CSES IMD Release of December 8, 2020.
For Italy (2006) work-related questions refer to the “head of family” rather than the respondent in the CSES Module 2 Full Release of December 15, 2015. Respondents not being the head of family were assigned code “5. Other.” These respondents have been recoded to “9. Missing” for Italy (2006) in the CSES IMD Release of December 8, 2020.
For the Netherlands (2006) respondents working for public/state enterprises or NGOs were assigned code “4. Other” in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “3. Tertiary sector” for the Netherlands (2006) in the CSES IMD Release of December 8, 2020.
For Slovenia (2004) respondents working for public/state enterprises or NGOs were assigned code “4. Other” in the CSES Module 2 Full Release of December 15, 2015. These respondents have been recoded to “3. Tertiary sector” for Slovenia (2004) in the CSES IMD Release of December 8, 2020.
For Sweden (2006) respondents who have never been working were assigned code “4. Other” in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “9. Missing” for Sweden (2006) in the CSES IMD Release of December 8, 2020.

IMD2019_1 (UNION MEMBERSHIP OF RESPONDENT): In the Standalone Modules 1-4, respondents not being a member of a union were assigned code 2. For the IMD, these respondents have been recoded to “0. No.”

IMD2019_2 (UNION MEMBERSHIP OF OTHERS IN HOUSEHOLD): In the Standalone Modules 1-4, household members not being a member of a union were assigned code 2. For the IMD, they have been recoded to “0. No.”

IMD2020 (SPOUSE: EMPLOYMENT STATUS): The Standalone CSES Modules 2-4 included several auxiliary codes 11 and 12 for employment statuses originally not envisaged by the original CSES coding scheme. These codes have been harmonized for IMD, thereby expanding the existing coding scheme by code “11. On temporary job leave”. Further, employed spouses from studies not differentiating employment by working hours were mostly coded as “1. Employed – Full Time” in the Standalone Modules, but have been recoded to “0. Employed – no working hours specified” in IMD. Finally, the following recodes have been applied to individual election studies:

For Australia (2013) 96 respondents gave ambiguous or multiple answers for this variable and were assigned code 11 in the CSES Module 4 Full Release of May 29, 2018. These respondents have been recoded to “99. Missing” for Australia (2013) in the CSES IMD Release of December 8, 2020.
For Mexico (2012) respondents without a partner or spouse were assigned code 11 in the CSES Module 4 Full Release of May 29, 2018. These respondents have been recoded to “99. Missing” for Mexico (2012) in the CSES IMD Release of December 8, 2020.
For New Zealand (2011) respondents engaging in an unpaid work outside the home were classified as “4. Employed – less than part time” in the CSES Module 4 Full Release of May 29, 2018. These respondents have been recoded to “10. Others, not in labor force” for New Zealand (2002) in the CSES IMD Release of December 8, 2020.
For New Zealand (2014) 47 respondents gave multiple answers for this variable and were assigned code 11 in the CSES Module 4 Full Release of May 29, 2018. These respondents have been recoded to “99. Missing” for New Zealand (2014) in the CSES IMD Release of December 8, 2020.
For Slovenia (2008) self-employed spouses were assigned code 11 in the CSES Module 3 Full Release of December 15, 2015. These spouses have been recoded to “1. Employed – full time” for Slovenia (2008) in the CSES IMD Release of December 8, 2020, to harmonize data with the Slovenian (2004) study.
For Slovenia (2008) retired spouses with additional employment were assigned code 12 in the CSES Module 3 Full Release of December 15, 2015. These spouses have been recoded to “7. Retired” for Slovenia (2008) in the CSES IMD Release of December 8, 2020, to harmonize data with the Slovenian (2004) study.

IMD2021_ISCO_88 (MAIN OCCUPATION (CSES MODULES 1-3)): This variable reports a spouse’s main occupation based on ISCO-88 two-digit codes for election studies included in CSES Modules 1-3. The following recodes have been applied:

CSES MODULE 1 categories “0. Not Applicable” and “97. No Occupation” were recoded to “999. Missing” in the CSES IMD Release of December 8, 2020.
For Hong Kong (1998 and 2000) wild codes (i.e., codes where the meaning of the classification is not ascertainable) 5, 7, 8 and 95 are included in the data which are neither envisaged by ISCO-88, nor documented in the CSES Module 1 Codebook. These values have been set to “999. Missing” for Hong Kong (1998 and 2000) in the CSES IMD Release of December 8, 2020, as their meaning could not be ascertained.
For Japan (2004) spouses without occupation were assigned code 90 in the CSES Module 2 Full Release of December 15, 2015. These spouses have been recoded to “999. Missing” for Japan (2004 in the CSES IMD Release of December 8, 2020.
For Latvia (2010) spouses employed as skilled workers were coded “60. Skilled agricultural and fishery workers” in the CSES Module 3 Full Release of December 15, 2015. These spouses have been recoded to “996. Other non-classifiable occupations” for Latvia (2010) in the CSES IMD Release of December 8, 2020, as skilled workers are a generic group not matching a single ISCO-88 category.
For Norway (1997) wild codes (i.e., codes where the meaning of the classification is not ascertainable) 25, 63 and 64 are included in the data which are neither envisaged by ISCO-88, nor documented in the CSES Module 1 Codebook. These values have been recoded to “999. Missing” for Norway (1997) in the CSES IMD Release of December 8, 2020, as the meaning of the codes could not be ascertained.

IMD2021_ISCO_08 (SPOUSE: MAIN OCCUPATION (CSES MODULE 4)): This variable reports a spouse’s main occupation based on ISCO-08 two-digit codes for election studies included in CSES Module 4. The following recodes have been applied:

For New Zealand (2011) wild codes 245, 533 and 913 (i.e., codes where the meaning of the classification is not ascertainable) are included in the data which are neither envisaged by ISCO-08, nor documented in the CSES Module 4 Codebook. These values have been recoded to “999. Missing” for New Zealand (2011) in the CSES IMD Release of December 8, 2020, as the meaning of the codes could not be ascertained.

IMD2022 (SPOUSE: SOCIO-ECONOMIC STATUS): Standalone Modules included several auxiliary codes 5 and 6 for groups not envisaged by the original CSES coding scheme. These codes have been harmonized for IMD as follows: Spouses out of labor force coded as 5 or 6 in the Standalone Modules have been set to “9. Missing” in the CSES IMD Release of December 8, 2020. The remaining non-classifiable groups were set to “5. Other.” Additionally, the following recodes have been applied to individual election studies:

For Austria (2008) public sector workers and officials are assigned code 5 in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “1. White Collar” for Austria (2008) in the CSES IMD Release of December 8, 2020.
For Germany (2009) civil servants are assigned code 6 in the CSES Module 3 Full Release of December 15, 2015. These spouses have been recoded to “1. White Collar” for Germany (2009) in the CSES IMD Release of December 8, 2020.
For Slovenia (2008) freelancers are assigned code 5 in the CSES Module 3 Full Release of December 15, 2015. These spouses have been recoded to “4. Self-Employed” for Slovenia (2008) in the CSES IMD Release of December 8, 2020.

IMD2023 (SPOUSE: EMPLOYMENT TYPE – PUBLIC OR PRIVATE): The Standalone Modules included several auxiliary codes 5 and 6 for employment types not envisaged by the original CSES coding scheme. These codes have been harmonized for IMD as follows: Farmers, self-employed spouses and helping family members coded as 5 or 6 in the Standalone Modules have been recoded to “2. Private Sector” in the CSES IMD Release of December 8, 2020. Spouses out of labor force coded as 5 or 6 in the Standalone Modules have been set to “9. Missing” in the CSES IMD Release of December 8, 2020. Additionally, the following recodes have been applied to individual election studies:

For Czech Republic (2006 and 2010) state administration and self-government administration employees are assigned code 5 in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “1. Public Sector” for Czech Republic (2006 and 2010) in the CSES IMD Release of December 8, 2020.
For Great Britain (2005) wild code (i.e., codes where the meaning of the classification is not ascertainable) 5 is included in the data which is not documented in the CSES Module 2 Codebook. This value has been recoded to “9. Missing” for Great Britain (2005) in the CSES IMD Release of December 8, 2020, as the meaning of the code could not be ascertained.

IMD2024 (SPOUSE: INDUSTRIAL SECTOR): The following recodes have been applied:

For Germany (2002 Telephone) respondents working for public services or public enterprises were assigned code “4. Other” in the CSES Module 3 Full Release of December 15, 2015. These respondents have been recoded to “3. Tertiary sector” for Germany (2002 Telephone) in the CSES IMD Release of December 8, 2020.
For Slovenia (2004) respondents working for public services or public enterprises were assigned code “4. Other” in the CSES Module 2 Full Release of December 15, 2015. These respondents have been recoded to “3. Tertiary sector” for Slovenia (2004) in the CSES IMD Release of December 8, 2020.

IMD2025 (BUSINESS OR EMPLOYER ASSOCIATION MEMBERSHIP): In the Standalone Modules 2-4, respondents not being a member of a business or employer association were assigned code 2. For the CSES IMD Release of December 8, 2020, these respondents have been recoded to “0. No.”

IMD2026 (FARMERS ASSOCIATION MEMBERSHIP): In the Standalone Modules 2-4, respondents not being a member of a farmers’ association were assigned code 2. For the CSES IMD Release of December 8, 2020, these respondents have been recoded to “0. No.”

IMD2027 (PROFESSIONAL ASSOCIATION MEMBERSHIP): In the Standalone Modules 2-4, respondents not being a member of a professional association were assigned code 2. For the CSES IMD Release of December 8, 2020, these respondents have been recoded to “0. No.”

IMD3009_E (LIKE-DISLIKE – LEADER E) for Turkey (2015) was set to missing in the CSES IMD Release of December 8, 2020. Data on like-dislike scale for leaders D (Selahattin Demirtas, HDP) and E (Mustafa Kamalak, SP) have the exact same distribution in the CSES Module 4 Full Release of May 29, 2018. Data is correct for Leader D, but should be set to missing for leader E.

IMD5034_1 (FORMAL COMPLAINTS AGAINST NATIONAL LEVEL RESULTS): In the Standalone CSES Modules 3 (of December 15, 2015) and 4 (of May 29,2018), polities that did not experience any formal complaints against the national level election results were assigned code 5. For the CSES IMD Release of December 8, 2020, these polities have been recoded to “0. No.”

IMD5034_2 (ELECTION IRREGULARITIES REPORTED): In the Standalone CSES Modules 3 (of December 15, 2015) and 4 (of May 29,2018), polities for which international election observers did not report election irregularities were assigned code 5. For the CSES IMD Release of December 8, 2020, these polities have been recoded to “0. No.”

IMD5036_1 (ELECTORAL ALLIANCES PERMITTED IN AN ELECTION): In the Standalone CSES Modules 1-4, value “No” is coded 5. For the CSES IMD Release of December 8, 2020, this has been recoded to “0. No.”

IMD5036_1 (ELECTORAL ALLIANCES PERMITTED IN AN ELECTION): For Australia (1996 and 2004) this variable is recoded from “Yes” (category 1) into correct value “No” (category 0) in the CSES IMD Release of December 8, 2020.

IMD5036_1 (ELECTORAL ALLIANCES PERMITTED IN AN ELECTION): For Sweden (2006 and 2014) this variable is recoded from “Yes” (category 1) into correct value “No” (category 0) in the CSES IMD Release of December 8, 2020.

IMD5036_1 (ELECTORAL ALLIANCES PERMITTED IN AN ELECTION): For Canada (2008, 2011 and 2015) this variable is recoded from “No” (category 0) into correct value “Yes” (category 1) in the CSES IMD Release of December 8, 2020.

IMD5036_3 (DID ANY ELECTORAL ALLIANCE FORM?) this variable is recoded from “0. No” into correct value “7. Not applicable” in the CSES IMD Release of December 8, 2020 for the following election studies (listed in alphabetical order):

Denmark (1998)
Japan (1996)
Sweden (1998)
Taiwan (1996)
Thailand (2001)
USA (1996).

Those election studies have not been coded as “not applicable” before because this category did not exist in the CSES Module 1 Full Release of December 15, 2015.

IMD5038 (REQUIREMENTS OF JOINT PARTY LISTS): For Great Britain (2015) this variable is incorrectly coded “5. No, joint parties are governed by the same rules as other parties” in the CSES Module 4 Full Release of May 29, 2018. The value has been correctly coded into “7. Not applicable” for Great Britain (2015) in the CSES IMD Release of December 8, 2020.

IMD5040_1 (MULTI PARTY ENDORSEMENTS): In the Standalone CSES Modules 1-4, polities in which multi-party endorsements were not permissible were assigned code 5. For the CSES IMD Release of December 8, 2020, these polities have been recoded to 0. Additionally, the following recodes have been applied:

For Brazil (2002) this variable is incorrectly coded “5. No” in the CSES Module 2 Full Release of December 15, 2015, suggesting that candidates could not run with the endorsement of more than one party. However, multi-party endorsements were permissible for all Brazilian elections included in CSES. Therefore, data has been recoded to “1. Yes” for Brazil (2002) in the CSES IMD Release of December 8, 2020.
For Iceland (all studies) this variable is incorrectly coded “1. Yes”, suggesting candidates could run with the endorsement of more than one party. However, since 1999, multi-party endorsements are not permissible in Iceland. Therefore, data have been recoded to “0. No” for Iceland (1999, 2003, 2007, 2009, 2013) in the CSES IMD Release of December 8, 2020.
For Japan (1996 and 2004) this variable is incorrectly coded “5. No” in CSES Modules 1 and 2 Full Releases of December 15, 2015, suggesting that candidates could not run with the endorsement of more than one party. However, multi-party endorsements were permissible for all Japanese elections included in CSES. Therefore, data has been recoded to “1. Yes” for Japan (1996 and 2004) in the CSES IMD Release of December 8, 2020.

IMD5040_2 (MULTI-PARTY ENDORSEMENT ON BALLOT): The coding scheme for multi-party endorsement on ballot applied in the CSES IMD Release of December 8, 2020 follows the classification used in the Standalone CSES Modules 2-4. Additionally, the following recodes have been applied:

For Brazil (2002) this variable is incorrectly coded “7. Not applicable” in the CSES Module 2 Full Release of December 15, 2015, suggesting that candidates could not run with the endorsement of more than one party. However, multi-party endorsements were permissible for all Brazilian elections included in CSES, even if they are not reflected on the ballot. Therefore, data have been recoded to “2. No party endorsements are indicated on the ballot paper” for Brazil (2002) in the CSES IMD Release of December 8, 2020.
For France (2002). This variable is incorrectly coded 3 in the CSES Module 2 Full Release of December 15, 2015, suggesting that multi-party endorsements are reflected on the ballot paper with the “candidate’s name appearing once, together with the names of all supporting parties”. However, the French (2002) election is concerned with Presidential elections, for which ballots indicate the candidate’s name only, without any reference to party endorsements. Therefore, data has been recoded to “2. No party endorsements are indicated on the ballot paper” for France (2002) in the CSES IMD Release of December 8, 2020.
For Great Britain (1997 and 2005) this variable is incorrectly coded “5. Yes, other” in CSES Modules 1 and 2 Full Releases of December 15, 2015. However, in Great Britain, candidates are permitted to put up a short description of themselves on the ballot paper, including party labels. Therefore, data have been recoded to “3. Yes, candidate’s name appears once, together with the names of all supporting parties” for Great Britain (1997 and 2005) in the CSES IMD Release of December 8, 2020.
For Japan (1996, 2007, 2013) this variable is incorrectly coded “7. Not applicable” in CSES Modules 1, 3 (Full Releases of December 15, 2015), and 4 (Full Release of May 29, 2018), suggesting that multi-party endorsements were not permissible in Japan. However, for all Japanese studies included in CSES, candidates could run with the endorsement of more than one party, even if not reflected on the ballot. Therefore, data have been recoded to “1. No” for Japan (1996), 2007 and 2013) in the CSES IMD Release of December 8, 2020.
For Iceland (all studies) this variable were coded inconsistently. However, since 1999, multi-party endorsements are not permissible in Iceland. Therefore, data in have been recoded to “7. Not applicable” for Iceland (1999, 2003, 2007, 2009, 2013) in the CSES IMD Release of December 8, 2020.
For Ireland (2002) this variable is incorrectly coded “1. No” in CSES Module 2 Full Release of December 15, 2015, suggesting that multi-party endorsements are not indicated on the ballot paper. However, in case of multi-party endorsements, the candidate’s name appears once, together with the names of all supporting parties. Therefore, data has been recoded to “3. Yes, candidate’s name appears as many times as there are different parties endorsing him/her, each time with the name of endorsing party” for Ireland (2002) in the CSES IMD Release of December 8, 2020.
For New Zealand (1996) this variable is incorrectly coded 3 in CSES Module 1 Full Release of December 15, 2015, suggesting that multi-party endorsements were reflected on the ballot paper with the “candidate’s name appearing once, together with the names of all supporting parties.” However, although candidates may have an endorsement from another party, usually the main party of the candidate is the one appearing on the ballot paper. Therefore, data has been recoded to “1. No” for New Zealand (1996) in the CSES IMD Release of December 8, 2020.
For Portugal (2002) this variable is incorrectly coded “5. Yes, other” in CSES Modules 1 and 2 Full Releases of December 15, 2015. However, multi-party endorsements in Portugal are not permissible, unless parties decide to form a pre-electoral coalition, running a joint list. Therefore, data have been recoded to “7. Not applicable” for Portugal (2002) in the CSES IMD Release of December 8, 2020.
For United States (1996 and 2004) this variable is incorrectly coded 3 in CSES Modules 1 and 2 Full Releases of December 15, 2015, suggesting that multi-party endorsements are reflected on the ballot paper with the “candidate’s name appearing once, together with the names of all supporting parties.” However, with the electoral rules varying by state, nearly all the options distinguished by IMD5040_2 are possible in the United States. Therefore, data have been recoded to “5. Other” for United States (1996 and 2004) in the CSES IMD Release of December 8, 2020.

IMD5041_1 (NUMBER OF ELECTORAL SEGMENTS (TIERS)): This variable is incorrectly coded 1 (one electoral segment) in CSES Modules 1 and 2 Full Releases of December 15, 2015 for the following election studies (listed in alphabetical order):

Denmark (1998 and 2001)
Iceland (1999 and 2003)
Norway (1997 and 2001)
Sweden (1998 and 2002)

Data was recoded to correct value 2 (two electoral segments) in the CSES IMD Release of December 8, 2020.

IMD5041_1 (NUMBER OF ELECTORAL SEGMENTS (TIERS)): For Mexico (2003) this variable is incorrectly coded 1 (one electoral segment) in the CSES Module 2 Full Releases of December 15, 2015. Data was recoded to correct value 2 (two electoral segments) for Mexico (2003) in the CSES IMD Release of December 8, 2020.

IMD5041_1 (NUMBER OF ELECTORAL SEGMENTS (TIERS)): For Greece (2009) this variable is incorrectly coded 2 (two electoral segments) in the CSES Module 3 Full Release of December 15, 2015. Therefore, data has been recoded to correct value 3 (three electoral segments) for Greece (2009) in the CSES IMD Release of December 8, 2020.

IMD5041_1 (NUMBER OF ELECTORAL SEGMENTS (TIERS)): For Poland (1997) this variable is incorrectly coded 2 (two electoral segments) in the CSES Module 1 Full Release of December 15, 2015. Therefore, data has been recoded to correct value 1 (one electoral segment) for Poland (1997) in the CSES IMD Release of December 8, 2020.

IMD5041_2 (LINKED ELECTORAL SEGMENTS (TIERS)): For Norway (2005) this variable is incorrectly coded “7. Not applicable” in the CSES Module 3 Full Release of December 15, 2015. Therefore, data has been recoded into correct value “1. Yes” for Norway (2005) in the CSES IMD Release of December 8, 2020.

IMD5041_2 (LINKED ELECTORAL SEGMENTS (TIERS)): For Mexico (2006, 2012 and 2015) this variable is incorrectly coded “1. Yes” in CSES Standalone Modules. Therefore, data have been recoded into correct value “9. Missing” for Mexico (2006, 2012 and 2015) in the CSES IMD Release of December 8, 2020.

IMD5042_1 (DEPENDENT FORMULA IN MIXED SYSTEMS): For Philippines (2010) this variable is incorrectly coded “7. Not applicable” in the CSES Module 3 Full Release of December 15, 2015. Therefore, data have been recoded into correct value “3. Dependent” for Philippines (2010) in the CSES IMD Release of December 8, 2020.

IMD5042_1 (DEPENDENT FORMULA IN MIXED SYSTEMS): For Philippines (2016) this variable is incorrectly coded “1. Independent” in the CSES Module 4 Full Release of May 29, 2018. Therefore, data have been recoded into correct value “3. Dependent” for Philippines (2016) in the CSES IMD Release of December 8, 2020.

IMD5042_1 (DEPENDENT FORMULA IN MIXED SYSTEMS): For Taiwan (2008) this variable is incorrectly coded “7. Not applicable” in the CSES Module 3 Full Release of December 15, 2015. Therefore, data have been recoded into correct value “1. Independent” for Taiwan (2008) in the CSES IMD Release of December 8, 2020.

IMD5042_1 (DEPENDENT FORMULA IN MIXED SYSTEMS): For Thailand (2007) this variable is incorrectly coded “7. Not applicable” in the CSES Module 3 Full Release of December 15, 2015. Therefore, data have been recoded into correct value “9. Missing” for Thailand (2007) in the CSES IMD Release of December 8, 2020.

IMD5042_2 (SUBTYPES OF MIXED ELECTORAL SYSTEMS): For Philippines (2010) this variable is incorrectly coded “7. Not applicable” in the CSES Module 3 Full Release of December 15, 2015. Therefore, data have been recoded into correct value “4. Correction” for Philippines (2010) in the CSES IMD Release of December 8, 2020.

IMD5042_2 (SUBTYPES OF MIXED ELECTORAL SYSTEMS): For Philippines (2016) this variable is incorrectly coded “72. Superposition” in the CSES Module 4 Full Release of May 29, 2018. Therefore, data have been recoded into correct value “4. Correction” for Philippines (2016) in the CSES IMD Release of December 8, 2020.

IMD5044 (FUSED VOTE): In the Standalone CSES Modules 3 (of December 15, 2015) and 4 (of May 29, 2018), polities in which there was no fused vote were assigned code 5. For the CSES IMD Release of December 8, 2020, these polities have been recoded to “0. No.” Additionally, not applicable codes for IMD5044 have been re-defined for the CSES IMD Release of December 8, 2020: Elections are now coded not applicable either if there is no directly elected President in the polity, or if Presidential and legislative elections were not scheduled on the same day, rendering a fused vote for both institutions impossible.

IMD5045_1 (AVERAGE DISTRICT MAGNITUDE – LOWER HOUSE (TIER 1)): The following recodes have been applied:

For Iceland (1999) the first electoral segment is incorrectly coded as having an average district magnitude of 6.20 in the CSES Module 1 Full Release of December 15, 2015. As the first tier returned 50 seats from 8 multi-member districts, the value has been recoded to 6.25 for Iceland (1999) in the CSES IMD Release of December 8, 2020.
For Iceland (2003) the first electoral segment is incorrectly coded as having an average district magnitude of 10.5 in the CSES Module 2 Full Release of December 15, 2015. However, for the 2003 election, constitutional amendments reduced the number of multi-member districts to 6, returning 54 seats in total. Therefore, data have been recoded to 9 for Iceland (2003) in the CSES IMD Release of December 8, 2020.
For Norway (1997) the first electoral segment is incorrectly coded as having an average district magnitude of 8.60 in the CSES Module 1 Full Release of December 15, 2015. However, for the 1997 election, the first electoral segment returned 157 seats in 19 multi-member districts. Therefore, data have been recoded to 8.26 for Norway (1997) in the CSES IMD Release of December 8, 2020.
For Spain (1996 and 2000) data is incorrectly coded as 7.00 in CSES Module 1 Full Release of December 15, 2015. As Spain operates one electoral segment with 350 seats being allocated in 52 multi-member districts data have been recoded to 6.73 for Spain (1996 and 2000) in the CSES IMD Release of December 8, 2020.
For Taiwan (2001 and 2004) the first electoral segment is incorrectly coded as having an average district magnitude of 5.68 in CSES Module 2 Full Release of December 15, 2015. However, for the respective elections, the first tier consisted of 168 seats from 29 multi-member districts. Therefore, data have been recoded to 5.79 for Taiwan (2001 and 2004) in the CSES IMD Release of December 8, 2020.
For Turkey (2011) data is incorrectly coded as 6.96 in CSES Module 3 Full Release of December 15, 2015. However, in the 2011 election, Turkey operated a one-tier electoral system with 550 seats awarded in 85 multi-member districts. Therefore, data has been recoded to 6.47 for Turkey (2011) in the CSES IMD Release of December 8, 2020.

IMD5045_2 (AVERAGE DISTRICT MAGNITUDE – LOWER HOUSE (TIER 2)): The following recodes have been applied:

For Denmark (1998 and 2001) data is incorrectly coded as having one electoral tier in CSES Modules 1 and 2 Full Releases of December 15, 2015. However, for the respective elections, Denmark employed two electoral segments, with the second tier awarding 40 compensatory seats on the national level. Therefore, data have been recoded to 40.00 for Denmark (1998 and 2001) in the CSES IMD Release of December 8, 2020.
For Germany (1998, 2002, 2005, and 2009) data were coded inconsistently. Inconsistencies occurred because average district magnitude in the second tier in Germany is liable to change depending on whether overhang or compensatory seats are allotted or not. Data in IMD5045_2 reflects the average district magnitude disregarding compensatory seats. Therefore, data has been recoded to 20.50 for Germany (1998) and 18.69 for Germany (2002), Germany (2005) and Germany (2009) in the CSES IMD Release of December 8, 2020.
For Hong Kong (1998 and 2000) data is incorrectly coded as having two electoral segments in CSES Module 1 Full Release of December 15, 2015, with the second segment referring to traditional functional constituency vote. However, since suffrage for the traditional FCs is limited, Hong Kong should be regarded as one-tier system until electoral reform in 2010. Therefore, data have been recoded to “997. Not applicable” for Hong Kong (1998 and 2000) in the CSES IMD Release of December 8, 2020.
For Iceland (1999) data is incorrectly coded as having one electoral segment in CSES Module 1 Full Release of December 15, 2015. However, Iceland employs two tiers for lower house elections. As the second national-level tier returned 13 seats in 1999, data have been recoded to 13.00 for Iceland (1999) in the CSES IMD Release of December 8, 2020.
For Iceland (2003) data is incorrectly coded as having one electoral segment in CSES Module 2 Full Release of December 15, 2015. However, Iceland employs two tiers for lower house elections. As the second national-level tier returned 9 seats in 2003, data have been recoded to 9.00 for Iceland (2003) in the CSES IMD Release of December 8, 2020.
For Mexico (2003) data is incorrectly coded as having one electoral segment in CSES Module 2 Full Release of December 15, 2015. However, for all studies included in CSES, Mexico employs a two-tier system for lower house elections, with the second tier returning 200 seats from 5 multi-member districts. Therefore, data have been recoded to 40.00 for Mexico (2003) in the CSES IMD Release of December 8, 2020.
For Norway (1997 and 2001) data were incorrectly coded as having one electoral tier in CSES Modules 1 and 2 Full Releases of December 15, 2015. However, for the respective elections, Norway employed two electoral segments, with the second tier awarding 8 adjustment seats on the national level. Hence, data have been recoded to 8.00 for Norway (1997 and 2001) in the CSES IMD Release of December 8, 2020.
For Slovenia (1996 and 2004) data is incorrectly coded as 1.00 in CSES Modules 1 and 2 Full Releases of December 15, 2015. That data referred to the two seats awarded to Hungarian and Italian minorities. In accordance with later Slovenian studies, data have been recoded to “997. Not applicable” for Slovenia (1996 and 2004) in the CSES IMD Release of December 8, 2020.
For Sweden (1998 and 2002) data were incorrectly coded as having one electoral tier in CSES Modules 1 and 2 Full Releases of December 15, 2015. However, for the respective elections, Sweden employed two electoral segments, with the second tier awarding 39 adjustments seats on the national level. Therefore, data have been recoded to 39.00 for Sweden (1998 and 2002) in the CSES IMD Release of December 8, 2020.
For Taiwan (2001 and 2004) data were incorrectly coded as having one electoral tier in CSES Module 2 Full Release of December 15, 2015. However, for the respective elections, Taiwan employed two electoral segments, with the second tier awarding 41 seats in a nation-wide district. Therefore, data have been recoded to 41.00 for Taiwan (2001 and 2004) in the CSES IMD Release of December 8, 2020.

IMD5046_1-4 (TRANSFERABLE VOTE): In the Standalone CSES Modules 2-4, polities that did not employ transferable votes were assigned code 5. For the CSES IMD Release of December 8, 2020, these polities have been recoded to “0. No.” Additionally, not applicable codes for IMD5046_ have been re-defined for the CSES IMD Release of December 8, 2020: Elections are now coded not applicable either if the electoral system employed does not have a second tier (IMD5046_2 & 4), or if the parliament is unicameral (IMD5046_3 & 4).

IMD5048 (REGIME: TYPE OF EXECUTIVE): The following recodes have been applied:

For Portugal (2015) data is incorrectly coded “1. Parliamentary regime” in CSES Module 4 Full Release of May 29, 2018. As Portugal’s President is directly elected and the government is accountable to both the President and parliament data have been recoded to “2. Mixed Regime” for Portugal (2015) in the CSES IMD Release of December 8, 2020.
For South Korea (2008) data is incorrectly coded “1. Parliamentary regime” in CSES Module 3 Full Release of December 15, 2015. As South Korea’s President is directly elected and parliament is not empowered to remove the government from office data have been recoded to “3. Presidential Regime” for South Korea (2008) in the CSES IMD Release of December 8, 2020.
For Taiwan (2008) data is incorrectly coded “3. Presidential regime” in CSES Module 3 Full Release of December 15, 2015. The executive Yuan in Taiwan is responsible to both the legislative Yuan and the President. Therefore, data have been recoded to “2. Mixed regime” for Taiwan (2008) in the CSES IMD Release of December 8, 2020.

IMD5061 (CONSTITUTIONAL FEDERAL STRUCTURE): In the Standalone CSES Modules 3 (of December 15, 2015) and 4 (of May 29, 2018), polities that did not employ a constitutional federal structure were assigned code 5. For the CSES IMD Release of December 8, 2020, these polities have been recoded to “0. No.”

IMD5061 (CONSTITUTIONAL FEDERAL STRUCTURE): For Philippines (2010) data is incorrectly coded “1. Yes” (federal state). However, the Philippines are a unitary presidential constitutional republic. Therefore, data have been recoded to “0. No” for Philippines (2010) in the CSES IMD Release of December 8, 2020.

Updates made to Phase II Release of October 17, 2019 since Phase I Release of December 4, 2018

CSES Integrated Module Dataset Phase II Release: Errata

Posted: October 17, 2019

Updates made to Phase II Release of October 17, 2019 since Phase I Release of December 4, 2018

Phase II of the CSES Integrated Module Dataset (IMD) of October 17, 2019, contains additions, updates, and improvements on the Phase I release of December 4, 2018. We advise users not to 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.

Overall, CSES IMD sees a set of 40+ variables being added to the dataset. The new variables added to IMD Phase II from IMD Phase I are listed below in the data changes section.

In addition to the extensive set of new variables, we provide users with a list of other changes and corrections made to CSES IMD Phase II compared to CSES IMD Phase I. We have divided this into data changes (prefix D), codebook changes (prefix C), and explanations concerning deviations between the CSES IMD and the Standalone CSES Modules (prefix S). Users are advised that this list is not comprehensive but instead highlights the main changes since CSES IMD Phase I, which was released on December 4, 2018.

DATA CHANGES

The following survey level variables have been added to the dataset:
- - - IMD3014 GOVERNMENT PERFORMANCE: GENERAL: This variable is available for election studies included in CSES Module 2 and CSES Module 3.
    - IMD3015_1-IMD3015_4 DICHOTOMIZED POLITICAL INFORMATION ITEMS.
    - IMD3015_A-IMD3015_D POLITICAL INFORMATION SCALES BY MODULE: Scales indicate the number of correct answers each respondent provided to the political information items included in the Standalone CSES Modules.
Several macro-level variables have been added to the dataset. These are:
- - - IMD5001_A-IMD5001_I Percent Vote Lower House ELECTION: Parties A-I.
    - IMD5002_A-IMD5002_I Percent Seats Lower House ELECTION: Parties A-I.
    - IMD5003_A-IMD5003_I Percent Vote Upper House ELECTION: Parties A-I.
    - IMD5004_A-IMD5004_I Percent Seats Upper House ELECTION: Parties A-I.
    - IMD5005_A-IMD5005_I Percent Vote, President: Parties A-I.
    - IMD5008_1-IMD5008_2 Party of the Prime Minister before and AFTER.
    - IMD5009_1-IMD5009_2 Party of the President before and AFTER
    - IMD5011_A-IMD5011_I Ideological Family: Parties A-I.
    - IMD5016_1-IMD5016_4 Votes Cast: across multiple tiers for the lower and upper house elections.
    - IMD5017_1-IMD5017_4 Voting procedure: across multiple tiers for the lower and upper house elections.
    - IMD5018_1-IMD5018_4 Voting rounds: across multiple tiers for the lower and upper house elections.
    - IMD5021_1-IMD5021_4 is there party threshold: across multiple tiers in the lower and upper house elections. This variable is available for election studies included in CSES Module 3 and CSES Module 4.
    - IMD5022_1-IMD5022_4 party threshold: across multiple tiers, in the lower and upper house This variable is available for election studies included in CSES Module 3 and CSES Module 4.
    - IMD5024_1-IMD5024_3 DATE election scheduleD.
    - IMD5025_1-IMD5015_3 DATE election HELD.
    - IMD5026_1 Number of legislative chambers.
    - IMD5026_2 Number of elected legislative chambers.
    - IMD5027 Size of the lower house.
    - IMD5028 Size of Cabinet Before: This variable is available for election studies included in CSES Module 2, CSES Module 3, and CSES Module 4.
    - IMD5029_A-IMD5029_I Number of Portfolios before: Parties A-I. This variable is available for election studies included in CSES Module 2, CSES Module 3, and CSES Module 4.
    - IMD5030 Size of Cabinet After: This variable is available for election studies included in CSES Module 2, CSES Module 3, and CSES Module 4.
    - IMD5031_A-IMD5031_I Number of Portfolios After: Parties A-I. This variable is available for election studies included in CSES Module 2, CSES Module 3, and CSES Module 4.
    - IMD5049 Age of the current regime.
    - IMD5052_1-IMD5052_3 GDP Growth: data for the election year and the two preceding years.
    - IMD5053_1-IMD5053_3 GDP per capita: data for the election year and the two preceding years.
    - IMD5054_1-IMD5054_3 Unemployment, total: data for the election year and the two preceding years.
    - IMD5055_1-IMD5055_3 Human development index: data for the election year and the two preceding years.
    - IMD5056_1-IMD5056_3 Inflation, GDP deflator: data for the election year and the two preceding years.
    - IMD5057_1-IMD5057_3 Population, total: data for the election year and the two preceding years.
    - IMD5058_1 Effective number of electoral parties.
    - IMD5058_2 Corrected effective number of electoral parties: This variable is available for election studies included in CSES Module 3 and CSES Module 4.
    - IMD5059_1 Effective number of parliamentary parties.
    - IMD5059_2 Corrected effective number of parliamentary parties: This variable is available for election studies included in CSES Module 3 and CSES Module 4.
IMD3009_A (Like/Dislike Leader A) was previously coded as missing for Israel 2013 due to an issue in the CSES Module 4 dataset. The variable was recoded, using the correct data for IMD3009_A.
IMD3002_UH_DC_1 (Current Upper House Election: Vote Choice – District Candidate 1) was previously coded as missing for Mexico (2006) in the CSES IMD Dataset. However, there was data available for this variable in the CSES Module 3 dataset, which was added for CSES IMD Phase 2.
Data for variable IMD3009_H (Like-Dislike Leader H) was added for Taiwan (2008).
For the Czech Republic (2010), data at the survey level for the party “2030003. Christian D. Union – Czech People’s Party (KDU-CDL)” was previously coded in variables IMD3007_G (Left-Right – Party G), IMD3008_G (Like-Dislike – Party G), and IMD3009_G (Like-Dislike – Leader G). For IMD Phase 2, the KDU-CDL has been assigned the alphabetical code F, however. Thus, data for the KDU-CDL is now coded in IMD3007_F, IMD3008_F, and IMD3009_F.
The Hungarian Party “Hungarian Democratic Forum (MDF)” was previously assigned the numerical code 3490007 and was recoded to the correct numerical code 3480007 in variables IMD3002_LH_DC (Current Lower House Election: Vote Choice – District Candidate), IMD3002_LH_PL (Current Lower House Election: Vote Choice – Party List), IMD3005_3 (Party Identification: Who), and IMD5000_G (Party G Identifier – Numerical).
The Bulgarian Party “Euroleft” was previously assigned two different numerical party codes, 1000040 and 1000041. The codes were harmonized to 1000040, and variable IMD3004_LH_PL (Previous Election Lower House: Vote Choice – Party List) for Bulgaria (2001) was recoded accordingly.
The Philippine Party “Let Life Prosper” was previously assigned two different numerical party codes, 6080144 and 6080145. The codes were harmonized to 6080144, and variable IMD3004_LH_PL (Previous Election Lower House: Vote Choice – Party List) for the Philippines (2016) was recoded accordingly.
IMD5000_H (Party H Identifier – Numerical) was previously coded as “9999999. MISSING” and has been recoded to “6430039. Peace and Democratic Party (BDP)” for Turkey (2011).
IMD5000_G (Party H Identifier – Numerical) and IMD5000_H (Party H Identifier – Numerical) were previously coded as “9999999. MISSING” and have been recoded to “6430039. Agrarian Party” (Party G) and “6430005. Pensioners Party” (Party H) for Russia (2004).
IMD5000_A (Party A Identifier – Numerical) and IMD5000_B (Party B Identifier – Numerical) were previously coded as “9999999. MISSING” and have been recoded to “4400101. A. Paulauskas (Independent)” (Party A) and “4400102. V. Adamkus (Independent)” (Party B) for Lithuania (1997).
IMD5000_H (Party H Identifier – Numerical) and IMD5000_I (Party I Identifier – Numerical) were previously coded as “9999999. MISSING” and have been recoded to “2500020. New Center (NC)” (Party H) and “2500018. Radical Party of the Left (RPG)” (Party I) for France (2007).
IMD5000_D (Party D Identifier – Numerical) was previously coded as “9999999. MISSING” and has been recoded to “64220300. Calin-Constantin-Anton Popescu-Tariceanu (Independent Candidate).”
IMD5000_F (Party F Identifier – Numerical) was previously coded as “9999999. MISSING” and has been recoded to “2030074. SNK European Democrats (SNK ED)” for the Czech Republic (2006).
IMD5000_G (Party G Identifier – Numerical) was previously coded as “9999999. MISSING” and has been recoded to “7050008. Slovenian National Party (SNP)” for Slovenia (1996).
IMD5000_I (Party I Identifier – Numerical) was previously coded as “9999999. MISSING” and has been recoded to “2760009. National Democratic Party (NDP)” for Germany (2013).
IMD5000_D (Party D Identifier – Numerical) was previously coded as “9999999. MISSING” and has been recoded to “1580007. Non-Partisan Solidarity Union (NPSU)” for Taiwan (2012).
Data for variables IMD5000-A-IMD5000_I (Party A-I Identifier – Numerical) were added for Albania (2005), Montenegro (2012), New Zealand (2008), and Sweden (2002).
Data for all macro variables related to the alphabetical CSES party codes A-I were added for Party H (People’s Monarchist Party, PPM) and Party I (Humanist Party, PH) for Portugal (2002).
Data for Variables IMD5012_A-IMD5012_E (Left-Right Expert Placement – Party A-E) was added for Iceland (1999).
Data for Variables IMD5012_A-IMD5012_E (Left-Right Expert Placement – Party A-E) was added for Russia (2000).
IMD1014_2 (Interview Timing – Number of Days Between Interview and First Second Round of Election), a number of cases were coded as “9999. MISSING” even though no second round took place. These cases were recoded to “9996. NOT APPLICABLE: NO SECOND ROUND.”
Variables IMD5000_C, IMD5000_E, and IMD5000_F (Party C, E, and F Identifier – Numerical) were recoded to missing for Belarus (2001).
Variable IMD5000_I (Party I Identifier – Numerical) was recoded to missing for Brazil (2006).
Variable IMD5000_I (Party I Identifier – Numerical) was recoded to missing for Slovakia (2010).
Variable IMD5000_G (Party G Identifier – Numerical) was recoded to missing for Latvia (2010).
Variable IMD3001_UH (Turnout – Current Upper House Election) was recoded from “9999996. NOT APPLICABLE: NO UPPER HOUSE ELECTION” to “999999. MISSING.”

CODEBOOK CHANGES

The CSES IMD Codebook now consists of four separate sections (previously three). Part 3 documents the harmonized numerical party and coalition codes. Part 4 documents the alphabetical party/coalition codes and the alphabetical leader codes.
The CSES IMD Codebook now contains a Bibliography detailing data sources (see part 1).
Parts 3 and 4 of the Codebook (Harmonized Numerical Party Codes and Alphabetical Party and Leader Codes) have been updated. More parties, for which information was previously not listed have been added, and parties or leaders for which there is no data included in CSES IMD Phase 2 have been deleted.
Variable names in the Codebook have been updated to correct for small inconsistencies in the names between the dataset and the Codebook.
For Romania (2014), the numerical party code 6420300 (Calin-Constantin-Anton Popescu-Tariceanu) was added to the CSES IMD Codebook Part 3.
The Hungarian Party “Hungarian Democratic Forum (MDF)” was previously assigned the numerical code 3490007 and was reassigned to the correct numerical code 3480007 in Part 3 of the codebook.
The Philippine Party “Let Life Prosper” was previously assigned two different numerical party codes, 6080144 and 6080145. The codes were harmonized to 6080144 listed accordingly in Part 3 of the codebook.
The Portuguese party Earth Party (MPT) was added to Codebook Part 4 (Party G, Portugal 2002).
For Russia (2004), two additional parties were assigned alphabetical codes, which were documented in the CSES IMD Codebook Part 4. The Agrarian Party was assigned the alphabetical code G, the Pensioners Party was assigned the alphabetical code H.
For Slovenia (1996), the Slovenian National Party was assigned the alphabetical code PARTY G and was added to CSES IMD Codebook Part 4.
For Chile (1999), the independent candidate Sara Larrain was assigned the alphabetical party code PARTY E and added to the CSES IMD Codebook Part 4. An election study note was also added to explain that relational data is only available for IMD5005_E (Percent Vote – President – Party E).
Lithuania (1997), the independent candidate A. Paulauskas was assigned alphabetical party code PARTY A, and the independent candidate V. Adamkus was assigned alphabetical party code B. Both were added to the CSES IMD Codebook Part 4. An election study note was also added to explain that relational data is only available for IMD5005_A (Percent Vote – President – Party A) and IMD5005_B (Percent Vote – President – Party B).
For France (2007), the alphabetical party codes PARTY H (New Center, NC) and PARTY I (Radical Party of the Left, RPG) were added to the CSES IMD Codebook Part 4.
For Germany (2013), the alphabetical party code PARTY I (National Democratic Party, NDP) was added to the CSES IMD Codebook Part 4.
For Taiwan (2012), the alphabetical party code PARTY D (Non-Partisan Solidarity Union, NPSU) was added to the CSES IMD Codebook Part 4.
For Belarus (2001), the alphabetical party codes C, E, and F were removed from the CSES IMD Codebook Part 3 as there is no relational data available in the dataset for any of the three respective parties (Republican Party of Labor and Justice, Belarusian Social-Democratic Assembly, Belarusian Socialist Sporting Party).
For Brazil (2006), the alphabetical party code for PARTY I was removed from the CSES IMD Codebook Part 3 as there is no relational data available in the dataset for the respective party (Socialism and Freedom Party, PSOL).
For Latvia (2010), the alphabetical party code for PARTY G was removed from the CSES IMD Codebook Part 3 as there is no relational data available in the dataset for the respective party (Made in Lativa).
For Mexico (2000), the alphabetical party codes for PARTY G and PARTY H were removed from the CSES IMD Codebook Part 3 as there is no relational data available in the dataset for these parties (National Action Party and Democratic Revolution Party).
For the Czech Republic (2006), PARTY F (SNK European Democrats, SNK ED) was added to the CSES IMD Codebook Part 4.
For Austria (2008), the documentation of alphabetical leader codes was removed since there is no data available for the relevant variables (IMD3009_A-IMD3009_I)
For Brazil (2010), Leader F (Circo Gomes, PSB) was added to the table “ALPHABETICAL LEADER CODES BY CSES MODULE: BRAZIL” in Part 4 of the CSES IMD Codebook
For Slovenia (1996), Leader G (Zmago Jelincic, SNS) was added to the table “ALPHABETICAL LEADER CODES BY CSES MODULE: SLOVENIA” in Part 4 of the CSES IMD Codebook.
For Taiwan (2008), Leader I (Su Tseng-Chang, DPP) was added to the table “ALPHABETICAL LEADER CODES BY CSES MODULE: TAIWAN” in Part 4 of the CSES IMD Codebook and Vincent Siew (KMT) was assigned Leader H.
For Portugal (2002), Leader E (Franciso Louca, BE) and Leader F (Garcia Pereira, PCTP/MRPP) were added to the table “ALPHABETICAL LEADER CODES BY CSES MODULE: PORTUGAL” in Part 4 of the CSES IMD Codebook.
The following leader entries in the CSES IMD Codebook Part 4 were removed because there is no corresponding data in the dataset (IMD Variables IMD3009_A-IMD3009_I): Leader C (Romania 1996, USA 1996), Leader D (Australia 2013, Kenya 2013, South Korea 2000, Ukraine 1998), Leader F (Peru 2011, Slovenia 1996), Leader G (Germany 2009, Netherlands 1998), Leader H (Peru 2016, Serbia 2012, Sweden 2006) and Leader I (Slovakia 2016).
For Turkey (2011), the Peoples’ Democratic Party (HDP) had erroneously been documented as PARTY D in the CSES IMD Codebook Part 4. In the data, PARTY D refers to the Peace and Democratic Party (BDP), and this was updated accordingly.

DEVIATIONS BETWEEN CSES IMD AND STANDALONE CSES MODULES

Below, we detail the primary deviations between Standalone CSES Modules and CSES IMD Phase II. CSES makes every effort to ensure that these deviations are minimal and that where they occur that details concerning them are listed in the CSES IMD Codebook and also here.

The following variables are not available in Standalone CSES Modules 1 and 2 but are available in CSES IMD. They are:
- - IMD5008_ Party of the Prime Minister Before and After
  - IMD5009_ Party of the President Before and After
  - IMD5018_ Voting rounds
  - IMD5025_ Date Election Held
  - IMD5027_ Size of the Lower House
  - IMD5049_Age of Current Regime
  - IMD5052_GDP Growth – Annual % – time t; time t-1; time t-2
  - IMD5053_GDP Per Capita, PPP- time t; time t-1; time t-2
  - IMD5054_Unemployment, Total – time t; time t-1; time t-2
  - IMD5055_Human Development Index – time t; time t-1; time t-2
  - IMD5056_Inflation, GDP Deflator (Annual %) – time t; time t-1; time t-2
  - IMD5057_Population, Total – time t; time t-1; time t-2
  - IMD5058_1 Effective Number of Electoral Parties
  - IMD5059_1 Effective Number of Parliamentary Parties
IMD5052_1 (GDP Growth Annual – time t) data were not available at the time of the Standalone CSES Module Full Release for Finland (2011) and thus were classified as “Missing.” However, this data has since become available and is coded in CSES IMD.
IMD5053_1 (GDP per capita – time t) data were not available at the time of the Standalone CSES Module Full Release for Finland (2011) and thus were classified as “Missing.” However, this data has since become available and is coded in CSES IMD.
IMD5054_ (Unemployment – time t; time t-1; time t-2) data were not available at the time of the Standalone CSES Module Full Release for specific studies and thus were classified as “Missing.” However, this data has since become available and is coded in CSES IMD. These studies are Brazil (2010), Finland (2011), Latvia (2010), Netherlands (2010), Peru (2011), Philippines (2010), Portugal (2009), Slovakia (2010), and Turkey (2011).
IMD5055_ (Human Development Index) data were not available at the time of the Standalone CSES Module Full Release for specific studies and thus were classified as “Missing.” However, this data has since become available and is coded in CSES IMD. For IMD5055_1 (Human Development Index – time t) these studies are Peru (2016), Philippines (2016), and Slovakia (2016). For IMD5055_2 (Human Development Index – time t-1) these are Norway (2005), Peru (2011), Poland (2005), and Turkey (2011). For IMD5055_3 (Human Development Index – time t-2) these are Brazil (2006), Czech Republic (2006), Mexico (2006), Norway (2005), Poland (2005), and Sweden (2006).
IMD5056_1 (Inflation, GDP Deflator (Annual %) – time t) data were not available at the time of the Standalone CSES Module Full Release for Finland (2011) and thus were classified as “Missing.” However, this data has since become available and is coded in CSES IMD.
IMD5057_ (Population, total – time t; time t-1, time t-2) data were not available at the time of the Standalone CSES Module Full Release for specific studies and thus were classified as “Missing.” However, this data has since become available and is coded in CSES IMD. These studies are Austria (2013), Finland (2011), Slovakia (2010), and Peru (2016).
IMD5001_A-I (Percent Vote – Lower House – Party A – I), IMD5002_A-I (Percent Seats – Lower House – Party A – I), IMD5003_A-I (Percent Vote – Upper House – Party A – I), IMD5004_A-I (Percent Seats – Upper House – Party A – I) and IMD5005_A-I (Percent Vote – President – Party A – I): CSES Module 1 and CSES Module 2 include a category “ALLIANCE MEMBER (NO INDIVIDUAL TALLY)” (Coded 997 in Module 1 and 996 in Module 2). This category was recoded to “999. MISSING” in CSES IMD.
IMD5001_D (Percent Vote – Lower House – Party D) has been coded to “997. NOT APPLICABLE” because the study refers to the Russian 2000 Presidential election. In the CSES Module 1, this variable is coded “0.”
IMD5003_F (Percent Vote – Upper House – Party F) and IMD5004_F (Percent Seats – Upper House – Party F) have been coded to “996. NOT APPLICABLE – UNICAMERAL SYSTEM” for Hong Kong (1998), because Hong Kong is a unicameral system. In CSES Module 1, these variables are coded “0” for Hong Kong (1998).
IMD5003_F (Percent Vote – Upper House – Party F) has been coded missing (999) for Spain (2000) because there is no PARTY F for Spain (2000). This variable is coded “997. NOT APPLICABLE” in CSES Module 1.
IMD5004_F (Percent Seats – Upper House – Party A – I) has been coded to missing (999) for Spain (2000) because Upper house election was held, but there is no PARTY F for Spain (2000). In CSES Module 1 this variable is coded “997. not applicable.”
IMD5003_D, E and F (Percent Vote – Upper House – Parties D, E and F), IMD5004_D, E and F (Percent Seats – Upper House – Parties D, E and F) and IMD5005_D, E and F (Percent Vote – President – Party D-F) have been coded to missing (999) for United States (1996) because there were no PARTY D, E, and F in the U.S. 1996 dataset. In CSES Module 1, these variables are coded “997. NOT APPLICABLE.”
IMD5004_F-I (Percent Seats – Upper House – Parties F-I) have been coded to missing (999) for Switzerland (2003), because there were no PARTY F, G, H and I in the Swiss 2003 dataset. In CSES Module 2, these variables are coded “997. not applicable.”
IMD5005_F (Percent Vote – President – Party F) has been coded to “996. not applicable – no role of President” because Hong Kong does not have a President. In CSES Module 1, this variable is coded “0” for Hong Kong (1998).
IMD5005_D, E, and F (Percent Vote – President – Parties D, E, and F) have been coded to missing (999) because there are no PARTY D, E, and F in the Taiwan 1996 dataset. In CSES Module 1, these variables are coded “997. NOT APPLICABLE.”
IMD5008_1 (Party of the Prime Minister before) and IMD5008_1 (Party of the Prime Minister after) for Belarus (2008) have been coded to “9999989. INDEPENDENT CANDIDATE” because Sergei Sidorsky (independent) was Prime Minister before and after the 2008 Belarusian election. In CSES Module 3, these variables are coded missing (99) for Belarus (2008).
IMD5008_1 (Party of the Prime Minister before) and IMD5008_1 (Party of the Prime Minister after) for France (2007) have been coded to “2500001. Union for a Popular Movement (UMP)” because Dominique de Villepin (UMP) was Prime Minister before the 2007 French Presidential election and François Fillon (UMP) was Prime Minister after the 2007 French Presidential election. In CSES Module 3, these variables are coded “1”, the numerical party code for the Green Party (Les Verts) for France (2007).
IMD5008_1 (Party of the Prime Minister before) and IMD5008_2 (Party of the Prime Minister after) for Hong Kong (2008) study has been coded to “9999997. NOT APPLICABLE” because there is no such post as the Prime Minister in Hong Kong. In CSES Module 3, these variables are coded missing (99) for Hong Kong.
IMD5008_1 (Party of the Prime Minister before) for Peru (2011) has been coded to “6040026. Peruvian Aprista Party (PAP)” because Rosario Fernandez (PAP) was Prime Minister before the 2011 Peruvian general election. In CSES Module 3, this variable is coded not applicable (97) for Peru (2011).
IMD5008_1 (Party of the Prime Minister before) and IMD5008_2 (Party of the Prime Minister after) for the Philippines (2010) has been coded to “9999997. NOT APPLICABLE” because there is no Prime Minister in the Philippines, with the President being both the head of government and the head of state. In CSES Module 3, these variables are coded missing (99) for the Philippines (2010).
IMD5008_1 (Party of the Prime Minister before) and IMD5008_2 (Party of the Prime Minister after) for South Korea (2008) have been coded to “4100001. New Frontier Party (NFP)” (known as Grand National Party until 2012) because Han Seung-soo (Grand National Party) was Prime Minister before and after the 2008 South Korean election. In CSES Module 3, these variables are coded “97. not applicable” for South Korea (2008).
IMD5008_1 (Party of the Prime Minister before) for Poland (2007) has been coded to “6160002. Law and Justice (PiS)” as Jaroslaw Kaczynski (PiS) was Prime Minister before the Polish 2007 election. In CSES Module 3, this variable is coded “6”, the numerical party code for Civic Platform (PO) for Poland (2007).
IMD5008_1 (Party of the Prime Minister before) for Slovenia (2008) has been coded to “7050001. Slovenian Democratic Party (SDS)” because Janez Jansa (SDS) was Prime Minister before the 2008 Slovenian election. In CSES Module 3, this variable is coded “2”, the numerical party code for Liberal Democracy of Slovenia (LDS) for Slovenia (2008).
IMD5008_2 (Party of the Prime Minister after) for Peru (2011) has been coded to “6040050. Peru Wins (UPP)” because Salomón Lerner Ghitis, the former campaign manager of the 2011 Peru Wins alliance, became Prime Minister after the 2011 Peruvian general election. In CSES Module 3, this variable is coded “97. NOT applicable” for Peru (2011).
IMD5008_2 (Party of the Prime Minister after) for Poland (2007) has been coded to “6160001. Civic Platform” as Donald Tusk (PO) became Prime Minister after the Polish 2007 election. In CSES Module 3, this variable is coded “8”, the numerical party code for the Polish Peasant Party (PSL) for Poland (2007).
IMD5008_2 (Party of the Prime Minister after) for Slovenia (2008) has been coded to “7050002. Social Democrats (SD)” because Borut Pahor (SD) became Prime Minister after the 2008 Slovenian election. In CSES Module 3, this variable is coded “1”, the numerical party code for the Democratic Party of Pensioners (DeSUS) for Slovenia (2008).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) for Belarus (2008) has been coded to “1120101. Alexander Lukashenko (Independent)” because Lukashenko became President of Belarus in 1994. Even though he runs in elections under an independent banner, he was assigned a separate numerical code in the CSES IMD because of his special role in the Belarusian political system. In CSES Module 3, these variables are coded missing (99) for Belarus (2008).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) for Croatia (2007) have been coded to “1910003. Croatian People’s Party – Liberal Democrats (HNS)” because Stjepan Mesic, endorsed by the HNS, was President before and after the 2007 Croatian election. In CSES Module 3, these variables are coded “97. not applicable” for Croatia (2007).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) for Czech Republic (2006 and 2010) have been coded to “2030002. Civic Democratic Party (ODS)” because for both studies, Vaclav Klaus (ODS) was President before and after the respective elections. In CSES Module 3, these variables are coded “97. not applicable” for Czech Republic (2006) and (2010).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) because for Estonia (2011) has been coded to “2330003. Social Democratic Party (SDE)” because Toomas Hendrik Ilves was President before and after the Estonian 2011 election. In CSES Module 3, these variables are coded “97. not applicable” for Estonia (2011).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) for France (2007) have been coded to “2500001. Union for a Popular Movement (UMP)” because Jacques Chirac (UMP) was President before the French 2007 Presidential election and Nicolas Sarkozy (UMP) was President after the French 2007 Presidential election. In CSES Module 3, these variables are coded “1”, the numerical party code for the Green Party (Les Verts) for France (2007).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) for Germany (2005) has been coded to “2760002. Christian Democratic Union (CDU)” because Horst Koehler (CDU) was President before and after the German 2005 election. In CSES Module 3, these variables are coded “97. not applicable” for Germany (2005).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) for Greece (2009) has been coded to “3000001. Pan-Hellenic Socialist Movement (PASOK)” because Karolos Papoulias (PASOK) was President before and after the 2009 Greek election. In CSES Module 3, these variables are coded “97. not applicable” for Greece (2009).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) for Hong Kong (2008) has been coded to “9999997. NOT APPLICABLE” because there is no such post as the President in Hong Kong. In CSES Module 3, these variables are coded “99. MISSING” for Hong Kong (2008).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after). That is for Iceland (2007, 2009 and 2013) have been coded to “9999989. INDEPENDENT CANDIDATE” because Olafur Ragnar Grimsson (independent) was President of Iceland from 1996 to 2016. For the respective Icelandic studies, these variables are coded “97. not applicable” in the Standalone CSES Module datasets.
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) for Israel (2006) has been coded to “3760001. Likud – The Consolidation (L)” because Moshe Katsav (Likud) was President before and after the 2006 Israeli election. In CSES Module 3, these variables are coded “97. not applicable” for Israel (2006).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) for Latvia (2010) study has been coded to “9999989. INDEPENDENT CANDIDATE” because Valdis Zatlers, independent before founding the Reform Party in 2011, was President of Latvia from 2007 to 2011. In CSES Module 3, these variables are coded “99. MISSING” for Latvia (2010).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) for Portugal (2009) has been coded to “6200001. Social Democratic Party (PSD)” because Anibal Cavaco Silva (PSD) was President before and after the 2009 Portuguese election. In CSES Module 3, these variables are coded “99. MISSING” for Portugal (2009).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) for Slovenia (2008) has been coded to “9999989. INDEPENDENT CANDIDATE” because Danilo Türk (independent) was President before and after the 2008 Slovenian election. In CSES Module 3, these variables are coded “97. not applicable” for Slovenia (2008).
IMD5009_1 (Party of the President before) and IMD5009_2 (Party of the President after) for Turkey (2011) has been recoded to “7920001. Justice and Development Party (AKP)” because Abdullah Gul (AKP) was President before and after the 2011 Turkish election. In CSES Module 3, these variables are coded “97. not applicable” for Turkey (2011).
IMD5011_F (Ideological Party Families – Party F) and IMD5011_G (Ideological Party Families – Party G) for Finland (2007), referring to Swedish People’s Party in Finland – RKP (party F) and Christian Democrats – KD (party G) have been recoded. In CSES IMD party F is coded to “12. ETHNIC PARTIES”, while party G is coded to “08. CHRISTIAN DEMOCRATIC PARTIES.” This is in accordance with the Finish 2007 Macro Report. In the CSES Module 3, party F is coded to “8. Christian Democratic Parties” while party G is coded to “12. Ethnic Parties” for Finland (2007).
IMD5016_1 (Votes cast, Lower House – 1^st segment (tier)) and IMD5016_3 (Votes cast, Upper House – 1^st segment (tier)) for Australia (1996 and 2004) have been coded to “91. OTHER (SEE ELECTION STUDY NOTES)” because Australia uses the Alternative Vote system (for Lower House) and single-transferable vote form of proportional representation (for Upper House). Hence, voters vote for as many candidates as there are on the ballot. Thus, the total number of votes varies across electoral districts. These studies are coded “1” for these variables in CSES Standalone Module datasets.
IMD5016_2 (Votes cast, Lower House – 2^nd segment (tier)) for Mexico (1997 and 2000) have been coded “97. NOT APPLICABLE” because voters in Mexico cast a single vote, which is further used to determine both state and national seat allocations. These studies are coded “1” for the variable “Votes cast Lower house – 2^nd segment (tier)” in CSES Standalone Module datasets.
IMD5016_2 (Votes cast, Lower House – 2^nd segment (tier)) for Norway (1997, 2001, 2005 and 2009) have been coded “91. OTHER (SEE ELECTION STUDY NOTES)” because Norwegian Parliament comprises 169 seats in two tiers. These studies are coded “97. NOT APPLICABLE” in CSES Standalone Module datasets.
IMD5016_2 (Votes cast, Lower House – 2^nd segment (tier)) for Slovenia (1996 and 2004) have been coded “97. NOT APPLICABLE” because voters in Slovenia cast a single vote in the open-list proportional system (one-tier system). These studies are coded “1” for the variable “Votes cast Lower house – 2^nd segment (tier)” in CSES Standalone Module datasets.
IMD5016_2 (Votes cast, Lower House – 2^nd segment (tier)), IMD5016_3 (Votes cast, Upper House – 1^st segment (tier)), and IMD5016_4 (Votes cast, Upper House – 2^nd segment (tier)) for Taiwan (1996) have been coded “97. NOT APPLICABLE” because this study refers to 1996 Presidential election, where voters cast single vote. The coding in CSES Module 1 dataset refers to the National assembly in Taiwan.
IMD5016_3 (Votes cast, Upper House – 1^st segment (tier)) for Czech Republic (1996 and 2002) have been coded “97. NOT APPLICABLE” because these studies do not refer to Upper House elections. These studies are coded “1” in CSES Standalone Module datasets.
IMD5016_3 (Votes cast, Upper House – 1^st segment (tier)) for France (2002) has been coded “97. NOT APPLICABLE” because members of the Senate (French Upper House) are indirectly elected. This study is coded “1” in CSES Module 2 dataset.
IMD5016_3 (Votes cast, Upper House – 1^st segment (tier)) for Germany (1998) has been coded “97. NOT APPLICABLE” because members of Bundesrat (German Upper House) are indirectly elected. This study is coded “3” in CSES Module 1 dataset.
IMD5016_3 (Votes cast, Upper House – 1^st segment (tier)) for Netherlands (1998 and 2002) have been coded “97. NOT APPLICABLE” because members of Senate (Dutch Upper House) are indirectly elected. Netherlands (1998) is coded “3” and Netherlands (2002) is coded “91” CSES Standalone Module datasets.
IMD5016_3 (Votes cast, Upper House – 1^st segment (tier)) for Philippines (2004) has been coded “12” because, in the Senate election, voters have one vote per candidate to be elected. This is in total 12 votes because half of the Senate is renewed at the single election. This study is coded “1” in CSES Module 2 dataset.
IMD5016_3 (Votes cast, Upper House – 1^st segment (tier)) for Poland (1997, 2005, 2007, and 2011) have been coded “91. OTHER (SEE ELECTION STUDY NOTES)” because voters in Senate election have as many votes as there are candidates to be elected in a district (between 1 and 4). These studies are coded “1” in CSES Standalone Module datasets.
IMD5016_3 (Votes cast, Upper House – 1^st segment (tier)) for Spain (1996, 2000, and 2004) have been coded “3” because each voter has three votes in 52 multi-member constituencies in Senate election in Spain. These studies are coded “1” in CSES Standalone Module datasets.
IMD5016_4 (Votes cast, Upper House – 2^nd segment (tier)) Belgium (1999 (Flanders and Wallonia) and 2003) have been coded “97. NOT APPLICABLE” because Senate elections at the time were held in a one-tier system. Since 2014, Senators in Belgium are not directly elected. Belgium (1999) study for Flanders is coded “99”, Belgium (1999) study for Wallonia is coded “9” and Belgium (2003) is coded “91” in CSES Standalone Module datasets.
IMD5017_1 (Voting procedure, Lower House – 1^st segment (tier)) for Brazil (2002) has been coded “4. OTHER (SEE ELECTION STUDY NOTES)” because in Brazil each political party presents a list of candidates, and voters can cast a vote either for a candidate, or for a party. This study is coded “2. PARTY LISTS” in CSES Module 2 dataset.
IMD5017_1 (Voting procedure, Lower House – 1^st segment (tier)) and IMD5017_3 (Voting procedure, Upper House – 1^st segment (tier)) for Chile (1999) has been coded “2. PARTY LISTS” because coalitions in Chile present a list of two candidates, normally from different parties, and voters cast a vote for one candidate from the list. This study is coded “1. CANDIDATES” in CSES Module 1 dataset.
IMD5017_1 (Voting procedure, Lower House – 1^st segment (tier)) for Denmark (1998) has been coded “4. OTHER (SEE ELECTION STUDY NOTES)” because voters choose to endorse a party, a candidate on a party list, or an independent candidate. This study is coded “2. PARTY LISTS” in CSES Module 1 dataset.
IMD5017_1 (Voting procedure, Lower House – 1^st segment (tier)) for Netherlands (1998 and 2002) have been coded “4. OTHER (SEE ELECTION STUDY NOTES)” because voters in the Netherlands cast one vote for the list and a candidate at the same time (matrix). These studies are coded “2. PARTY LISTS” in CSES Standalone Module datasets.
IMD5017_1 (Voting procedure, Lower House – 1^st segment (tier)) for Peru (2011) has been coded “2. PARTY LISTS” because voters cast one vote for a party list, but they can also express a preferential vote for some specific candidate optionally, though not many voters do that. This study is coded “1. CANDIDATES” in CSES Module 3 dataset.
IMD5017_1 (Voting procedure, Lower House – 1^st segment (tier)) for Switzerland (1999 and 2003) have been coded “4. OTHER (SEE ELECTION STUDY NOTES)” because Swiss voters can vote for a party list; endorse specific candidate from a list; add candidates from other parties instead of a candidate of the list; delete candidates or vote twice for the same candidate. These studies are coded “2. PARTY LISTS” in CSES Standalone Module datasets.
IMD5017_2 (Voting procedure, Lower House – 2^nd segment (tier)) for Mexico (2003) has been coded “2. PARTY LISTS” because, in Mexico, each voters’ vote is counted twice; once for the single-member district contest, and a second time for the regional PR contest. This study is coded “7. NOT APPLICABLE” in CSES Module 2 dataset.
IMD5017_2 (Voting procedure, Lower House – 2^nd segment (tier)) for Norway (1997, 2001, 2005 and 2009) have been coded “2. PARTY LISTS” because voters in Norway cast one single vote based on party lists. These studies are coded “7. NOT APPLICABLE” in CSES Standalone Module datasets.
IMD5017_2 (Voting procedure, Lower House – 2^nd segment (tier)) for Slovenia (1996 and 2004) have been coded “7. NOT APPLICABLE” because Slovenia is a one-tier system. These studies are coded “1. CANDIDATES” in CSES Standalone Module datasets because members of Italian and Hungarian indigenous ethnic communities are entitled to 2 deputy seats, one for each community, and were considered as the second tier.
IMD5017_2 (Voting procedure, Lower House – 2^nd segment (tier)) for Taiwan (1996) has been coded “7. NOT APPLICABLE” because before 2008 reform, Taiwan was a one-tier system. This study is coded “2. PARTY LISTS” in CSES Module 1 dataset.
IMD5017_3 (Voting procedure, Upper House – 1^st segment (tier)) for Czech Republic (1996 and 2002) have been coded “7. NOT APPLICABLE” because they do not refer to Upper house election. These studies are coded “1. CANDIDATES” in CSES Standalone Module datasets.
IMD5017_3 (Voting procedure, Upper House – 1^st segment (tier)) for France (2002) has been coded “7. NOT APPLICABLE” because Senators in France are indirectly elected. This study is coded “4. OTHER (SEE ELECTION STUDY NOTES)” in CSES Module 2 dataset.
IMD5017_3 (Voting procedure, Upper House – 1^st segment (tier)) for Germany (1998) has been coded “7. NOT APPLICABLE” because members of Bundesrat (Upper house) in Germany are indirectly elected. This study is coded “4. OTHER (SEE ELECTION STUDY NOTES)” in CSES Module 1 dataset.
IMD5017_3 (Voting procedure, Upper House – 1^st segment (tier)) and IMD5017_4 (Voting procedure, Upper House – 2^nd segment (tier)) for Mexico (1997, 2000 and 2003) have been coded “1. CANDIDATES” because in Mexico each vote in Upper house election is counted twice: once for the 3-seat multi-member districts contest, and a second time for the national PR contest. These studies are coded as “2. PARTY LISTS” for the first segment (tier) and “7. NOT APPLICABLE” for second segment (tier) of upper house election in CSES Standalone Module datasets.
IMD5017_3 (Voting procedure, Upper House – 1^st segment (tier)) for Netherlands (1998 and 2002) have been coded “7. NOT APPLICABLE” because members of the Upper house in Netherlands are indirectly elected. These studies are coded “4. OTHER (SEE ELECTION STUDY NOTES)” in CSES Standalone Module datasets.
IMD5017_3 (Voting procedure, Upper House – 1^st segment (tier)) for Romania (2009 and 2014) have been coded “7. NOT APPLICABLE” because these studies do not concern the Upper house election. These studies are coded “1. CANDIDATES” in CSES Standalone Module datasets.
IMD5017_3 (Voting procedure, Upper House – 1^st segment (tier)) for Romania (2012) has been coded “4. OTHER (SEE ELECTION STUDY NOTES),” which differs from Romania (1996 and 2004) studies. This is because the law changed, and from 2008 to 2016 Romania used mixed member proportional system. This study is coded “1. CANDIDATES” in CSES Module 4 dataset.
IMD5017_3 (Voting procedure, Upper House – 1^st segment (tier)) for Spain (1996, 2000 and 2008) have been coded to “4. OTHER (SEE ELECTION STUDY NOTES)” because for Upper house election in Spain parties present list of three candidates, voters mark up to three candidates, from any party. In total, 4 candidates are elected who won the most votes in district. These studies are coded “1. CANDIDATES” in CSES Standalone Module datasets.
IMD5018_1 (Voting rounds, Lower House – 1^st segment (tier)) for Belarus (2008) has been coded “1” because all 110 deputies were elected in the first round of the election. This study is coded “2” in CSES Module 3 dataset.
IMD5018_1 (Voting rounds, Lower House – 1^st segment (tier)) and IMD5018_2 (Voting rounds, Lower House – 2^nd segment (tier)) for Japan (2007 and 2013) have been coded to “97. NOT APPLICABLE” because the studies do not concern the lower house election. These studies are coded “1” in CSES Standalone Module datasets.
IMD5018_2 (Voting rounds, Lower House – 2^nd segment (tier)) for Hong Kong (2008) has been coded “1” because, in geographical constituency elections, there is only one round of voting. This study is coded “2” in CSES Module 3 dataset.
IMD5018_2 (Voting rounds, Lower House – 2^nd segment (tier)) for Norway (2005 and 2009) have been coded “1” because voters cast only single ballot, even though their vote directly impacts the selection of 150 members from 19 multi-member districts using proportional representation, but also influences the dispersion of the 19 “member at large seats.” These studies are coded “97. NOT APPLICABLE” in CSES Module 3 dataset.
IMD5018_3 (Voting rounds, Upper House – 1^st segment (tier)) for Czech Republic (2006 and 2010) have been coded “2” because the study covers upper house elections which were held in two rounds. These studies are coded “97. NOT APPLICABLE” in CSES Module 3 datasets.
IMD5021_2 (Is there party threshold, Lower House – 2^nd segment (tier)) and IMD5021_4 (Is there party threshold, Upper House – 2^nd segment (tier)) for Romania (2009) has been coded “7. NOT APPLICABLE” because the study concerns the Presidential election. This study is coded “1” in CSES Module 3 dataset and that coding refers to 2008 Parliamentary election in Romania.
IMD5021_4 (Is there party threshold, Upper House – 2^nd segment (tier)) for Czech Republic (2013) has been coded to “7. NOT APPLICABLE” because the study does not refer to upper house election. This study is coded “5” in CSES Module 4 datasets.
IMD5022_1 (Party threshold, Lower House – 1^st segment (tier)) for South Africa (2009 and 2014) have been coded “0” because South Africa is one national district with a magnitude of four hundred, and there is no legal threshold for gaining representation. These studies are coded “5” for 2009, and “0.25” for 2014 in CSES Standalone Module datasets.
IMD5022_2 (Party threshold, Lower House – 2^nd segment (tier)) and IMD5022_3 (Party threshold, Upper House – 1^st segment (tier)) for Canada (2008) have been coded to “97. NOT APPLICABLE” because Canada is using the “first-past-the-post” system, which is a one-tier system, and the 2008 study does not concern Upper house election. This study is coded “7” in CSES Module 3 dataset.
IMD5022_2 (Party threshold, Lower House – 2^nd segment (tier)) for New Zealand (2008) has been coded “96. OTHER THRESHOLD (SEE ELECTION STUDY NOTES)” because New Zealand has an alternative threshold, parties can obtain seats by winning 5% of total votes nationally, or by winning one of the 70 constituency seats. This study is coded “5” in CSES Module 3 dataset.
IMD5022_2 (Party threshold, Lower House – 2^nd segment (tier)) for South Korea (2008) has been coded “96. OTHER THRESHOLD (SEE ELECTION STUDY NOTES)” because South Korea has an alternative threshold, parties can obtain seats by winning 3% of total votes nationally, or by winning at least 5 primary district seats. This study is coded “3” in CSES Module 3 dataset.
IMD5022_2 (Party threshold, Lower House – 2^nd segment (tier)) and IMD5022_4 (Party threshold, Upper House – 2^nd segment (tier)) for Romania (2009) has been coded “97. NOT APPLICABLE” because the study concerns Presidential election. The study is coded “96. OTHER THRESHOLD (SEE ELECTION STUDY NOTES)” for 2008 Parliamentary election in CSES Module 3 dataset.
IMD5022_2 (Party threshold, Lower House – 2^nd segment (tier)), IMD5022_3 (Party threshold, Upper House – 1^st segment (tier)) and IMD5022_4 (Party threshold, Upper House – 2^nd segment (tier)) for South Africa (2009) has been coded “97. NOT APPLICABLE” because the study does not concern the Upper house election, and South Africa is a one-tier system. The study is coded “7” in CSES Module 3 dataset.
IMD5022_2 (Party threshold, Lower House – 2^nd segment (tier)), IMD5022_3 (Party threshold, Upper House – 1^st segment (tier)), and IMD5022_4 (Party threshold, Upper House – 2^nd segment (tier)) for Latvia (2010) has been coded “97. NOT APPLICABLE” because Latvia is a single-tier system for Lower house election and 2010 study does not concern the Upper house election. The study is coded “7” in CSES Module 3 dataset.
IMD5022_3 (Party threshold, Upper House – 1^st segment (tier)) for Czech Republic (2013) has been coded “97. NOT APPLICABLE” because the study does not concern the Upper house election. This study is coded “0” in CSES Module 4 dataset.
IMD5022_3 (Party threshold, Upper House – 1^st segment (tier)) for Ireland (2007) has been coded “97. NOT APPLICABLE” because members of the Upper house in Ireland are indirectly elected. This study is coded “0” in CSES Module 3 dataset.
IMD5022_3 (Party threshold, Upper House – 1^st segment (tier)) and IMD5022_4 (Party threshold, Upper House – 2^nd segment (tier)) for Peru (2011) has been coded “97. NOT APPLICABLE” because Peru has a unicameral parliament. This study is coded “7” in CSES Module 3 dataset.
IMD5029_A (Government Composition Before – Party A) and IMD5029_H (Government Composition Before – Party H) for Portugal (2015) have been coded “8” (for party A – Social Democratic Party) and “4” (for party H – People’s Party). These two parties contested the 2015 election as an electoral coalition, but they are assigned separate alphabetical codes. Independents held the remaining 3 seats. This study is coded as 12 portfolios for party A, 4 portfolios for party E, and 4 portfolios for party G in CSES Module 4 dataset.
IMD5031_D (Government Composition After – Party D) for Canada (2011 and 2015) have been coded “0” because that party did not hold any ministry after the 2011 or 2015 election. This variable is coded “6” in CSES Module 4 dataset.
IMD5031_A (Government Composition After – Party A) for Great Britain (2015) has been coded “22” because 22 is the size of the cabinet, and the Conservative party (party A) held all ministries after the 2015 election. This variable is coded “23” in CSES Module 4 dataset.
IMD5049 (Age of the Current Regime) in CSES IMD reflects the latest version of PolityIV. Using this latest version, we discovered the following deviations between Standalone CSES Modules and CSES IMD, namely:

- - For Czech Republic (2010) CSES IMD codes this variable “17,” reflecting the last regime change in the Czech Republic noted by PolityIV in 1993. In CSES Module 3, this variable is coded “16” for the Czech Republic (2010).
  - For Denmark (2007) CSES IMD codes this variable “62,” reflecting the last regime change in Denmark noted by PolityIV in 1945. In CSES Module 3, this variable is coded “61” for Denmark (2007);
  - For New Zealand (2011) CSES IMD codes this variable “134,” reflecting the last regime change in New Zealand noted by PolityIV in 1877. In CSES Module 3, this variable is coded “135” for New Zealand (2011);
  - For Netherlands (2006) CSES IMD codes this variable “61,” reflecting the last regime change in the Netherlands noted by PolityIV in 1945. In CSES Module 3, this variable is coded “60” for the Netherlands (2006);
  - For Netherlands (2010) CSES IMD codes this variable to “65,” reflecting the last regime change in the Netherlands noted by PolityIV in 1945. In CSES Module 3, this variable coded “64” for the Netherlands (2010).