CSES Module 4: 2011-2016 Errata

Updates and error corrections.

CURRENT ERRATA

Errata listed in this section apply to the current version of CSES Module 4, and users should consider whether to apply the items to their files.

CSES Module 4 errata discovered with Full Release of May 29, 2018, in the development of CSES IMD Phase 2

CSES Module 4: Dataset & Codebook Errata

Posted: October 17, 2019

 

Errata discovered with Full Release of May 29, 2018,  in the development of CSES IMD Phase 2

 

With the release of the second phase of CSES Integrated Module Dataset (IMD), the CSES Secretariat identified some issues with CSES Module 4 Full Release of December 15, 2015. Below, we provide a list of these issues that users of CSES Module 4 (release date: December 15, 2015) in the future may wish to take note of concerning analyses.

 

  • VARIABLE D3012_A (LIKE-DISLIKE – LEADER A): For Israel (2013), the values coded in D3012_A are incorrect and will be corrected in a future re-release. In the last Full Release of CSES Module 4 dated May 29, 2018, this error remains. However, the CSES IMD Release of October 17, 2019, these respondents were recoded to “99. MISSING” in CSES IMD.

 

  • VARIABLE D5009_A (GOVERNMENT COMPOSITION BEFORE ELECTION – PARTY A): For Canada (2015), portfolios are coded under D5009_A, indicating that PARTY A (Liberal Party) was the party of the incumbent government. However, the party of the incumbent government was the Conservative Party (PARTY B). In the last Full Release of CSES Module 4 dated May 29, 2018, this error remains. However, in the CSES IMD Release of October 17, 2019, this variable is correctly coded for Canada (2015).

 

  • VARIABLES D5009_A, D5009_E and D5009_H (NUMBER OF PORTFOLIOS AFTER – PARTY A, PARTY E, and PARTY H) for Portugal (2015) are coded as 12 portfolios for party A (Social Democratic Party), 4 portfolios for party E (Party for People, Animals, and Nature), and 4 portfolios for party G (Workers Communist Party. Two parties that contested election as electoral coalition are Social Democratic Party (party A) and People’s Party (party H). They are assigned separate alphabetical codes. Party A should have 8 portfolios, and party H 4. The remaining 3 seats were held by independents. Parties E and G did not have portfolios before Portuguese (2015) election. In the last Full Release of CSES Module 4 dated May 29, 2018, this error remains. However, in the CSES IMD Release of October 17, 2019, these variables are correctly coded for Portugal (2015).

 

  • VARIABLE D5013_D (NUMBER OF PORTFOLIOS AFTER – PARTY D) is coded “6” for Canada (2011 and 2015). Party D (Bloc Quebecois for both studies)did not hold any ministry after the 2011 or 2015 election. In the last Full Release of CSES Module 4 dated May 29, 2018, this error remains. However, in the CSES IMD Release of October 17, 2019, these variables are correctly coded to “0” for Canada (2011 and 2015).

 

  • VARIABLE D5013_A (NUMBER OF PORTFOLIOS AFTER – PARTY A) is coded “23” for Great Britain (2015). Conservative party (party A) held all ministries after the 2015 election – in total 22 ministries. In the last Full Release of CSES Module 4 dated May 29, 2018, this error remains. However, in the CSES IMD Release of October 17, 2019, this variable is correctly coded to “22” for Great Britain (2015).

 

  • VARIABLE D5038_3 (VOTES CAST – UPPER – 1ST SEGMENT) is coded “1” for Poland (2011). In Upper House Election in Poland, voters have as many votes as there are candidates to be elected in a district (between 1 and 4). In the last Full Release of CSES Module 4 dated May 29, 2018, this error remains. However, in the CSES IMD Release of October 17, 2019, this variable is coded “91. OTHER (SEE ELECTION STUDY NOTES)” for Poland (2011).

 

  • VARIABLE D5046_1 (PARTY THRESHOLD – LOWER – 1ST SEGMENT) is coded “5” for South Africa (2014). However, South Africa is one national district with a magnitude of four hundred, and there is no legal threshold for gaining representation. In the last Full Release of CSES Module 4 dated May 29, 2018, this error remains. However, in the CSES IMD Release of October 17, 2019, this variable is coded “0” for South Africa (2014).

 

  • CODEBOOK PART 3 (PARTIES AND LEADERS): For Greece (2015), the numerical party code “7” refers to the party “The Panhellenic Socialist Movement – PASOK.” The alphabetical party code PARTY G also refers to PASOK and not Democratic Alignment as written in the Parties and Leaders table. In the last Full Release of CSES Module 4 dated May 29, 2018, this error remains. However, the CSES IMD Release of October 17, 2019, this error has been corrected and incorporated into the dataset.

 

CSES Module 4 Full Release: May 29, 2018 Version

CSES Module 4 Full Release: Data Set Errata
Posted: May 29, 2018

Updates made to the Full Release of May 29, 2018, since the Advance Release of April 11, 2017

The Full Release of CSES Module 4 of May 29, 2018, contains additions, updates, and improvements to the Fourth Advance Release of April 11, 2017. Please do not use files from the previous release with files from the current release. Variable names and attributes, as well as the data, have sometimes been changed.

Seven new election studies have been added to the file, bringing the number of election studies included in Module 4 to 45 in total. The studies added to the Full Release Dataset from the Fourth Advance Release are in alphabetical order: Argentina (2015), Canada (2015), Greece (2015), Latvia (2011), Latvia (2014), Peru (2016), and Romania (2014).

Below, we provide a list of some of the primary changes and corrections to previously released Module 4 data. Users are advised that this is list is not comprehensive but instead highlights the main changes both to data and documentation since the previous release.

Data changes:

    1. To allow for data bridging with other macro sources, new variables have been added to the dataset. These variables are:
      … Variables D5200_A – D5200_I which contain party identifiers for the Manifesto Research on Political Representation (MARPOR / previously CMP – https://manifesto-project.wzb.eu/) project, for CSES parties A to I.
      … Variables D5201_A – D5201_I which contain party identifiers from Parliaments and Governments Database (ParlGov – http://www.parlgov.org/) for CSES parties A to I.
    2. District data for France (2012), Romania (2012) and Thailand (2011) have been added to the dataset.
    3. District data Mexico (2012) have been changed. In the previous version of the dataset states were used as primary electoral districts due to lack of data on the actual primary electoral districts. This has now been corrected and changes are implemented as of Module 4 Full Release.
    4. Variables D3025_2_A_PT, D3025_4_A_PT, D3025_2_B_PT and D3025_4_B_PT (dichotomized and original Political information 2nd and 4th item) have been added to the dataset. The data on these variables are available for Canada (2011) and Ireland (2011). These countries ran a pre-test version of the CSES Module 4 study and there was some discrepancy in the political information items collected by these studies as a result. We provide these data in separate variables to allow analysts to make a choice about the inclusion of these data in their analyses.
    5. Variable D1010_1 (Sample Weight) for Canada (2011) have been added to the dataset.
    6. Variable D2011 (Occupation) for Ireland (2011) have been added to the dataset.
    7. Variable D2012 (Socio-economic Status) for Slovakia (2016) have been added to the dataset.
    8. Variable D3009 (Who is in power can make difference) for Canada (2011) have been added to the dataset.
    9. Variables D3022_ (Mobilization: Personal Contact) and D3023_ (Mobilization: Personal Contact – Means of Contact) for Canada (2011) have been added to the dataset.
    10. Variable D4002 (Number of Candidates in District) for Sweden (2012) have been added to the dataset.
    11. Variable D4004_4 (district election results for PARTY D (Sinn Fein)) for Ireland (2011) have been added to the dataset.
    12. Variables D5091_1 (Health Expenditure – % GDP) and D5093_1 (Central Government Debt, Total – % GDP) for New Zealand (2014) have been added to the dataset.
    13. The variable D1005 (ID Variable – Respondent) contained an incorrect digit indicating two sample components for New Zealand (2011) when in fact there was one. The variable is recoded for this study and the fourth digit is now “zero.”
    14. Demographic weight (variable D1010_2) for the Mexico (2015) study has been updated. The new variable corrects for the fact that the previous version of the weight variable was constructed for a sample of 2400 respondents (all participants in the study), but the CSES questionnaire was administered on a sample of 1200 respondents.
    15. Variable D2010 (Current employment status) for Finland (2015) study was recoded and differs slightly from the variable in a fourth advance release. The Finish study included more codes than the CSES options, and these are now recoded to be more in line with CSES standard for this variable.
    16. In France (2012) study some respondents answered “zero” on the question about a number of persons in the household (D2021). These data were previously coded to missing.
    17. Variable D2026 (Religious Denomination) for Canada (2011) is recoded. Respondents who answered “none, don’t have one / Atheist” in Canadian study and were previously coded as CSES category “8200. Atheist” are now recoded to “8300. None.”
    18. Vote choice variables (D3006_LH_DC and D3008_LH_DC), as well as, district data variables (D4004_E – D4004_I) for Ireland (2011) are recoded such that they reflect the electoral system of Ireland correctly (district candidate system), and they differ from the ones published as of fourth advance release (where Ireland was coded as party list system).
    19. Vote choice variable (D3006_) for Mexico (2012) has been recoded to D3006_LH_DC (Vote Choice – District Candidate). It was previously coded as D3006_LH_PL (Vote Choice – Party List).
    20. New categories, “996. Not applicable: Unicameral system” and “997. Not applicable: No upper house election”, are introduced for variables D3008_UH_PL, D3008_UH_DC_1 and D3008_UH_DC_2. For these variables missing code is changed from “99” to “999.”
    21. For variable D3006_LH_PF (Current Election – Lower House – Preference Vote), Greece (2012), Poland (2011), Finland (2015), and Iceland (2013) are now coded to missing.
    22. For variable D3006_UH_PF (Current Election Upper House – Preference Vote), Kenya (2013) is now coded to “997. Not applicable: No Upper House election.”
    23. For the variable D3006_UH_PF (Current Election Upper House – Preference Vote), the missing code is changed from 9 to 999. Further, codes 996 (“Not applicable: unicameral system”) and 997 (“Not applicable: No upper house election”) are introduced, and studies are coded to 996 or 997 accordingly.
    24. Variables D3011_ (Like/Dislike Parties) and D3012_ (Like/Dislike Party Leaders) are recoded for Canada (2011). These questions measured likeability on a 0 to 100 scale in Canadian Election Study whereas CSES measures likeability on a 0-10 scale. In the initial recoding, some respondents were mistakenly coded as “don’t know any of the parties,” “don’t know,” “refused to answer,” and “missing.”
    25. For variable D3018_3 (Which party do you feel close to) for Canada (2011), 5 respondents that were previously coded as “91. None of the parties/candidates” are now coded to “90. Other party (not further specified).” This follows CSES standard for the given variable.
    26. Variable D3019 (Mobilization: Institutional Contact) is recoded for South Africa (2014). In the previous version, respondents who were not contacted by any party were mistakenly coded as if they had been contacted by at least one party.
    27. Variables D3020_2 (Mobilization: Institutional Contact – By Mail) and D3020_6 (Mobilization: Institutional Contact – Social Network or Web) are now coded as missing for Canada (2011) because the two means of contact were asked as a single question in Canadian Election Study and could not be distinguished for CSES standard variables.
    28. The variable D3021_1 (Mobilization: Institutional Contact) was recoded for Philippines (2016), so that previous code 94 is now 68, 95 is now 72, and 96 is now 64. The previous version of the data for Philippines (2016) contained codes “94”, “95” and “96” which are invalid CSES codes for this variable.
    29. For Sweden (2014) code “10” is recoded to “90”: for the following variables: D3021_3, D3021_4, D3021_5, D3021_6, D3021_7, D3021_9, and D3021_10. The code referred to “other party” category and is now coded to the standard CSES category “90. Other party / candidate (not further specified).”
    30. The variable D3027_2 (Ownership: Business or Property or Farm or Livestock) is recoded for Ireland (2011). The initial distribution for this variable was inversed and incorrect.
    31. The variable D4003 (Number of party lists in district) is recoded and corrected for Sweden (2014) and Finland (2015).
    32. “Not applicable” codes for the district data variables (variables D4001-D4005 inclusive) have been adjusted to ensure that not applicable is coded as consistently as possible throughout the dataset. The following changes have been made: … D4001:
      • “000. NOT APPLICABLE” is now “997. NOT APPLICABLE”
      • … D4001_N: “000. NOT APPLICABLE” is now “997. NOT APPLICABLE”
      • … D4002: “0000. NOT APPLICABLE” is now “9997. NOT APPLICABLE”
      • … D4002_N: “0000. NOT APPLICABLE” is now “9997. NOT APPLICABLE”
      • … D4003: “000. NOT APPLICABLE” is now “997. NOT APPLICABLE”
      • … D4003_N: “000. NOT APPLICABLE” is now “997. NOT APPLICABLE”
    33. New categories, “996. Not applicable: Unicameral system” and “997. Not applicable: No upper house election”, are introduced for variables D5003_ (Percent vote – Upper House) and D5004_ (Percent seats – Upper House).
    34. New categories, “996. Not applicable: No role of President” and “997. Not applicable: No Presidential election”, are introduced for variables D5005_ (Percent vote – President).
    35. Variable D5100 (Number of Legislative Chambers) is recoded from 1 to 2 for France (2012) and Great Britain (2015) to correctly reflect the bicameral systems of both countries.
    36. Variable D5015 (Number of parties participating in election) is recoded to 15 for Switzerland (2011), which was the number of party lists competing in 2011 elections. The previous version of the variable contained information that 17 parties were participating in the elections.
    37. Variable D5042_1 (Transferable Votes – Lower – 1st Segment) is recoded to “7. Non applicable” for Canada (2011). It was previously incorrectly coded as “5. No”.
    38. Variable D5054 (Regime: Type of Executive) is recoded to “3. Presidential Democracy” for the USA (2012) to reflect the nature of US political system correctly. Previously, it was incorrectly coded.
    39. Variables D5016_G, D5016_H and D5016_I are now coded for the United States (2012). These were previously coded as D5016_C, D5016_D and D5016_E. The CSES standard is to use alphabetical codes G, H and I for parties that have not reached the threshold of 1% at the national level.
    40. Variable D5006_2 (ELECTORAL TURNOUT – TURNOUT AS A PERCENTAGE OF THE VOTING AGE POPULATION (VAP)) has been updated for several countries to reflect clarification in measurement aim of the variable.

 

Codebook changes:

    1. The CSES Module 4 Codebook now consists of six separate sections (previously five). Part 6 is new and contains an Overview of Study Design and Weights. It provides users with an overview of the design of each study included in CSES as well as the polity weights provided for each Election Study. Previously, this information had been detailed in Part 1 of the CSES Codebook.
    2. Detailed information about how CSES codes parties/coalitions and leaders is now available in the Codebook. This information can be found in Parts 1 and 2 of the CSES Codebook by searching for “CSES MODULE 4 CODING OF PARTIES/COALITIONS & LEADERS.”
    3. Overview of CSES Module 4 Data File Particulars has been added to Part 1 of the Codebook. This includes information about type of data, the geographical coverage of the study and the total variable count.
    4. All six parts of the Codebook now contain instructions for users on how to navigate the CSES Module 4 Codebook electronically. As the Codebook is very extensive, this information has been added to help users quickly find the information they are looking for. The information can be found in the Codebook by searching for the following: “))) HOW TO NAVIGATE THE CSES MODULE 4 CDEBOOK”
    5. Information about the identification variables has been updated in Part 1 of the Codebook, including detailed distinctions between Election Study identifiers, Polity identifiers, and Respondent identifiers.
    6. The CSES Bibliography (in part 1 of the Codebook) has been updated and sorted alphabetically.
    7. Part 3 of the Codebook (Parties and Leaders) has been updated. More parties, for which information was previously not listed has been added.
    8. A brief overview explanation of the processing check conducted by the CSES Module 4 Dataset by the CSES Secretariat has been added to Part 1 of the Codebook.
    9. For all macro variables, CSES Codebook now lists studies for which data is missing in variables note.
    10. Variable names in the Codebook have been updated to correct for small inconsistencies in the names between the dataset and the Codebook.
    11. List of tables is now available in each part of the CSES Codebook. Tables are now listed at the beginning of the Codebook. To allow for easy navigation, tables can be located by searching for “+++” in the CSES Codebook.
    12. A table in Part 1 of the Codebook have been updated and now contains information about studies included in the dataset, number of observations per study, fieldwork date for every study and mode of data collection in each country.
    13. A table has been added to the Codebook (Part 2) providing users with more information on the measurement of household income (D2020) by different studies. Some studies measure monthly income while others measure annual income. The table provides users with an overview of what measure each polity used.
    14. Table with a total number of electoral districts per polity and the total number of electoral districts represented in CSES data has been updated and now contains a percentage of all districts represented in the data per polity.
    15. Primary Electoral Districts table for Turkey (2015) has been updated listing 47 districts from which are respondents in the mentioned study.
    16. Additional sources for the South African 2014 election summary have been added. The previous version of the summary contained only one reference.
    17. Variable note for variable D1007 (Sample component) have been updated further explaining that the use of the variable is to “capture information about subsets of respondents that are meaningful but are not captured by other variables.”
    18. Election Study Note for variable D1007 (Sample Component) for United States (2012) has been updated to clarify that respondents who voted on Election day are coded as “1”, while early voters are coded as “2.”
    19. Election Study Notes for variable D2003 (Education) for Czech Republic (2013), Japan (2013) and Montenegro (2012) have been added explaining the recoding scheme from original study to ISCED format CSES uses.
    20. Election Study Note for variable D2003 (Education Level) for Iceland (2013) has been updated to clarify for the way variable is coded, since the Iceland Election Study asked two questions for education (one for completed education level and other for education that respondents have not finished).
    21. Election Study Note for variable D2003 (Education) for Sweden (2014) has been added explaining how the “Swedish Education Terminology (SUN)” system has been recoded to ISCED format CSES uses.
    22. Variable note for D2010 (Current Employment Status) has been updated with explanation of additional codes CSES uses on this variable.
    23. Election Study Notes for variables D2012 – D2019 have been updated to be more consistent in way Codebook report similar situations in data collection process from various election studies.
    24. Election Study Note for variable D2032 (Primary Electoral District) for United States (2012) has been updated to clarify that district data was collected for the presidential system at the state level because the US uses electoral college system that operates at the level of states.
    25. Table with a total number of electoral districts per polity and the total number of electoral districts represented in CSES data has been updated for United States (2012) and now has correct information that respondents for US study are from 41 districts.
    26. Variable Note for D3028_ (Find another job – respondent/spouse) has been updated to further clarify the CSES convention for cases in which these questions have been asked to respondents out of labor force. The CSES principle is not to recode these inconsistences but to report it in the Codebook.
    27. Variable note for D3016 (Optional Alternative Scale – Self-placement) has been updated with more precise information on what has been asked as this optional/alternative scale in the studies that have data on this variables.
    28. Variable D5006_1 and D5006_2 – the meaning of these two variables has been clarified in the variable notes.
    29. English names of parties for Norway (2013) in part 3 (Parties and leaders) have been updated and now contain a more precise translation.
    30. Party and leader table for Israel (2013) has been updated to correct for an English translation for Party A, Likud, and is now “National Liberal Party.”
    31. Party and leader table for Germany (2013) has been updated and alphabetical code “I” is added for party 10 (NPD).
    32. Party and leader table for New Zealand (2014) has been updated for Party E with the precise name “Conservative Party” and the correct abbreviation “CP.”
    33. Party and leader table for Phillipines (2016) has been updated for Party 24 with correct party name “One Patriotic Coalition of Marginalized Nationals (1-Pacman).”
    34. Party and leader table for Romania (2012) has been updated for Party 14 with correct party name “National Union for the Progress of Romania.”
    35. Party and leader table for Czech Republic (2013) has been updated, and code 22 has been deleted from the table since there is no party coded with “22” in the dataset.
    36. Party and leader table for Thailand (2011) has been updated with correct names of party 19 “New Aspiration Party” (English name) and part 20 “Phak Prachakon Thai” (Thai name).
    37. Party and leader table for Ireland (2011) has been updated with the original name of Party E – “Comhaontas Aontaithe an Chle (ULA).”
    38. Election Study Note for Parties and leaders table for Australia (2013) has been added explaining that there is data for LEADER G, despite the fact that there is no G code in the table. Data exist because this code refers to Julia Gillard, at the time outgoing Prime Minister and leader of the Australian Labor Party (ALP) until June 26, 2013.
    39. Election Study Notes for Parties and leaders table for Montenegro (2012) have been updated to more precisely convey the information about electoral alliances in Montenegro 2012 election and to which parties, part of the alliance, certain variables in the dataset refer.

ALREADY APPLIED

The following errata have already been applied to the most recent version of the dataset, and can be ignored by users of the current version of CSES Module 4.

CSES Module 4 Fourth Advance Release: April 11, 2017 Version

CSES Module 4 Fourth Advance Release: Data Set Errata
Posted: April 11, 2017

Updates made to the Advance Release of April 11, 2017, since the Advance Release of June 22, 2016

The fourth advance release of April 11, 2017 contains additions, updates, and improvements to the third advance release of June 22, 2016. Please do not use files from the previous release with files from the current release. Variable names and attributes, as well as the data, have sometimes been changed.

Nine new election studies have been added to the file, bringing the number of election studies included so far in Module 4 to thirty-eight in total. The studies added as of the Fourth Advance Release are: Finland (2015), Great Britain (2015), Hong Kong (2012), Kenya (2013), New Zealand (2014), Philippines (2016), Romania (2012), Slovakia (2016), and South Africa (2014).

Some of the changes and corrections to previously released Module 4 data are listed here, though this is not a comprehensive list:

    1. The appendix of the CSES Module 4 codebook now consists of three sections (previously one). The three sections provide separate information about the party and leader codes (Appendix I), the primary electoral districts (Appendix II), and election summaries (Appendix III).
    2. All sections of the Module 4 codebook (Section 1, Section 2 and Appendix I) now include more extensive information about the CSES coding schemes of parties, leaders, and alliances.
    3. National Democratic Party of Germany (NPD) was added to the Party/Leader table for Germany (2013). The party was included in the dataset but not in the party table.
    4. The values and value lables for variable D1003 (ID VARIABLE – ELECTION STUDY NUMERIC POLITY) for Sweden (2014) and the Czech Republic (2013) were recoded to correct an error in the initial coding fo the election year.
    5. Variables D1003, D1005, and D1007 previously indicated that respondents for Norway (2013) originated from two separate sample components. However, this is not the case and the variables were recoded accordingly.
    6. The variable D1005 (ID VARIABLE – RESPONDENT) was corrected for all cases for Sweden (2014).
    7. Variable D1007 (SAMPLE COMPONENT) for Sweden (2014) previously indiated that there was only one sample component. However, there were four sample components included in the Swden (2014) study, and this has been corrected.
    8. All missing cases for Norway (2013) for variable D1028 (DATE QUESTIONNAIRE ADMINISTERED – YEAR) were recoded to the correct CSES missing code “9999. MISSING”. These had been coded as “99” before.
    9. Variable D1030 (LANGUAGE OF QUESTIONNAIRE ADMINISTRATION) was recoded for all respondents for Brazil (2014) from “1. AFRIKAAS” to “097. PORTUGESE”.
    10. Data for South Korea (2012) for variable D2003 (EDUCATION) were added to dataset.
    11. Data for variable D2015 (SPOUSE: CURRENT EMPLOYMENT STATUS) for Sweden (2014) were recoded to “99. MISSING”. These cases had been coded as “9. PERMANENTLY DISABLED” in the previous release.
    12. The classification schemes for D3006_LH_PF and D3006_UH_PF were revised to include a new category, “2. RESPONDENT CAST PREFERENCE VOTE IN AV/STV SYSTEM”. This is to take into acount that a preference vote can be excercised in AV/STV systems, too and that it is distinct to a preference vote in a PR List System. The data for Australia (2013) was recoded accordingly to code “2” for variable D3006_LH_PF.
    13. Variable D2027 (LANGUAGE USUALLY SPOKEN AT HOME) for Turkey (2015) was corrected and recoded after a number of wrong codes were detected.
    14. Variable D3006_LH_PF was recoded from “9. MISSING” to “0. CANDIDATE PREFERENCE VOTE IS NOT APPLICABLE” for all respondents for Canada (2011) because Canadian voters cannot cast a preference vote in the lower house election.
    15. Variable D3006_LH_PF for for Ireland (2011) were recoded from “0. CANDIDATE PREFERENCE VOTE IS NOT APPLICABLE” to “9. MISSING” because voters can cast a preference vote in Ireland but these data are unavailable.
    16. Answers to to D3006_LH_PL and D3008_LH_PL for Canada (2011) were recoded to D3006_LH_DC and D3008_LH_DC to reflect that Canadian voters elect candidates at the district level and not on the basis of party lists. Further, the original data included a number of respondents that had refused to answer the vote choice questions, current and previous. These had been coded as “99. MISSING” but have been recoded to “97. VOLUNTEERED: REFUSED”.
    17. Data for variables D3008_UH_DC_1 and D3008_UH_DC_2 were added for Brazil (2014).
    18. For variable D3018_3 (WHICH PARTY DO YOU FEEL CLOSEST TO), 177 respondents for Montenegro (2012) had been coded as feeling close to Bosnjacka Stranka (PARTY 10). This number was incorrect and was corrected accordingly.
    19. Variables D3025_2_A, D3025_2_B, D3025_4_A, and D3025_4_B were recoded to missing for all cases for Canada (2011). These items had been included in a pretest in the original study and differed from the final CSES questionnaire.
    20. Thirteen new district variables were added to account for countries in which nationwide electoral districts are applied. The thirteen new variables have variables names which end in “_N” (for example, D4001_N).
    21. District variables (D4001, D4003, D4004_A-D4004_F and D4005) were recoded for Turkey (2015) after one respondent was found to have been incorrectly coded for all of these variables. This resulted from a wrongly included district “Adapazari” which is part of the larger district “Sakarya”. The district table and the coding were adopted accordingly.
    22. Variable D4003 (NUMBER OF PARTY LISTS IN DISTRICT) was recoded from “999. MISSING” to “000. NOT APPLICABLE” for all respondents for Canada (2011) because there are candidates competing in a district but not party lists.
    23. D5015 (NUMBER OF PARTIES PARTICIPATING IN ELECTION) was recoded for respondents for New Zealand (2011) to 13 to reflect the number of parties that received votes at tier 2.
    24. Variables D5023* and D5024* have been corrected for Israel (2013) to indicate the correct date for which the election was scheduled (October 22, 2013) and the date on which the elections were held (January 22, 2013).
    25. D5023_2 and D5024_2 were recoded for respondents for Czech Republic (2013) to reflect that elections were held on two days. It is not coded as “96. SEE ELECTION STUDY NOTES”
    26. D5046_2 was recoded for respondents for New Zealand (2011) to 96.00 OTHER THRESHOLD [SEE ELECTION STUDY NOTES] (Previously coded as “5”) as New Zealand operates an alternative threshold. D5047_2 was recoded for respondents for New Zealand (2011) to 4. OTHER [SEE ELECTION STUDY NOTES] (Previously coded as “2”) as New Zealand operates an alternative threshold.
    27. D5073 was recoded for respondents for New Zealand (2011) to “41.7%”. This is to reflect the constant expected value rather than the actual value that varies depending on uberhangmandates
    28. Missing values for D5080_1, D5081_1, D5082_1 and D5082_1 were corrected for Portugal (2015).
    29. Missing codes for D5081_1 – D5081_3 (CDP PER CAPITA) were adjusted from 99999. MISSING to 999999. MISSING.
    30. The net Migration Rate data (D5088 – NET MIGRATION RATE) has been separated into two separate measures, D5088_1 (NET MIGRATION RATE 2010-2015) and D5088_2 (NET MIGRATION RATE 2005-2010).
    31. D5086 was recoded for respondents for Brazil (2014) to “43”. This previous value was a miscode.
    32. The data for D5091_1 through D5091_3 (HEALTH EXPENDITURE (% GDP) TIME) were updated for all countries as we identifed inconsistencies in the previously released data. 
CSES Module 4 Third Advance Release: June 22, 2016 Version

List of known minor issues in the Third Advance Release of June 22, 2016.

The following minor issues were discovered late in processing of the Third Advance Release, and will be researched and corrected as appropriate in a future release:

  1. The storage format for variables D5014, D5015, D5046_1, D5104, and D5105 is not optimized appropriately and may display with more decimal points than are necessary, but the data are correct.
  2. When reading the raw data into STATA using the associated syntax users, STATA may assign more decimal places and precision than are present in the original data due to the methods by which STATA stores data. This is just how STATA works. But we think it should not change analyses substantively, in any case.
  3. Variable D5023_1 (DATE ELECTION SCHEDULED – MONTH) For Israel (2013) is coded for all cases as “22” when in fact the data the election was scheduled was October 22, 2013.
  4. One case in variable D1028 (DATE QUESTIONNAIRE ADMINISTERED) is coded as “99”.
  5. In variable D5080_1 (GDP GROWTH), Portugal (2015) is coded as “999” whereas the missing data code for that variable is “99”. In variable D5081_1 (GDP PER CAPITA) and D5081_2, Portugal (2015) is coded as “999” whereas the missing data code for that variable is “9999”. In D5084_1 (POPULATION, TOTAL), Portugal (2015) is coded as “999” instead of the correct missing data code.(
  6. In variable D5081_1 (GDP PER CAPITA), Norway (2013) is coded as “99999” whereas the missing data code for that variable is “9999”. In variables D5081_2 and D5081_3, Norway (2013) is coded as a number larger than the missing data value and should be discarded as missing data.
  7. In variable D5084_1, Turkey (2015) and Mexico (2015) are assigned unreasonably large scores and they should instead be assigned to missing (999999999).
  8. We are aware of some erorrs in variable D5091_1 (HEALTH EXPENDITURE) for a number of election studies.
  9. Austria (2013) has 96 cases which show 0 seats in the district (variable D4001).
  10. In variable D4005, Turkey (2015) has a single case with a turnout of “0”.
  11. There are ten cases who list that they came to the country prior to when they were born (variables D2001 and D2034).

Updates made to the Advance Release of June 22, 2016, since the Advance Release of March 20, 2015

The third advance release of June 22, 2016 contains additions, updates, and improvements to the second advance release of March 20, 2016. Please do not use files from the previous release with files from the current release. Variable names and attributes, as well as the data, have sometimes been changed.

Twelve new election studies have been added to the file, bringing the number of election studies included so far in Module 4 to twenty-nine in total. The studies added as of the Third Advance Release are: Brazil (2014), Bulgaria (2014), Canada (2011), Czech Republic (2013), Israel (2013), Mexico (2015), Norway (2013), Portugal (2015), Slovenia (2011), South Korea (2012), Sweden (2014), and Turkey (2015).

Some of the changes and corrections to previously released Module 4 data are listed here, though this is not a comprehensive list:

    1. To facilitate merging CSES data with other (macro-) datasets, CSES decided to introduce two additional ID variables across all modules. The new variable 1006_UN (ID VARIABLE – POLITY UN CODE) contains the UN Country Codes. These are also contained in variable 1006 but here are followed by a fourth digit. The new variable 1006_NAM (ID VARIABLE – POLITY NAME) contains verbatim country names.
    2. Variable D3008_UH_DC has been differentiated into D3008_UH_DC_1 and D3008_UH_DC_2 to accomodate cases in which two votes can be cast for upper house district candidates (in this release notably Brazil (2014)).
    3. Because of data availability limitations, variable D5087 (INTERNATIONAL MIGRANT STOCK (% POPULATION)) was removed from this current module.
    4. On variable D2030 (ETHNICITY), Montenegro (2012) and New Zealand (2011) had idiosyncratic codes for the value ‘Other, not further specified’. This has been corrected so that the generic code ‘996. OTHER: NOT SPECIFIED’ is now also used for these countries.
    5. Variable D4002 (NUMBER OF CANDIDATES IN DISTRICT) has been coded ‘000. NOT APPLICABLE’ for Austria (2013), Iceland (2013), Montenegro (2012), and Serbia (2012).
    6. Variable D4003 (NUMBER OF PARTY LISTS IN DISTRICT) has been coded ‘000. NOT APPLICABLE’ for Austria (2013), Germany (2013), France (2012), Ireland (2011), Japan (2013), Mexico (2012), New Zealand (2011), Thailand (2011), Taiwan (2012), and the United States of America (2012).
    7. For Germany (2013), variable D5013_B and D5013_E (NUMBER OF PORTFOLIOS AFTER) has been recoded as codes for parties B and E (SPD and CSU) had been incorrectly allocated.
    8. For Greece (2012), variable D3018_3 (WHICH PARTY DO YOU FEEL CLOSEST TO), several party codes had to be recoded because of previously erroneous assignment.
    9. For Greece (2012), variable D3028_1 (FIND ANOTHER JOB – RESPONDENT), code 6 has been recoded to 4.
    10. For Iceland (2013), variable D5041_1 (PARTY LISTS – LOWER – 1ST SEGMENT (TIER)) has been recoded from ‘1. CLOSED’ to ‘2. OPEN’.
    11. The vote choice data for Ireland (2011) has been coded in variable variable D3006_LH_DC (CURRENT LOWER HOUSE ELECTION: VOTE CHOICE – DISTRICT CANDIDATE). Before, it was erroneously coded in variable D3006_LH_PL (CURRENT LOWER HOUSE ELECTION: VOTE CHOICE – PARTY LIST).
    12. For Mexico (2012), the variable D2003 (EDUCATION OF RESPONDENT) has been recoded. In the prior release, incomplete educational categories had erroneously been coded as if they were completed educational attainments. This has been corrected.
    13. For Mexico (2012) variables D5044_1 to D5044_4 (concerning cumpolsory voting) have been recoded. They had all been erroneously coded ‘5. NO’. As voting is actually compulsory in Mexico, D5044_1 (COMPULSORY VOTING – LOWER – 1ST SEGMENT (TIER)) and D5044_3 (COMPULSORY VOTING – UPPER – 1ST SEGMENT (TIER)) have been recoded to ‘3. YES; WITHOUT SANCTION FOR VIOLATION’. As no seperate votes are cast for the second segments of both houses, the variables concerning the second segments, D5044_2 (COMPULSORY VOTING – LOWER – 2ND SEGMENT (TIER)) and D5044_4 (COMPULSORY VOTING – UPPER – 2ND SEGMENT (TIER)) have been coded ‘7. NOT APPLICABLE’.
    14. For New Zealand (2011), variable D1006 (ID COMPONENT – POLITY) has been recoded into 5540.
    15. For New Zealand (2011), variable D2026 (RELIGIOUS DENOMINATION), code 9600 has been recoded to 9002.
    16. For Switzerland (2011), variables D3025_1_A (DICHOTOMIZED POLITICAL INFORMATION ITEM – 1ST) – D3025_4_B (ORIGINAL POLITICAL INFORMATION ITEM – 4TH) have been amended such that codes ‘7. REFUSED’ and ‘8. DON’T KNOW’ now also contain cases. For the last release, these cases had erroneously been coded into ‘9. MISSING’.
    17. For the United States (2012), variable D2020 (HOUSEHOLD INCOME) has been added. Furthermore, the variable D2029 (RACE) has been slightly adjusted. An explanation has been inserted in the election study note for D2029.
    18. For the United States (2012), variable D3021_1 (MOBILIZATION: INSTITUTIONAL CONTACT – WHO – MULTIPLE MENTIONS – FIRST MENTION), code 5 has been recoded to 90.
    19. For the United States (2012), variable D5004_A and D5004_B PERCENT (SEATS – UPPER HOUSE) have been recoded so that percentage number of seats refer to the election in question and do not include holdover seats.
CSES Module 4 Second Advance Release: March 20, 2015 Version

Posted: March 20, 2015; Updated January 17, 2016

Updates made to the Advance Release of March 20, 2015, since the Advance Release of April 28, 2014

The second advance release of March 20, 2015 contains additions, updates, and improvements to the first advance release of April 28, 2014. Please do not use files from the previous release with files from the current release. Variable names and attributes, as well as the data, have sometimes been changed.

Some of the changes between the first and second advance Releases are listed here, though this is not a comprehensive list:

  1. Seven new election studies have been added to the file, bringing the number of election studies included so far in Module 4 to 17 in total. The studies added as of the Second Advance Release are: Australia (2013), Germany (2013), Greece (2012), Iceland (2013), Japan (2013), Poland (2011), and United States (2012).
  2. In the New Zealand (2011) election study, the coding of the party positions on the left-right scale for Parties E-H inclusive were incorrect due to a coding error. The scores for parties E, F, G, and H have been revised.

Posted: Apr 28, 2014
Excessive precision after the decimal, in variables with decimals

In the First Advance Release, for variables with decimal points in them, the exported variables have precision after the decimal point in excess of what is required or appropriate.

For example, the value “10.977” might be stored and presented as “10.97700023651120” when it is only intended to be precise to three digits to the right of the decimal.

This is a function of the storage format in STATA (the statistical package in which CSES processes its recent files) and the export method used by the project, and will be corrected in future releases.

We have added some formatting statements in the STATA syntax (“do”) file to partially address this, but it will be more noticeable in other statistical packages.

Please refer to the variables section of the codebook to be informed as to what precision is appropriate for each variable with a decimal in it.

CSES Module 4 First Advance Release: April 28, 2014 Version

(no errata listed)