=========================================================================== COMPARATIVE STUDY OF ELECTORAL SYSTEMS (CSES) - MODULE 4 (2011-2016) CODEBOOK PART 6: STUDY DESIGNS AND WEIGHTS FULL RELEASE - MAY 29, 2018 CSES Secretariat www.cses.org =========================================================================== HOW TO CITE THE STUDY: The Comparative Study of Electoral Systems (www.cses.org). CSES MODULE 4 FULL RELEASE [dataset and documentation]. May 29, 2018 version. doi:10.7804/cses.module4.2018-05-29 These materials are based on work supported by the American National Science Foundation (www.nsf.gov) under grant numbers SES-0817701, SES-1154687, and SES-1420973, the GESIS - Leibniz Institute for the Social Sciences, the University of Michigan, in-kind support of participating election studies, the many organizations that sponsor planning meetings and conferences, and the numerous organizations that fund national election studies by CSES collaborators. Any opinions, findings and conclusions, or recommendations expressed in these materials are those of the author(s) and do not necessarily reflect the views of the funding organizations. =========================================================================== =========================================================================== TABLE OF CONTENTS =========================================================================== ))) IMPORTANT NOTE REGARDING FULL RELEASE ))) OVERVIEW OF "CODEBOOK PART 6: STUDY DESIGNS AND WEIGHTS" ))) HOW TO NAVIGATE THE CSES MODULE 4 CODEBOOK ))) LIST OF TABLES IN CODEBOOK PART 6 ))) OVERVIEW OF STUDY DESIGN AND WEIGHTS BY POLITY >>> POLITY WEIGHTS IN CSES >>> TYPE OF POLITY WEIGHTS BY ELECTION STUDY >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ARGENTINA (2015) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - AUSTRALIA (2013) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - AUSTRIA (2013) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - BRAZIL (2014) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - BULGARIA (2014) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - CANADA (2011) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - CANADA (2015) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - CZECH REPUBLIC (2013) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - FINLAND (2015) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - FRANCE (2012) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - GERMANY (2013) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - GREAT BRITAIN (2015) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - GREECE (2012) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - GREECE (2015) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - HONG KONG (2012) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ICELAND (2013) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - IRELAND (2011) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ISRAEL (2013) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - JAPAN (2013) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - KENYA (2013) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - LATVIA (2011) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - LATVIA (2014) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - MEXICO (2012) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - MEXICO (2015) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - MONTENEGRO (2012) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - NEW ZEALAND (2011) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - NEW ZEALAND (2014) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - NORWAY (2013) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - PERU (2016) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - PHILIPPINES (2016) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - POLAND (2011) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - PORTUGAL (2015) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ROMANIA (2012) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ROMANIA (2014) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SERBIA (2012) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SLOVAKIA (2016) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SLOVENIA (2011) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SOUTH AFRICA (2014) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SOUTH KOREA (2012) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SWEDEN (2014) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SWITZERLAND (2011) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - TAIWAN (2012) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - THAILAND (2011) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - TURKEY (2015) >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - UNITED STATES (2012) =========================================================================== ))) IMPORTANT NOTE REGARDING FULL RELEASE =========================================================================== This dataset and all accompanying documentation is the "Full Release" of CSES Module 4 (2011-2016). Users of the Final Release may wish to monitor the errata for CSES Module 4 on the CSES website, to check for known errors which may impact their analyses. To view errata for CSES Module 4, go to the Data Center on the CSES website, navigate to the CSES Module 4 download page, and click on the Errata link in the gray box to the right of the page. =========================================================================== ))) OVERVIEW OF "CODEBOOK PART 6: STUDY DESIGNS AND WEIGHTS" =========================================================================== Part 6 of the CSES Codebook provides users with general information about the design of each election study included in Module 4 as well as the original polity level weights for each study. The purpose of these overviews is to provide users with a quick summary of the study and the weights including information about the sample size, the fieldwork period, sample selection procedures, the mode(s) of interview, and the language(s) the survey was administered in. For more detailed information about each study's design and its weights, users are advised to consult the Design Reports for each polity included in the study which are available on the CSES Module 4 Study Page (see http://www.cses.org/datacenter/module4/module4.htm). =========================================================================== ))) HOW TO NAVIGATE THE CSES MODULE 4 CODEBOOK =========================================================================== In the CSES Module 4 dataset, all variables begin with the letter "D" (D being the fourth letter of the English alphabet and thus signifying Module 4). The CSES codebook is especially extensive and users are advised that the best way to navigate it is electronically. It is a .txt format which allows it to be accessed via a variety of programmes. The CSES Codebook can be navigated quickly in the electronic files, with the following commands allowing for quick searching: ))) = Section Header >>> = Sub-section Header 1 <<>> = Sub-section Header 2 +++ = Tables VARIABLES NOTES = Notes for particular variables ELECTION STUDY NOTES = Notes for a particular election study For further details on the CSES Module 4 documentation, users are advised to consult Part 1 of the CSES codebook. =========================================================================== ))) LIST OF TABLES IN CODEBOOK PART 6 =========================================================================== Below, we list the Tables located in Codebook Part 6. Tables can be accessed in the electronic version of the CSES Codebook by searching for "+++". - TYPE OF POLITY WEIGHTS BY INDIVIDUAL ELECTION STUDIES =========================================================================== ))) OVERVIEW OF STUDY DESIGN AND WEIGHTS BY POLITY =========================================================================== --------------------------------------------------------------------------- >>> POLITY WEIGHTS IN CSES --------------------------------------------------------------------------- CSES provides users with up to three original weights from each national election study (see variable D1010_) namely: - SAMPLE WEIGHT (variable D1010_1): intended to correct for unequal selection probabilities resulting from booster samples procedures for selection within the household, non-response, or other sample design features - DEMOGRAPHIC WEIGHT (variable D1010_2): intended to adjust sample distributions of socio-demographic characteristics to more closely resemble the characteristics of the population - POLITICAL WEIGHT (variable D1010_3): intended to reconcile discrepancies in the reported electoral behavior of respondents vis-a-vis official electoral counts. Users are advised to read carefully about the different weights in CSES to ascertain whether their analyses should be subjected to weighting and if so which kind. The CSES project does not provide advice as to which weights are appropriate to use in particular circumstances. This is best left to analysts to decide based on their detailed knowledge of the research question under investigation. We advise analysts to consult each Polity's Design Report on the CSES Module 4 Study Page (see http://www.cses.org/datacenter/module4/module4.htm). For information on derivative weights calculated for the Cross-National Dataset, users should consult Part 1 and Part 2 (variables D1011-D1014) of the CSES Codebook. --------------------------------------------------------------------------- >>> TYPE OF POLITY WEIGHTS BY ELECTION STUDY --------------------------------------------------------------------------- | +++ TABLE: TYPE OF POLITY WEIGHTS | BY INDIVIDUAL ELECTION STUDIES | | Sample Demographic Political | POLITY (ELEC YEAR) Weight Weight Weight | ----------------------------------------------------------- | AUSTRALIA (2013) - X - | AUSTRIA (2013) - X - | BRAZIL (2014) X - - | BULGARIA (2014) - X - | CANADA (2011) X - - | CANADA (2015) X - - | CZECH REPUBLIC (2013) - - X | FINLAND (2015) - X X | FRANCE (2012) - X X | GERMANY (2013) X X - | GREAT BRITAIN (2015) X X - | GREECE (2012) - X - | GREECE (2015) - X X | HONG KONG (2012) - X - | IRELAND (2011) - X X | JAPAN (2013) - X - | KENYA (2013) X - - | LATVIA (2011) X X - | LATVIA (2014) X - - | MEXICO (2012) X X - | MEXICO (2015) - X - | MONTENEGRO (2012) - X - | NEW ZEALAND (2011) - X X | NEW ZEALAND (2014) - X X | NORWAY (2013) - X - | PERU (2016) X X X | PHILIPPINES (2016) X - - | POLAND (2011) - X - | PORTUGAL (2015) - X - | ROMANIA (2012) X X X | ROMANIA (2014) X X X | SERBIA (2012) X X - | SLOVAKIA (2016) - X - | SOUTH AFRICA (2014) - X - | SWITZERLAND (2011) X - X | TAIWAN (2012) - X - | THAILAND (2011) X - - | TURKEY (2015) - X - | UNITED STATES (2012) X X - | ----------------------------------------------------------- | KEY: X = available; - = not available. | | Weights are unavailable for ARGENTINA (2015), ICELAND (2013), | ISRAEL (2013), SLOVENIA (2011), SOUTH KOREA (2012) and | SWEDEN (2014). --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ARGENTINA (2015) --------------------------------------------------------------------------- <<>> STUDY DESIGN Elections in Argentina were held on October 25, 2015. The post-election Interviewing began on November 21 (27 days after the Election day) and lasted for 39 days, ending on December 30, 2015. The face-to-face mode was employed for all interviews. The Argentinian election study was conducted as a panel study in two waves. CSES questions were distributed in the second wave, which contained in total 1149 respondents. Of those respondents, 780 had been sampled initially for the first survey (panel attrition between two waves was 68%), and 369 respondents were a refresh sample. The fieldwork was conducted by MBC Mori. The questionnaire was administered in the Spanish language, and the target population for the study was all Argentian citizens older than 18 years. A multistage sample design with a total of six stages was used. In the first stage, the country's 23 provinces (plus the federal district of Buenos Aires) were divided into four electoral strata based on 2011 presidential election results, using as the stratifying variable the percentage of valid votes obtained by incumbent Kirchner. Provinces were classified into these four strata and cities/towns were selected within each stratum. Stages two through six were conducted according to standard household sampling procedures, using different types of probabilistic selection procedures at stages two to five, where the population is composed of geographic areas or households. These procedures included sampling of census fractions and census radiuses with Probability Proportional to Size (stages 2 and 3) as well as a simple random sampling of blocks within census radiuses (stage 4) and systematic random sampling of households within blocks (stage 5). Finally, a quota procedure was employed for stage six where individuals within households were sampled. More detailed information about the sampling procedure is available in the Design Report. <<>> POLITY WEIGHTS No weights provided. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - AUSTRALIA (2013) --------------------------------------------------------------------------- <<>> STUDY DESIGN The Australian Election Study was carried out between September 6, 2013 and January 7, 2014. The mode of the interview was a mail-back questionnaire. The fieldwork period began one day before the Federal Election on September 7, and in total took 123 days. The questionnaire was administered in English and the fieldwork was carried out by the Survey Research Centre Pty Ltd. The study collected information from voters, that is Australian adults aged 18 years or over who were enrolled on the electoral register, and eligible to vote. A stratified random sampling procedure was used to compile a sample of respondents from the Australian Electoral Register which is administered by the Australia Electoral Commission (AEC). Initially, 15,000 records were provided, 100 from each electorate. The sample was provided by the AEC for use only for this study. Due to budget considerations, only 12,200 of the original sample was selected for the initial mail-out. The following strategy was used for sub-selection: 1) Randomly select 81 voters from each electorate (totally 12,150) 2) After Step 1, randomly select 1 additional voter from 50 randomly selected electorates. Sample records with an overseas mailing address were removed from the mailing sample. A total of 3,955 Surveys (3,379 via mail-back and 576 online) were received. <<>> POLITY WEIGHTS The Australian Election Study provides a DEMOGRAPHIC WEIGHT. This weight was calculated to correct for the following: the disproportionate probability of selection, non-response in the survey, and to match known demographic characteristics of the population. This weight was calculated using rim weighting adjusting to four benchmarks. Figures for age, gender, and the state of residence were obtained from the AEC enrollment data for the 2013 election. Figures for voting came from the AEC Final Election Vote Tallies. For the category "(Voted Informal)/Did not vote," "Voted informally" is bracketed as this response is included in the dataset but not in the benchmark. The numbers represent only the number of enrolled voters who did not vote, i.e., the difference between the vote tally and the number of enrollments). Counts of invalid votes were obtained. However, it was impossible to ascertain how many invalid votes were deliberate (in which the case the respondent would give the "invalid vote" response) and how many were unintended (in which case the respondent would most likely not realize their vote was invalid and respond with their voting intention). --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - AUSTRIA (2013) --------------------------------------------------------------------------- <<>> STUDY DESIGN The elections in Austria were held on September 29, 2013. The survey interviewing began on October 1, and lasted for 29 days. The interviews were conducted via the telephone by Jaksch & Partner Institute for statistics. All interviews were conducted in the German language. The total of 1000 interviews have been completed for the Austrian study, and the sample is meant to be representative of the Austrian population, persons aged 16 and older, eligible to vote in the national election 2013. The sampling method employed was stratified random sampling. The primary sampling factor was the nine Austrian provinces ("Bundeslaender"). Only within these provinces, an unlimited random sample was drawn. The provinces were then divided into administrative districts proportional to their population size. Phone numbers were randomly selected using dual sampling frame (RDD: 89.1% and RLD: 10.9%). The last birthday method was used to randomly select the respondent within each household. The characteristics used for stratification were geographic (provinces, administrative districts and city/town sizes). <<>> POLITY WEIGHTS The Austrian Election Study provides a DEMOGRAPHIC WEIGHT. These are necessary to even out minor offsets in the sample that did not perfectly coincide with the target population. It was created on the basis of socio-demographic characteristics (gender, age, education, household size, region/Bundesland, and employment status). --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - BRAZIL (2014) --------------------------------------------------------------------------- <<>> STUDY DESIGN The survey was administered between November 1, 2014, and November 19, 2014. The interviewing started 5 days after the second round of elections in Brazil. The face-to-face mode was used for the study, and interviews were conducted in the Portuguese language. The fieldwork was coordinated by IBOPE Inteligencia. Sampling was conducted in three stages and is meant to be representative of the Brazilian population over 16 years that is registered to vote. Random probability sampling was implemented in two of the three stages: I) Cities within stratum (state) [Systematic PPS (Probability Proportional to Size)]; II) Census tracts within cities [Systematic PPS]; III) The household selection was made by quota: gender, age, education and the line of work. The primary sampling units are composed of cities. The sample is stratified per state. In case of states with metropolitan areas, its universe is stratified across metropolitan areas and countryside. The cities are probabilistically selected through systematic PPS (Probability Proportional to Size) in each stratum, based on the population of those above 16 years of age. Through this method, 194 cities were selected. The total of 3136 interviews were completed in Brazilian study. For research reasons, the state of Sao Paulo has a complement sample in the capital, suburbs and interior to allow interpretation of the results for each of these areas. The selection of cities in the periphery and interior follows the same criterion for other strata, i.e., systematic PPS based on population aged 16 or more. <<>> POLITY WEIGHTS The Brazilian Election Study provides a SAMPLE WEIGHT. It is constructed to correct for oversampling in the state of Sao Paulo. Oversampling occurred because of deliberate oversampling of the Sao Paulo region (for more details see study design description above). The weights are constructed so that complementary sample of the state of Sao Paulo has the proportional effect in the whole sample. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - BULGARIA (2014) --------------------------------------------------------------------------- <<>> STUDY DESIGN The post-election survey, containing the CSES Module 4 questionnaire, was conducted between January 23-31, 2015, slightly more than three months after the election. All respondents were surveyed in face-to-face mode and interviews were conducted in the Bulgarian language. The fieldwork was coordinated by TNS BBSS SEE research agency in Bulgaria. The total of 999 interviews were completed, and the sample is meant to be representative of adult (18+ years old) population of people living in Bulgaria. The sampling points were distributed proportionally to regional and urbanization characteristics of the population. The number of sampling points in each region for the different type of settlements was calculated based on the ratio regional center/other urban area/rural area. On a regional level, separately for each region, the settlements belonging to the same urbanization group were put in alphabetical order. The settlements included in the survey were selected at random from that list. Additionally, for the biggest cities (capital Sofia, Plovdiv, and Varna), sampling points were distributed in smaller administrative units (municipalities) proportionally to the size of the population in the respective city. Institutionalized persons were excluded from the sample, which is estimated to be 3% of total population. <<>> POLITY WEIGHTS The Bulgarian Election Study provides a DEMOGRAPHIC WEIGHT. It ensures that the sample is representative of the population. Weights tend to adjust for gender, age, urbanization level, and region of residence (NUTSII-level). These weights were constructed by using a weight matrix based on information about the statistical distribution of the population according to the National Statistical Institute of Bulgaria. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - CANADA (2011) --------------------------------------------------------------------------- <<>> STUDY DESIGN The study was part of a panel study. A pre-election campaign study (CPS) started on March 26, 2011. It originally included new 2011 RDD sample respondents (n=3,458), and individuals who had already participated in the Canadian Election Study in 2004, 2006 or 2008 (n = 850). However, the latter were removed from the CSES data. The interviewing was conducted via phone, for this phase of the study, and performed by Institute for Social Research and Jolicoeur & Associates which did interviewing in Quebec. The RDD sample for the 2011 campaign study was designed to represent the adult population of Canada: Canadian citizens 18 years of age or older who speak one of Canada's official languages, English or French, and reside in private homes in the ten Canadian provinces. The sample was selected via a modified random digit dialing (RDD) sample along with the birthday selection method within households. Residents of the three Northern territories were excluded from the sample as well as institutionalized persons and military personnel. That makes around 1.5% of the population. Additionally, 1.1% of households in Canada are without a phone. The interviews were conducted in English and French languages. The sample for the post-election surveys was comprised of respondents to the CPS. At the end of the CPS, the interviewer ensured that they had a first name or some other identifier (such as the respondent's initials or position in the household, e.g., mother). This information, as well as the sex and year of birth of the CPS respondent, and the respondent's telephone number was recorded on a "cover sheet." The elections in Canada were held on May 2, 2011. Calling for the post-election study (PES) started on May 3. The interviewer called and asked for the person by name or identifier. By day 30, more than 80% of the interviews were completed, but small numbers of interviews were only completed by early July. In total, 3,362 PES interviews were completed, of which 2,595 were from the new 2011 RDD sample. At the end of the post-election survey, respondents were asked to provide their address so they could be sent the mail-back survey (MBS). 30% of the PES respondents declined to provide an address. The PES respondents who provided mailing addresses received up to five contacts encouraging them to complete and return the mail-back questionnaire. The number of mail-back completions was 1,567. Of those who provided an address and were sent the MBS, 67% returned a completed questionnaire, and this represents 47% of all PES respondents. <<>> POLITY WEIGHTS The Canadian Election Study provides a SAMPLE WEIGHT. In order to produce national estimates, this weight correct for both the unequal probability of selection at the household stage and the unequal probability of selection based on province of residence. The weight variable is the product of two original weights, the household, and the province weight. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - CANADA (2015) --------------------------------------------------------------------------- <<>> STUDY DESIGN The study was part of the panel study. Three waves of the panel were conducted. The first wave, campaign period study (CPS) began on September 8 2015. By the October 19, the Election Day, 4,202 interviews were completed in the first wave. Interviews were conducted using the CATI (Computer Assisted Telephone Interviewing) technique in English and French. Fieldwork was conducted by Institute for Social Research. Respondents were drawn from random digit-dialing (RDD), a sample representing the adult population of Canada: Canadian citizens 18 years of age or older who speak one of Canada's official languages, English or French, and reside in private homes in the ten Canadian provinces. The birthday selection method was used to select respondents within the households. Residents of the Northern territories were excluded from the sample since they are difficult to sample. Additionally, institutionalized persons, as well as military personnel, were excluded from the sample. The excluded territories and groups contain around 1.5% of the population. Post-election study (PES) began on October 20, the day after the election. All respondents from the first (CPS) wave of the study, that could be contacted and wanted to take part in the survey, were interviewed. Again, the CATI technique was used for the PES interviewing. After 64 days, by the December 23 (last day of fieldwork for PES) 2,988 interviews were completed. The PES reinterview rate in 2015 was 71%. The third wave of the study was conducted as a mail-back survey (MBS). In the PES respondents were asked at the end to provide their addresses so that they could be sent the mail-back survey. The mail-back survey was hence sent only to those respondents who voluntarily gave the address and agreed to participate in this part of the study. Of those who provided an address at the end of the PES (70%) and were sent the MBS, 62% percent returned a completed questionnaire. In total, 1,289 respondents completed and sent back the survey. CSES questions were distributed among these three different components. <<>> POLITY WEIGHTS The Canadian Election Study provides a SAMPLE WEIGHT. It corrects for unequal probabilities of selection at the household stage and the unequal probabilities of selection based on province of residence. The provided variable is the product of household weight and a province weight. Household weight is included because the probability of an adult member of the household being selected for an interview varies inversely with the number of eligible adults living in that household. Regarding the province weight, because the sample was stratified, the distribution is not proportional to the population size of the provinces. Therefore, the data must be weighted before national estimates are derived. Weights are obtained by dividing the proportion of households in the province by the proportion of the households in the sample for that province. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - CZECH REPUBLIC (2013) --------------------------------------------------------------------------- <<>> STUDY DESIGN The interviewing for the Czech election study started two days after the elections in the country, and the survey was administered between October 28, 2013 and November 14, 2013. The mode of interviews was face-to-face/in person in the Czech language. The fieldwork was conducted by CVVM (Center for public opinion research) at the Institute of Sociology, Czech Academy of Sciences. For the Czech study, respondents were chosen according to a quota system which was designed for each of the 14 Czech regions (which correspond to the electoral districts). Quotas were based on age, gender, and education. The sample includes 1,653 respondents (Czech citizens older than 18 years) and is meant to be representative of all Czech Republican citizens of 18 years and older. Around 1% of the population, people who don't live in a house or apartment could not be reached, and that is why were excluded from the sample. <<>> POLITY WEIGHTS The Czech Election Study provides a POLITICAL WEIGHT. The weight adjusts the sample to percise results of the national election and were constructed based on election results. Parties who received less than 3% of the vote in the election were included in a combined category, labeled "other parties." --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - FINLAND (2015) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES Module 4 was implemented in the 2015 Finland Post-Election study that was carried out in the aftermath of the lower house elections which took place on April 19, 2015. The survey was administered between April 24 and July 7, 2015 (fieldwork lasting in total 74 days), using face-to-face interviews and drop-off questionnaires, in Finnish and Swedish language. The fieldwork was conducted by TNS Gallup Oy. The sample is a quota sample that is meant to be representative of all Finish citizens, 18 years and older, who were entitled to vote in 2015 Finish parliamentary election. In the first stage of the sampling procedure, the population was divided based on gender and age. The number of respondents required for the sample was then divided along NUTS2 regions and in a next step, along municipality lines. Interviewers were conducted with those persons in targeted households that fulfilled the age and gender criteria. Within their areas, interviewers themselves could choose any household, i.e., they were not provided with a list of target households by the survey company. Excluded from the sample were all Finish citizens living in the region of Asland Islands (equivalent to 0.49% of the total eligible population). Overall, 11,500 people were selected from the sampling frame. 1,587 of the selected people completed the interview. Of the 1,587 respondents, 684 returned the drop-off questionnaire which they were offered to fill out immediately after the interview (to be returned by mail). The drop-off questionnaire included questions on eight different variables from the CSES core module. For more information, see election study notes for variables D1007 (Sample Component) and D1023 (Survey Mode) in the Part 2 of CSES Codebook. <<>> POLITY WEIGHTS The Finnish Election Study provides a combined DEMOGRAPHIC and POLITICAL WEIGHT. These weights are constructed to correct the distributions for those of the eligible Finnish voting population with respect to mother tongue, age, gender, electoral district distribution and an actual vote share of parties in the election. The weights are combined and thus both the weights listed under variable D1010_2 and D1010_3 for Finland (2015) are identical. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - FRANCE (2012) --------------------------------------------------------------------------- <<>> STUDY DESIGN The interviewing for France election study began on May 10, 2012, 4 days after the second round of Elections in France, and lasted until June 9, 2012, lasting in total 30 days. TNS-Sofres conducted fieldwork. Interviews were held face-to-face in the French language. The sample is meant to be representative of the eligible population, French citizens, who are 18 years or older and registered to vote. Institutionalized persons and 4.5% of the population living in the overseas territories were excluded from the sampling frame because of the costs linked to interviewing in these regions. The sample was drawn as a stratified three stage probability sample. For the stratification, a table with 21 regions and five agglomeration classes were built. Each cell contained the corresponding population size. In the first stage of the sampling, 400 primary sampling units (districts) were allocated to the defined cells, using Cox's 'Method of controlled rounding.' The PSUs (municipalities, arrondissements of bigger cities or villages grouped in cantons) were selected from the cells according to the allocation proportional to population size. In the second stage, 5 (+20) addresses of the PSUs were selected via the random route. Failed addresses could be replaced by an extension of the random route. In stage three, individuals within households were selected via the 'Last-Birthday-Method.' In total, 2009 interviews were completed in the France election study. <<>> POLITY WEIGHTS The French Election Study provides two weights: a DEMOGRAPHIC WEIGHT and a POLITICAL WEIGHT. The demographic weights correct the distribution of known sociodemographic characteristics (sex, age, and occupation). The political weight adjusts vote to the official election results of the first and the second round of the presidential election. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - GERMANY (2013) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES module is part of the German post-election cross-sectional study, for which the fieldwork was conducted between September 23 and December 23, 2013, by MARPLAN Media- und Sozialforschungsgesellschaft mbH. Data collection started one day after the federal election on September 22 and lasted for 92 days in total. The questionnaire was administered in German and in person (CAPI). The sample is designed to be representative of all German citizens resident in the Federal Republic of Germany, who were living in private households, aged 18 or older, and who were eligible to vote. Originally the sample also included an additional number of 19 persons of age 16-17, but these were excluded from the CSES as they are not eligible to vote. The sampling method used was the Address-Random Method (ADM) design, with all private German households as the base population. 306 sampling points (voting districts) were randomly selected - 211 in West Germany and 95 in East Germany. In West Germany, 1,400 interviews or 6.6 interviews per sampling point on average were aimed for and in East Germany 700, or 7.4 interviews per sampling point on average, resulting in an oversampling of East Germany. In every sampling point, 60 households' addresses were randomly selected using the random route sampling. From this list of confirmed addresses, 25 households per sampling point were selected randomly. The third step was the selection of the target person via the Kish-Selection-Grid. A minimum of four attempts were made to contact a potential respondent, at different times of the day and days of the week, before declaring it a non-contact. There are 1889 completed interviews in the sample. <<>> POLITY WEIGHTS The German Election Study provides two weights: a SAMPLE WEIGHT and a DEMOGRAPHIC WEIGHT. The sample weight controls for the oversampling of East Germans and household size. The demographic weight controls for education, age, gender and size of the communities. User are advised that the weights were calculated by the German Election Study while still including 19 respondents who were ineligible to vote in the 2013 election. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - GREAT BRITAIN (2015) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES module was part of the British Election Study post-election cross- sectional survey. The survey was conducted between May 7 and September 29, 2015, a timespan of 146 days, with data collection beginning the day of the election. GfK UK Ltd was responsible for data collection. Whereas the main study was conducted face to face (CAPI), the CSES module was administered separately via self-completion after the interviewer had left the household. Respondents were free to complete the questionnaire either via paper-and- pencil (PAPI) or online (CAWI). All interviews were conducted in English. The sample was designed to be representative of citizens living in Great Britain, aged 18+ and who were eligible to vote in the 2015 general election. Those who are eligible to vote are British Citizens, Citizen of the Republic of Ireland or citizen of a Commonwealth Country who has a "right to remain" in the UK. The sample was selected based on a multi-stage design, summarized as: 1) A stratified random sample of 300 Parliamentary constituencies (primary sampling units) 2) Two Lower Layer Super Output Areas (LSOAs) per constituency were selected with probability proportional to size 3) Selection of addresses from the Small user Postcode Address File (PAF) 4) One individual randomly selected per address by the interviewer The exclusion of the Northern islands from the sample led to the exclusion of about 0.25% of the eligible population. Additionally, some sections of the population fell outside this sampling frame, e.g. elderly people in residential care, prisoners and military personnel living in defense establishments. This led to the exclusion of another 2% of the eligible population. 1567 respondents completed the main study and the CSES module. Before the study, all respondents received an incentive of 5 GBP together with a letter announcing the survey. After completion of the main study, participants received gift vouchers with varying values. Another voucher worth 5 GPB was given to those respondents who announced participation in the self- administered CSES module. For initial non-contacts, incentives were raised from 5 to 15 GBP. For much more detailed information on each step of the sampling and methods, please refer to the design report. <<>> POLITY WEIGHTS The British Election Study includes two different weights: a SAMPLE WEIGHT and a DEMOGRAPHIC WEIGHT. The Sample weight corrects for unequal selection probabilities, which might have occurred whenever an interviewer was left with the following choices: 1. To select a dwelling (typically flats) from a selected address 2. To select an individual household from a dwelling 3. To select an eligible person from the household A person's chance of being interviewed was calculated by multiplying the number of dwellings by the number of households within the selected dwelling, multiplied by the number of adults in the selected household. As the probability of selection is the inverse of this number, respondents were weighted by the result of the multiplication. The selection weight was capped at 5 (affecting 14 cases). Afterwards, the weight was rescaled to arrive at the original sample size. The Demographic weight accounts for differing levels of response from various groups in the population. For this second weight, demographics that were corrected were age, gender, and region, based on the Office for National Statistics (ONS) mid-year population estimates for 2014. The demographic weight was calculated after the selection weights had been applied. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - GREECE (2012) --------------------------------------------------------------------------- <<>> STUDY DESIGN The 2012 Hellenic (Greek) National Election Voter Study was conducted as a mixed-mode survey in the period between October 19, 2012 and January 5, 2013. Fieldwork started 124 days after the second 2012 parliamentary election on June 17, 2017 and lasted for 79 days. The Aristotle University of Thessaloniki, School of Political Sciences, Laboratory of Applied Political Research and To The Point Research Consulting Communication S.A. were responsible for data collection. The questionnaire was administered in Greek. The sample was designed to be representative of Greek voters, that is, Greek citizens with a minimum age of 18 registered to vote (with registration taking place automatically for all citizens born in 1994 or later). Half of the sample was designed to be collected online, complemented by face-to-face interviews. This is because collaborators expected that the 55+ age group would be greatly under-represented in the web survey sample, as the rural population. In contrast, the regions of Attiki (region of the capital Athens) and Thessaloniki were expected to be over-represented. The online sample was selected randomly (using RDD) by area proportional to total population. Thus, households were the primary sampling unit. The selected respondents were called and asked to provide their email address in case they agreed to participate. The face-to-face sample was designed as a cluster sample, with the regions of Attiki and Thessaloniki being excluded because of prior expectations outlined above. In each selected cluster, systematic sampling was used, i.e. the interviewers selected the first house randomly, continuing in the cluster by selecting every kth house. Additionally, a quota on age was employed, with 30% of respondents aged 54 or younger. There are 1029 completed interviews in the sample, of which 500 were realized face-to-face and 529 online. <<>> POLITY WEIGHTS The Greek Election Study provides one combined DEMOGRAPHIC and POLITICAL WEIGHT. It was constructed using the method of ranking, adjusting for gender, age, education, region, and valid votes in the 2012 Greek elections. The weights are combined and thus both the weights listed under variable D1010_2 and D1010_3 for Greece (2012) are identical. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - GREECE (2015) --------------------------------------------------------------------------- <<>> STUDY DESIGN The 2015 Hellenic (Greek) National Election Voter Study was conducted as a mixed-mode survey in the period between June 12 and September 8, 2015. Fieldwork started 138 days after the legislative election on January 25, 2015 and lasted for 89 days. The Aristotle University of Thessaloniki, School of Political Sciences was responsible for data collection. Interviews were conducted in Greek. The sample is meant to be representative of Greek citizens with a minimum age of 18 who were registered to vote (with registration taking place automatically for all citizens born in 1996 or later). Initially, respondents were recruited via telephone (CATI), with households being the primary sampling unit. Therefore, Greek households without a phone were excluded from the sampling frame (about 18 % of the eligible population). The main data collection phase was implemented as a web survey (CAWI). For respondents neither having access to the internet nor an e-mail address, telephone interviews (CATI) were conducted as an auxiliary method. The recruitment of the respondents was based on a random selection of telephone numbers, that is, Random Digital Dialing (RDD). Before the initial selection process, the geographical area was sampled at the first stage, the area code was identified and the telephone numbers were completed during the second stage by random sampling. The recruitment process lasted from June 12 until July 16. The individual who answered the phone call was asked to participate in the survey after verifying that the respondent was eligible to vote.Selected respondents were asked to provide their e-mail address in case they agreed to participate in a web survey conducted by the Aristotle University of Thessaloniki. Every e-mail address collected during the recruitment phase was uploaded to the epolls.gr web survey system and an e-mail with the invitation and the link to the survey was sent to the respondent. Invitations were also sent to about 400 respondents of the 2012 voter study who had indicated that they would be willing to participate in future ELNES surveys. In the following weeks after the initial invitation, reminders were sent to the respondents who had not completed the questionnaire. To increase the response rate of the survey, a maximum of six follow-up reminders was sent via email to the respondents. 1008 respondents completed the interview. <<>> POLITY WEIGHTS The Greek Election Study provides one combined DEMOGRAPHIC and POLITICAL WEIGHT. It was constructed using the method of raking, adjusting for gender, age, education, region and vote share (valid votes). If there were vote share differences between the sample estimates and the official election results prior to adjusting to the share of valid votes, a final round of raking was run including vote share as an additional variable. The weights are combined and thus both the weights listed under variable D1010_2 and D1010_3 for Greece (2015) are identical. A detailed report is available at: Andreadis, Ioannis. Weights for the 2015 Hellenic Voter Study. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 06/08/2016. http://doi.org/10.3886/E74764V1, Date accessed: February 6, 2017. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - HONG KONG (2012) --------------------------------------------------------------------------- <<>> STUDY DESIGN The study was conceptualized as a post-election telephone survey and conducted between September 13 and September 21, 2012, a period of nine days. Fieldwork realized by the Public Governance Programme started four days after the Legislative Council election on September 9. The questionnaire was administered in Chinese (Cantonese and Putonghua). The sample is meant to be representative of the eligible (Chinese- speaking) voting population in Hong Kong, which is defined as permanent residents of Hong Kong who are aged 18 or above and registered to vote. Institutionalized persons, who make up to 0.3% of the population, were excluded from the sample. The same is true for military personnel, who are from mainland China and hence do not have permanent residency. Other possible exclusions could result from an unknown percentage of residents without a fixed-line phone connection. The residential fixed-line penetration rate of 2012, which is 100.6%, was used as a proxy. As some households have more than one fixed line, the number of households without a fixed line is unknown. The total percentage of excluded persons from the sample was estimated to accumulate to less than 1%. Two stages of random sampling were used to generate the sample. In the first stage, samples of phone numbers were selected randomly from the pool of the Hong Kong Telephone Directory. The last two digits of the sample numbers were exchanged with random numbers. In the second stage, individual respondents were randomly selected from the eligible members of the respective households, which were the primary sampling unit. Interviews were conducted via the CATI mode. All interviewers were undergraduate students of the Lingnan University who had received basic training in survey interviews and the CATI system. Before declaring a household a non-sample, five trials of contact at different times and days were made. The average number of contact attempts made per successful interview was 1.69. If a household refused the interview on the first call, a second contact attempt was made to ask for the interview. In total, the dataset contains 1044 completed interviews. <<>> POLITY WEIGHTS The Hong Kong Election Study includes a DEMOGRAPHIC WEIGHT. It was constructed to correct for known divergences of the sample in comparison to the population of registered voters. The weights take the distributions of age, sex, and level of education into account, as indicated by official statistics. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ICELAND (2013) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES Module was included in a post-election telephone study which was fielded between May 4 and September 10, 2013, amounting to a period of 130 days. Data collection was realized by the Social Science Research Institute of the University of Iceland and started one week after the parliamentary election on April 27, 2013. The questionnaire was administered in Icelandic. The sample is meant to be representative of Icelandic voters between 18 and 80 years of age. Eligibility requirements include the age of 18 onwards and Icelandic citizenship. Voters do not need to register themselves before elections. Apart from persons under age or non- citizens, a group of persons listed on a "Do not call" list in the national register was excluded from the sampling frame. This latter group amounts to around 9.9% of the total eligible population. Further, persons without access to a landline or mobile phone could not be sampled. According to the Post and Telecom Administration in Iceland, this pertains to less than 1% of the population. The sample is a simple random sample of voters living in Iceland, which was drawn from the National Register. Therefore, individuals were the primary sampling unit. Every individual had the same probability to be selected. Before the respondents were interviewed by phone, a letter was sent to them. Persons who did not speak Icelandic, were not reachable within the time- limit of the survey, too sick to reply or passed away were designated as non-sample. The average number of necessary attempts to contact a person per household was 5.4 times. In total, the dataset includes 1479 completed interviews. <<>> POLITY WEIGHTS No weights provided. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - IRELAND (2011) --------------------------------------------------------------------------- <<>> STUDY DESIGN The study was conceptualized as a mixed- mode post-election survey and conducted between March 6 and April 10, a period of 36 days. Data collection started nine days after the Irish general election on February 25, 2011. The interviews of the main survey were conducted in English and in person, supplemented by a drop-off self-completion questionnaire. CSES questions were spread through both questionnaires. The Irish sample restricts to Irish citizens aged 18 years and older, excluding institutionalized persons and military personnel based on a military facility (about 3% of the total eligible population). The sampling process involved multiple stages. The primary sampling unit consisted of 320 geographical locations, so-called district electoral divisions (DEDs), which are low-level administrative areas.In these DEDs covering all 43 electoral districts, a non-business address was chosen at random from a geo-directory as the starting point for a random walk. The interviewer was asked to call into every fifth house rather than one consecutive house after another. At each household, the interviewer interviewed the person with the next birthday who fitted into an unfilled quota. This approach was continued until all sampling point-specific quotas were achieved. Quotas were set on gender, age, and social class, making use of the Irish Census lists. In total 1853 people of voting age were interviewed, with 18 to 66 respondents per constituency. The employed interviewers were all professional interviews from the Red C Research & Marketing Ltd, the Irish research center responsible for data collection. Note that the Irish election study was a CSES pilot study, based on an earlier pilot version of the CSES Module 4 questionnaire, and not the final version. Any differences with the final version of the CSES Module 4 questionnaire are spelled out in the Election Study notes for IRELAND (2011). <<>> POLITY WEIGHTS The Irish Election Study includes two different weights: a DEMOGRAPHIC WEIGHT and a POLITICAL WEIGHT. The demographic weight ensures that the sample reflects the Irish population. It is based on age, gender, and social class, according to the Irish Central Statistics Office. The political weight akes the demographic weight into account and ensures that the voting sample matches the outcome of the distribution of votes in the 2011 election. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ISRAEL (2013) --------------------------------------------------------------------------- <<>> STUDY DESIGN The study was conceptualized as a post- election telephone survey which was fielded between February 18 and March 13, 2013, a period of 24 days. Data collection was conducted by the B.I. and Lucille Cohen Institute for Public Opinion Research and started 27 days after the Israeli legislative election on January 22, 2013. The questionnaire was administered in Hebrew, Arabic, and Russian. The study is meant to be representative of the population of citizens of the state of Israel, aged 18 or older. Institutionalized persons were excluded from the sampling frame. The same holds for the 23,6 % of the population without a stationary phone and an unknown proportion of the population with unlisted telephone numbers. All survey respondents were sampled in a three-stage random sampling process. The primary sampling units were statistical areas, stratified by religion, the geographical area, religiosity and the socio-economic level of respondents. All of the statistical areas were sorted by strata, such that the probability of each statistical area to be included in the sample was proportionate to the size of its population. In the next stage of selection, households were sampled in every stratum, based on matching the telephone number database with the statistical area sampled. The list of telephone numbers used for surveys included stationary telephone line owners, numbered in sequential order. The sampling was performed out of this list. In the final stage, each interviewer received a list of households (telephone numbers) to be interviewed. Within the selected household one adult 18 years old or above was interviewed, without any further specification. The resulting sample consists of 1,017 completed interviews. <<>> POLITY WEIGHTS No weights provided. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - JAPAN (2013) --------------------------------------------------------------------------- <<>> STUDY DESIGN The post-election study for CSES Module 4 was carried out between July 22 and August 25, 2013, as a face-to-face survey. Fieldwork lasted for 35 days. Data collection was realized by the Nippon Research Center, starting the day after the Japanese Upper House election on July 21, 2013. The questionnaire was administered in Japanese and translated back into English. The original sample was part of a panel study with a population of registered Japanese voters aged 20 years or older. Additional addresses were added to this sample to enlarge the total sample size. The CSES module is included in the last wave of the four-wave panel survey. Therefore, the CSES survey contains three different types of respondents: a) The respondents from the first wave: n1 = 365. b) The respondents drawn for the second wave from the same PSU: n2 = 535. c) The respondents drawn for the fourth wave from the same PSU: n4 = 1037. The total number of respondents is 1,937: N = n1 + n2 + n4. Please note that no additional sample was drawn for the third wave. The total panel attrition rate between the first wave of the study and the wave that included the CSES Module is 85.1%, as 365 of the 2443 respondents of the first panel wave were successfully re-interviewed in wave four. The initial sample was selected through stratified two-stage random sampling. Electoral district's “chiten” (comparable to precincts in the US) were the primary sampling unit, stratified by region (eleven categories) and city size (five categories). This resulted in an 11 x 5 table with 55 cells. The number of chitens (150 in total) was assigned to each cell proportion- ally to the size of the population. The probabilities of the particular chiten's selection were proportional to the relative size of the chiten. Lists of registered voters compiled by municipalities served as the sampling frame for the second stage of the sampling process. Nippon Research Center fielded 143 interviewers for 150 primary sampling units during the data collection period. Interviewers were mostly female. Before the interview, respondents received a postcard announcing the interview. Additionally, respondents received a voucher worth 1000 Japanese Yen (roughly 10 USD) after the completion of the interview. <<>> POLITY WEIGHTS The Japanese Election Study contains a DEMOGRAPHIC WEIGHT. It adjusts the sample distributions to the Japanese population. To create the weight, age was coded into seven categories (20s, 30s, 40s, 50s, 60s, 70s, and 80s), while the area comprises six categories (Hokkaido, Tohoku, Kanto, Tokai/Hokuriku/Shinetsu, Kinki and Shikoku/Chugoku/Kyushu). --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - KENYA (2013) --------------------------------------------------------------------------- <<>> STUDY DESIGN The post-election study was administered between October 21 and November 28, 2013, as a face-to-face survey. Fieldwork lasted for 39 days. Data collection was realized by the Institute for Development Studies, University of Nairobi, starting seven months and 18 days after the Kenyan general election on March 4, 2013. Interviews were conducted in English, Luo, Kamba, Kikuyu, and Kiswahili. The design is a representative cross-sectional sampling based on a stratified multistage cluster sampling methodology. The sample consists of 1200 respondents, and only Kenyan citizens of voting age (18 and older) were targeted. Inmates and military personnel were excluded from the sample. The primary sampling unit (PSU) was the census enumerator area, the secondary sampling unit (SSU) was the parliamentary constituency. The 210 SSUs were stratified by province and within each province by whether a constituency was predominantly urban or rural. 30 constituencies were selected with probability proportionate to sample size (PPPS), out of which ten were urban and 20 were rural. Within each constituency, Enumeration Areas (EAs) were randomly selected (PPPS). In turn, a fixed number of households were randomly selected from each sampled EA. Furthermore, a quota was employed for gender. <<>> POLITY WEIGHTS The Kenyan Election Study contains a SAMPLE WEIGHT. It makes the sample representative of the population. Of the 30 constituencies selected for the study, some were oversampled to obtain a representative set of answers for the MP questions. Weights were constructed by province and compensate for the disproportionate probability of selection. They are meant to match the population by region and the urban-rural population distribution. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - LATVIA (2011) --------------------------------------------------------------------------- <<>> STUDY DESIGN The study was a post- election face-to-face survey and conducted between October 16 and November 7, 2011. TNS Latvia was responsible for data collection, starting 29 days after the Latvian election on September 17. Interviews were conducted in Latvian and Russian. The sample is meant to be representative of Latvian citizens residing in Latvia, who are between 18 and 74 years old. About five percent of the eligible population were excluded from the sampling frame. The sample can be described as a stratified multistage-quasi-random probability sample. In total there were four stages within the sampling process: 1) the selection of sampling points; 2) the selection of addresses; 3) the selection of households; and 4) the selection of respondents. Primary sampling units were sampling points which were selected system- atically from the complete list of Latvia's populated points, with the number of citizens as a size measure (Probability Proportional to Size). Populated points were sorted by region, district and urbanization level to keep regional and urban proportions in the sample. The planned number of sample points was calculated with the assumption of 10 interviews in each sample point. A random number determined the point in the list of cities or parishes from which the selection began. In the second sampling stage, the random starting address for an interviewer's route was selected within each sampling point. The selection within urban areas was based on lists of dwellings (addresses). In a third stage, households were selected according to route instructions - each second dwelling in apartment buildings, each second home in an area of individual houses; ascending or descending se- quence varying on odd and even sizes of streets. Finally, respondents were sampled from households by "the principle of the youngest man in the household". This principle demands that at each address, the interviewer clarifies who of the household members being at home is the youngest man. If no man is at home, the youngest woman within the household is asked to participate in the interview. If a respondent refused to participate or had not been contacted within two visits, the respondent was not replaced by another household member. After participating in the interview, respondents received a small gift (chocolate or pen). The sample contains 1004 completed interviews. <<>> POLITY WEIGHTS The Latvian Election Study contained a combined SAMPLE and DEMOGRAPHIC WEIGHT. It accounts for unequal selection probabilities. The weight was constructed by taking into age, nationality, region, area of residency, and gender. The weights are combined and thus both the weights listed under variable D1010_2 and D1010_3 for Latvia (2011) are identical. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - LATVIA (2014) --------------------------------------------------------------------------- <<>> STUDY DESIGN The study was conducted as a face-to-face, post-election survey. Data collection, conducted by TNS Latvia, started on November 7, 2014, 34 days after the Latvian parliamentary election. The last interviews were conducted on November 20, 2014. Questionnaires were administered in Latvian and Russian. The sample is representative of the Citizens of the Republic of Latvia, aged 18-74, residing in Latvia. About five percent of the eligible population were excluded from the sampling frame. Respondents were selected in a four- stage quasi-random probability sample. For the first stage, a systematic PPS-sample (Probability Proportional to Size) of sampling points was selected from the complete list of Latvia's populated points, with the number of citizens as a size measure. Populated points were sorted by region, district and urbanization level to keep regional and urban proportions in the sample. The planned number of sample points was calculated with the assumption of 10 interviews in each sample point. A random number determined the point in the list of cities or parishes, from which the selection began. Starting from this point, considering certain steps, cities or parishes, to be included in the sample, were selected. From each sampling point, a random starting address for an interviewer' route was selected, based on lists of dwellings. Households were then selected according to the route instructions – each second dwelling in apartment buildings, each second home in an area of individual houses; ascending or descending sequence varying on odd and even sides of streets. In the last step, the youngest man that was present in the household was selected to be interviewed. If no man was at home during the interview, the youngest woman within the household was asked to participate in the interview. If a respondent refused to participate or had not been contacted within two visits, the respondent was not replaced by another household member. After participating in the interview, respondents received a small gift (chocolate or pen). 1036 respondents completed the interview. <<>> POLITY WEIGHTS The Latvian Election Study contained a combined SAMPLE WEIGHT to get the sample division as close as possible to the universe. Criteria that were used for the data weighting are age, nationality, region, type of residence, and gender. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - MEXICO (2012) --------------------------------------------------------------------------- <<>> STUDY DESIGN This post-election study was conducted face- to-face between July 13 and July 19, 2012, and focused on the 2012 Mexican presidential and legislative elections. CAMPO, S. C. was responsible for data collection, which lasted for seven days and started 12 days after the general election on July 1. The questionnaire was administered in Spanish. The study is representative of the national population aged 18 years or older, excluding those who were institutionalized at the time the survey was being conducted (approx. 0.06% of the total eligible population). The sample was selected by a multistage procedure. The universe was divided into four regions: states with a PAN governor, Northern States with a PRI governor, Center-South states with PRI governors and states with a PRD governor. An independent sample was drawn within each region. In each region, precincts were ordered by the vote for PRI in the last election and divided into four groups of the same number of precincts. Within each group, precincts are clustered by county. This resulted in groups from the same county with a similar vote for PRI, which served as the primary sampling unit. The first selection stage was done with this list, clustering precincts within each group with probability proportional to size (PPS), being turnout the size of the cluster. In the second stage, precincts were selected with PPS. Fieldwork teams received the sample of precincts. In the field, the third stage, blocks were randomly selected in the precinct area. In each block, houses were selected following systematic methods of random start. Within each household, respondents were chosen with the last-birthday method. In case the selected household member could not be interviewed after two contact attempts, interviewers moved on to the third household on the left. The sample consists of 2400 completed interviews. <<>> POLITY WEIGHTS The Mexican Election Study contains two weights, a SAMPLE WEIGHTS and a DEMOGRAPHIC WEIGHT. The sample weight corrects for unequal selection probability of sampling units. The DEMOGRAPHIC WEIGHT corrects for non-response and to match the sample to known population characteristics (gender and age), based on census data from 2010. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - MEXICO (2015) --------------------------------------------------------------------------- <<>> STUDY DESIGN The study was conceptualized as a post- election survey and administered between June 20 and June 28, 2015 employing face-to-face interviews. CAMPO, S. C. was responsible for data collection, which started 13 days after the Mexican legislative election on June 7, 2015. Interviews were conducted in Spanish. The study is representative of the national population aged 18 years or older, excluding those who were institutionalized at the time the survey was being conducted (approx. 0.06% of the total eligible population). The sampling was done in a multistage procedure. The universe was divided into three regions: states with a PAN governor, states with a PRI governor, and states with a PRD governor (stratification). An independent sample was drawn within each region. In each region, precincts were ordered by the vote for PRI in the last election and divided into four groups of the same number of precincts. Within each group, precincts were clustered by county (municipality), resulting in groups with a similar vote for the PRI and located in the same county (municipality). These groups served as the primary sampling unit. The first stage selection was executed on the basis of this list, clustering precincts within each group with probability proportional to size (PPS), being turnout the size of the cluster. In the second stage, precincts were selected with PPS. Fieldwork teams received the sample of precincts. In the third stage of sampling, blocks were randomly selected in the precinct area. In each block, houses were selected following systematic methods of random start. Respondents within the households were randomly selected by listing all household members over 18 years and determining which birthday was closest to the current date. In case the selected household member could not be interviewed after two contact attempts, interviewers moved on to the third household on the left. The sample consists of 1197 completed interviews. <<>> POLITY WEIGHTS The Mexican Election Study contains two weights, a SAMPLE WEIGHTS and a DEMOGRAPHIC WEIGHT. The sample weight corrects for unequal selection probability of sampling units. The DEMOGRAPHIC WEIGHT corrects for non-response and to match the sample to known population characteristics (gender and age), based on census data from 2010. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - MONTENEGRO (2012) --------------------------------------------------------------------------- <<>> STUDY DESIGN The post-election study was conducted as a face-to-face survey, from February to March 2013. Data collection was realized by the De Facto Consultancy, starting around four months after the parliamentary election on October 14, 2012. Interviews were conducted in Montenegrin. The sample is meant to be representative of the eligible Montenegrin population aged 18 years or older. It consists of mainly five ethnic groups, Montenegrins, Serbs, Bosniaks, Albanians, and Muslims. The sampling was conceptualized as a multistage random procedure, with three stages in total: In the first stage, the population was divided into three regions, stratified by the proportion of the population, the distribution of ethnic groups within a region, age groups, and the region's geography, economy and history. These regions were the primary sampling unit. Second, 'mjesna zajednica' (the smallest municipal unit which is at the same time a unit for voting) were chosen randomly within a region. In the third and final stage, a random procedure based on 'step-and-go' was employed inside of 'mjesna zajednica' to choose the household. The last person to have had their birthday within the household was interviewed. 967 respondents completed the interview. <<>> POLITY WEIGHTS The Montenegrin Election Study contains a DEMOGRAPHIC WEIGHT, which is based on census data from 2011. The demographic weight is composed of two separate weights, one adjusting for the four main national groups, the other adjusting for gender and age groups (18-34, 35-54, 55+). The weight included in D1010_2 is a multiplication of the two individual weights. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - NEW ZEALAND (2011) --------------------------------------------------------------------------- <<>> STUDY DESIGN The interviews from New Zealand were conducted within a post-election study, from November 30, 2011, through April 4, 2012. Data collection was realized by the Centre for Methods and Policy Applications in the Social Sciences (COMPASS) and started four days after the general election on November 26, 2011. Respondents were contacted by mail and could respond with a self- administered mail- back questionnaire or with an online version of the questionnaire. Both versions were administered in English. The sample is meant to be representative of persons aged 18 and over on the electoral rolls, which cover 94% of the eligible voters. The primary sampling unit were individuals clustered within age groups (18-26 and 27 and over) and electorates (general and Maori), with over- samples of the young (18-26) and those in Maori electorates. For the CSES release, the over-sampled groups have been re-sampled to bring them in line with the proportions of those groups on the electoral rolls. As an incentive to participate in the study, collaborators conducted a 300 New Zealand dollar draw. A further effort was made to convert refusals. Specifically, respondents not answering the survey received a reminder postcard, a second questionnaire, a final reminder and a thank-you postcard. The CSES sample consists of 1374 completed interviews. <<>> POLITY WEIGHTS The New Zealand Election Study includes two weights, a DEMOGRAPHIC WEIGHT and a combined DEMOGRAPHIC-POLITICAL WEIGHT. The demographic weight adjusts the sample distribution to the real household income distribution (quintiles). Based on this demographic weight, the political weight adjusts for the electoral outcome in terms of both party vote and turnout. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - NEW ZEALAND (2014) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES Module 4 was part of the 2014 New Zealand general election post- election study, which was fielded between September 23, 2014 and February 6, 2015. Data collection was realized by the Centre for Methods and Policy Applications in the Social Sciences (COMPASS) and started three days after the general election on September 20, 2014. Respondents were contacted by mail and could respond with a self-administered mail-back questionnaire or with an online version of the questionnaire. Both versions were administered in English. The sample is representative of all eligible voters (minimum age: 18 years) who were listed on the electoral rolls (approx. 92% of all New Zealan citizens who are eligible to vote are enrolled). Respondents were randomly sampled from the electoral rolls. The primary sampling unit were individuals clustered within age groups (18-26 and 27 and over) and electorates (general and Maori), with over-samples of the young (18-26) and those in Maori electorates. For the CSES release, the oversampled groups have been re- sampled to bring them in line with the proportions of those groups on the electoral rolls. As an incentive to participate in the study, collaborators conducted a 300 New Zealand dollar draw. A further effort was made to convert refusals. Specifically, respondents not answering the survey received a reminder postcard, a second questionnaire, a final reminder and a thank-you postcard. The total N in the sample was 5200, and 1412 interviews were completed. After adjusting for oversampling of the young voters (18-26 years) and the Maori, the remaining sample for CSES has a total N of 1224 respondents. <<>> POLITY WEIGHTS The New Zealand Election Study includes two weights, a DEMOGRAPHIC WEIGHT and a combined DEMOGRAPHIC-POLITICAL WEIGHT. the demographic weight adjusts for age and education to match known characteristics in the population. Based on this demographic weight, the political weight adjusts for party vote and turnout – predominantly weighting up nonvoters, who were less likely to have participated in the survey. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - NORWAY (2013) --------------------------------------------------------------------------- <<>> STUDY DESIGN The study is a post-election, mixed-mode survey, which was fielded between September 10, 2013, and January 30, 2014. Data collection was realized by Statistics Norway, starting the day after the Norwegian parliamentary election on September 8 and 9, 2013. Interviews were conducted in Norwegian, either by phone or face-to-face. Those whose phone number was not listed were interviewed face-to-face. The sample is meant to be representative of the Norwegian citizens aged between 18 and 80. Norwegians living abroad for more than six months were excluded from the sampling frame (1.6 % of the eligible population). The sample was drawn from the public register as a cross-sectional probability sample. It is part of a rolling cross-sectional panel study with about half of the sample being interviewed for the second time. The sample consists of 1727 respondents aged 18-79 who were sampled in a two-stage design by Statistics Norway. The sampling frame is based on registers covering all Norwegian inhabitants. In the first stage, the country was divided into 363 primary sampling units (psus), based on the local munici- palities. All psus with more than 30,000 inhabitants and some with a population number between 25,000 and 30,000 constitute separate strata. For the remaining psus, strata were formed as homogeneously as possible. Stratification variables were industrial structure, number of inhabitants, centrality, communication structures, commuting patterns, trade areas and (local) media coverage. The stratification was done separately for each county and in such a way that no psu had less than 7% of the total population in its stratum. Thus, it is possible to give unbiased estimates for each county. After stratification, one psu was selected from each stratum. Those psus which constituted separate strata were included with certainty. For the remaining strata, one unit was drawn with probability proportional to the number of inhabitants. In the second stage of the sampling, survey units (individuals) were drawn from the population register, selected from the 109 sampling areas, using systematic random sampling. The sampling fraction at the second stage is proportional to the inverse selection probability at the first stage. The final sample then is self-weighting when both stages are taken into consideration. If a sampled unit from the panel sample had moved out of their original psu, it was still included in the sample. Before the interview, respondents received a letter announcing the study. <<>> POLITY WEIGHTS The Norwegian Election Study contains a DEMOGRAPHIC WEIGHT. The weight was constructed after collaborators found non-response in the survey was related to particular demographic chracterisitics. The weight adjusts for individual turnout (controlled against electoral roll), gender, age, and education, dividing the sample into 36 different strata. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - PERU (2016) --------------------------------------------------------------------------- <<>> STUDY DESIGN The survey employed face-to-face interviews and was conducted between May 7 and May 17, 2016, between the rounds of the Peruvian presidential election. The Instituto de Opinion Publica de la Pontificia Universidad Catolica del Peru was responsable for data collection, starting 27 days after the first election round on April 10, 2016. Interviews were conducted in Spanish. The sample is meant to be representative of registered Peruvian voters with a minimum age of 18. The 35 provinces of 19 regions of the country where the interviews were done covered approximately 67.7% of the national population, such that 32.3 % of the total eligible population were excluded from the sampling frame. The same holds for institutionalized persons (0.3%), which were not interviewed either. The sample was selected by a multi-stage stratified sampling procedure with stratification according to geographical units. Clusters representing either housing blocks in cities or towns in rural areas were defined as primary sampling units. The random selection at the first stage was executed with probabilities proportional to cluster size. At the second stage, households were randomly selected from the drawn clusters using systematic sampling. Within households, individuals were chosen according to sex and age quotas with a maximum of one individual being interviewed per household. Households were contacted once, and non- residential sample points, vacant households, or households in which all members were ineligible were declared non- samples. No non-sample replacement methods were used. However, in cases of refusal in a selected household, replacement was performed in the adjacent houses in the same housing block. The sample contains 1572 completed interviews. <<>> POLITY WEIGHTS The Peru Election Study contains two weights, a SAMPLE WEIGHT and a POLITICAL WEIGHT. Both weights account for unequal probabilities of selection. The first combines these design weights with demographic weights based on the population distribution of age and gender. The second weighting variable combines the design weights with political weights based on the results of the first presidential election round. The political weights are absent for 123 cases: 105 respondents who did not cast a vote in the presidential elections (D3005_PR_1) and 18 respondents who did not specify their vote choice in the presidential elections (D3006_PR_1). These cases were coded 0 on the political weighting variable and will thus be dropped from the analysis if the weight variable is applied to analysis. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - PHILIPPINES (2016) --------------------------------------------------------------------------- <<>> STUDY DESIGN The study was conducted with face-to-face interviews which were held between June 24 and June 27, 2016. The Social Weather Stations were responsible for data collection, which started 46 days after the Philippine presidential and legislative elections on May 9, 2016. The survey was fielded in seven languages (Filipino, Iluko, Hiligaynon, Cebuano, Bicol, Waray and Maranao) and is meant to be representative of Philippine citizens aged 18 years and older. Multi-stage probability sampling was used to select sample spots. In a first step, the country was divided into four study areas: The National Capital Region (NCR), Balance Luzon, Visayas, and Mindanao. From all areas except the National Capital region, 21 provinces were selected altogether, with probability proportional to size within each region. In a second stage, 140 sample municipalities were identified within the sample provinces and the National Capital Region. From these, 60 sample spots, the primary sampling unit (Barangays), were selected within each of the four regions, with probability proportional to size (PPS). In the next stage, in each of the sample spots, five sample households were chosen by systematic sampling. Designated starting points were randomly assigned - it was either: 1) a municipal/barangay hall, 2.) a school, 3.) the barangay captain's house, 4.) a church/chapel/mosque, 5.) a health facility, or 6.) a basket- ball court. From there, a random start from the first to the 6th household was defined. Thus, if a particular spot had a random start of 4, the first sample household should be the 4th household from the designated starting point. Subsequent sample households were chosen using a fixed interval of 5 households in between the sampled ones; i.e. every 6th household was sampled. In the last stage of the sampling, a respondent was randomly chosen among the household members who were 18 years of age or older, using a probability selection table. The sample contains 1200 completed interviews. <<>> POLITY WEIGHTS The Philippines Election Study includes a SAMPLE WEIGHT. To yield representative figures at the national level, a census-based population weight was constructed for the survey data. Since the sample is equally allocated across the four major areas (National Capital Region or NCR, the rest of Luzon outside NCR, Visayas and Mindanao), weights are used to adjust to known area population distribution. The weight projection is computed by dividing the projected population in the area by the sample size of the same area. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - POLAND (2011) --------------------------------------------------------------------------- <<>> STUDY DESIGN This post-electoral study was carried out via face-to-face interviews between October 20 and November 13, 2011. Fieldwork, which was conducted by the Public Opinion Research Center (Centrum Badania Opinii Spolecznej, CBOS), started 11 days after the Polish parliamentary election on October 9, 2012. The questionnaire was administrated in Polish. The study collects information from voters (Polish citizens aged 18 years or over, registered and eligible to vote). Eligible citizens are automatically registered as voters. The sample can be characterized as a stratified multistage probability sample, with municipalities as the primary sampling unit. In the first stage, 81 strata were selected based on 16 regions (voivodeship) and the size of the municipalities (six categories). Strata were obtained as an intersection of region and municipality size. As not every type of municipality occurs within each region, the number of strata was 81 instead of 96. In a second stage, municipalities were sampled in each stratum according to Hartley-Rao scheme. Next, for each selected municipality one street (or village in case of rural areas) was drawn. In each stratum, at least two streets or villages were sampled. Finally, for each street (or village) eight respondents were drawn based on their National Identification Number. The sample consists of 1919 completed interviews. <<>> POLITY WEIGHTS The Polish Election Study contains one DEMOGRAPHIC WEIGHT, adjusting the sample to population estimates for gender, age, level of education and size of residence. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - PORTUGAL (2015) --------------------------------------------------------------------------- <<>> STUDY DESIGN The post-election study was administered between October 17 and December 9, 2015, as a face-to-face survey. GfK Portugal - Metris was responsible for fieldwork, which started 13 days after the Portuguese legislative election on October 4, 2015. Interviews were conducted in Portuguese. The sample is meant to be representative of individuals aged 18 years or older, who were living in private households in continental Portugal at the time the survey was being conducted. Portuguese citizenship was not required to participate in the interview. The autonomous regions (archipelagos) of Madeira and Azores were excluded from the sample frame (4,87% of the eligible population). The same holds for institutionalized persons (about 0,13%). The sample was stratified according to NUTS and habitat. As the autonomous regions of Madeira and Azores were not included in the sample, stratification of primary sampling units (urban areas and rural districts) was done in the five first-level NUTS existing in Continental Portugal: Lisboa (the capital), Norte, Centro, Alentejo, and Algarve. The random selection of towns was carried out in accordance with the places included in the General Population Census, considering the matrix sampling of NUTS and habitat. In each designated town, sample points were selected randomly (105 in total). Afterward, households were selected through the random-route method. Finally, in each household, the person who had the most recent birthday was selected for the interview. Substitution of individuals was permitted. If the selected person was not available, three more visits were made in an attempt to reach the potential interviewee. After a fourth unsuccessful visit, the selected individual was replaced by another person with similar characteristics (sex, age group and habitat) from a different household. Prior to the interview, a letter was sent to respondents. The sample consists of 1499 completed interviews. <<>> POLITY WEIGHTS The Portuguese Election Study includes a DEMOGRAPHIC WEIGHT. It is designed to match known demographic characteristics of the population and incorporates region, gender, age, and education. To correct for non-response individuals in under-represented groups got a weight larger than 1, while those in over-represented groups got a weight smaller than 1. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ROMANIA (2012) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES Module 4 was implemented in the 2012 Romania Post-Election study and consists of two components, a cross-sectional component and a post- election survey from a two-wave Pre-/Post-Election panel. Interviews were carried out in the aftermath of the lower house elections, which took place on December 9, 2012. Fieldwork, which was conducted by KANTAR TNS (CSOP), started six days after the election and lasted from December 15, 2012, to January 30, 2013. Interviews were conducted face-to-face in Romanian. This sample is meant to be representative of the non-institutionalized eligible Romanian population, i.e. Romanian citizens of at least 18 years. As the sampling frame is based on Romanian's electoral register, only registered voters were sampled. Although registration happens automatically, the register only covers persons listed by local authorities. Therefore, unlisted persons were excluded, as were institutionalized persons (amounting to around 0,5% of the total eligible population). The sample can be described as a stratified two-stage probability sample. Stratification was based on 18 cultural areas and seven types of municipalities (poor communes, medium developed communes, developed communes, cities with less than 30,000 inhabitants, cities of 30-100,000 inhabitants, cities of 100-200,000 inhabitants, cities with more than 200,000 inhabitants), resulting in 96 different strata. Voting districts from the Local Elections in 2012 were used as the primary sampling unit (PSU). The number of PSUs to be selected within a stratum was obtained by multiplying the sample size n with the share pi of each stratum in the electorate (i.e. proportional to the number of registered voters within a stratum). Then, ten respondents were chosen randomly from the electoral register of each selected voting district, resulting in the main sample. Another ten respondents were selected for a reserve list to compensate for refusals, using the same procedure. The final sample included 1,248 households for the panel component (1,080 completed interviews) and 1,763 for the cross-sectional component (1,203 completed interviews). The panel attrition rate between the pre- and the post-study of the panel component was 13.5%. Respondents of the panel-component were offered a 10 RON (about 3 USD) food coupon after participation, or in case they were reluctant to participate. Overall, the 2012 Romanian sample includes 2,283 respondents (20% of the interviews were verified in the aftermath of the study). <<>> POLITY WEIGHTS The 2012 Romanian Election Study includes three weights: a SAMPLE WEIGHT, a DEMOGRAPHIC WEIGHT, and a POLITICAL WEIGHT. The Sample Weight (D1010_1) corrects for the variance in the probability of being selected into the sample, an issue that arose from different poll sizes. The demographic weight (D1010_2) is designed to match known characteristics in the population (gender and age) and further accounts for residence (urban/rural) and region (Transylvania, Muntenia, Moldavia, and Bucharest). The political weight (D1010_3) corrects for discrepancies between the reported vote choice and the actual election results. The demographic and the political weight are cumulative weights, that is, the demographic weight also takes the sample weight into account and the political weight takes both the sample and the demographic weight into account. The weights were created separately for each of the two sample components (see D1007). Therefore, they can be applied when working with the full sample. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ROMANIA (2014) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES Module 4 was implemented in the 2014 Romania Post-Election study. Interviews were carried out in the aftermath of the second round of the Presidential Elections which took place on November 16, 2014 (Round one took place on November 2, 2014). Fieldwork conducted by KANTAR TNS (CSOP) started four days after the second round on November 20, 2014, and ended on December 8, 2014. Interviews were conducted in Romanian. This sample is meant to be representative of the non-institutionalized eligible Romanian population, i.e. Romanian citizens of at least 18 years. As the sampling frame is based on Romanian's electoral register, only registered voters were sampled. Although registration happens automatically, the register only covers persons listed by local authorities. Therefore, unlisted persons were excluded, as were institutionalized persons (amounting to around 0,5% of the total eligible population). The sample can be described as a stratified two-stage probability sample, with stratification of the primary sampling units (voting districts for the 2014 presidential elections) proportional to the number of secondary units (adults registered in the electoral register). Specifically, voting districts were stratified with regard to 8 regions of development and 7 types of municipalities (rural areas under 2000 inhabitants, with 2000 to 4000 inhabitants and more than 4000 inhabitants; as well as cities under 30,000 inhabitants, cities with 30,000 to 100,000 inhabitants, cities with a population of 100,000 to 200,000 people, and cities with more than 200,000 inhabitants). This way, 53 strata were created. Voting districts were selected from each stratum, proportional to their number of registered voters. In the second stage, ten respondents were selected by systematic random sampling from the electoral register of each selected voting district, resulting in the main sample. A reserve sample of ten respondents was selected to compensate for refusals, using the same procedure. The final sample includes 1112 completed interviews. 20% of the interviews were verified in the aftermath of the study. <<>> POLITY WEIGHTS The 2014 Romanian Election Study includes three weights: a SAMPLE WEIGHT, a DEMOGRAPHIC WEIGHT, and a POLITICAL WEIGHT. The Sample Weight (D1010_1) corrects for the variance in selection probabilities of individual respondents. The demographic weight (D1010_2) is designed to adjust the sample to known characteristics in the population (gender and age) and further accounts for residence (urban/rural) and region. Finally, the political weight (D1010_3) corrects for discrepancies between the reported vote choice and the actual election results for both Presidential Election Rounds simultaneously. The demographic and the political weight are cumulative weights, that is, the demographic weight also takes the sample weight into account and the political weight includes both the sample and the demographic weight. All weights were created by dividing the known population distributions by the sample statistics. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SERBIA (2012) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES Module 4 was included in the Serbian Post-Election Study, running from December 21, 2012, until February 10, 2013, using face-to-face, computer-assisted interviews. Ipsos Strategic Marketing conducted fieldwork, starting 229 days (i.e. 7,5 months) after the Serbian parliamentary election on May 6, 2012, which was held simultaneously with the first round of the presidential election. Interviews were conducted in Serbian, but the questionnaires were also prepared in Hungarian and Albanian. The Serbia Public Opinion Study is designed to be a representative national sample of eligible voters in Serbia (i.e. adult citizens aged 18 on election day older). Institutionalized persons and military persons were excluded from the sampling frame, as were the regions of Kosovo and Metohija (about 8% of the total eligible population). The sample is address-based, using the database provided by the national post office of Serbia (Posta Srbije). For mail delivery, Posta Srbije divides the country into geographic regions named Reons. At the first sampling stage, the project team randomly selected a set of 150 lines from the database, i.e. intersections of settlements and postal Reons. These lines were the primary sampling unit. For each Line, the Posta Srbije was asked to provide the addresses from odd-numbered PAKs (Postal Address Codes) for residential households. In addition to the addresses, the Posta Srbije also provided the number of households per address. This allowed the project team to convert the list of addresses to a list of households. From among the list of households, 3300 households were randomly selected with equal probability, aiming for 1800 completed interviews. If more than one household member was found to be eligible, the respondent was randomly selected from within the respective household. After completing the interview, respondents received a package of coffee as a thank-you gift. Furthermore, for respondents who were reluctant to participate, interviewers left a letter at the household. Additionally, they were offered another package of coffee. The sample contains 1568 completed interviews. <<>> POLITY WEIGHTS The Serbian Election Study includes two weights, a SAMPLE WEIGHT and a DEMOGRAPHIC WEIGHT. The sample weight compensates for differing numbers of eligible persons in the household and an oversampling of low-response areas. The demographic weight adjusts the sample to the known urban-rural population distribution (based on 2011 census data). --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SLOVAKIA (2016) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES Module 4 was implemented in the 2016 Slovak Post-Election study that was carried out in the aftermath of the lower house elections that took place on March 5, 2016. TNS Slovakia s.r.o. was responsible for data collection, which started 222 days (i.e. 7 months and 8 days) after the election. The survey was administered between October 13 and November 28, 2016. Interviews were conducted face-to-face and in Slovak. This sample is meant to be representative of the eligible Slovak population, i.e. citizens of at least 18 years who have permanent residence in Slovakia. The sample was randomly drawn in two stages. First, territorial units, the primary sampling units, were randomly drawn from the Slovak register containing all communities (i.e., villages and cities). Selection was based on the proportion of eligible voters within a community and community size. In total, 220 territorial units were drawn. Households for interviews within these territories were then selected by random walk. Three to eight interviews were carried out in each of the selected units. Of all eligible persons within the household, the person with the closest birthday was selected for the interview. The final sample includes 1150 completed interviews. 10% of the interviews were verified by calling respondents back and asking them about their characteristics and control questions that were included in the questionnaire. <<>> POLITY WEIGHTS The Slovakian Election Study contains a DEMOGRAPHIC WEIGHT that corrects for characteristics of the target population (age, gender, education level, size of resident's community, and region). The weight can also be helpful in reducing discrepancies in the reported vote choice, particularly for the Direction-Social Democracy (Smer) party whose electorate is overrepresented in the sample. For more information, see Election Study Note on D3006_LH_PL in the Variable Section of the Codebook. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SLOVENIA (2011) --------------------------------------------------------------------------- <<>> STUDY DESIGN This post-electoral study was carried out via face-to-face interviews between March 29 and May 28, 2012. Fieldwork, which was realized by CJMMK (Public Opinion and Mass Communication Research Centre), started 116 days (i.e., 3 months, 25 days) after the Slovenian parliamentary election on December 4, 2011. The questionnaire was administrated in Slovenian. The study's sampling frame corresponds to the Central Register of Population (CRP) and includes all residents older than 18 years with a permanent address, including both citizens and non-citizens. Institutionalized people were excluded from the sampling frame (less than 1% of the total eligible population). The study employed a two-stage stratified random sample from the Central Register of Population, where every population unit has an equal selection probability. So-called clusters of enumeration areas (CEAs) were the primary sampling unit (PSU). In the first stage, 120 PSUs were selected with probability proportional to the size of the respective CEA. The CEAs were stratified according to 12 regions x 6 types of settlements. At the second stage, 15 people were selected (120x15) by systematic random selection inside each CEA. Persons selected from the Central Population Register were identified by name and address. No efforts were made to persuade respondents who were reluctant to be interviewed. However, respondents received a letter prior to the study. The sample consists of 1031 completed interviews. <<>> POLITY WEIGHTS No weights provided. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SOUTH AFRICA (2014) --------------------------------------------------------------------------- <<>> STUDY DESIGN The study was conceptualized as a post-election face-to-face survey and conducted between February 2 and February 26, 2015. Fieldwork was realized by Citizen Surveys and started 271 days (i.e., 8 months, 26 days) after the South African general election on May 7, 2014. Interviews were conducted in eleven different languages (English, Afrikaans, Ndebele, IsiXhosa/Xhosa, SePedi/North Sotho, SeSotho/South Sotho, SetTswana/Tswana, Shangaan, Swazi, Venda, and IsiZulu/Zulu). The sample is meant to be representative of adult South African citizens, aged 18 or above. Institutionalized persons and military personnel were excluded from the sampling frame (about 0,43% of the total eligible population). Within each EA, interviewers were sent to a random starting point which had been marked on a map of the EA. From this starting point, interviewers turned into a designated direction, selecting every 10th household by considering dwellings on both sides of the street. Once the household was chosen, the interviewer had to randomly select the individual respondent. A quota was employed for gender, such that every other respondent had to be female. For the random selection of respondents within the household, numbered cards were assigned to each member. Afterwards, the head of the household was asked to randomly select one of these. In case a household was vacant, all members of a household were ineligible, the selected person did not answer after two callbacks or refused to participate, interviewers were instructed to move to the next house in the walking pattern. Substitution within a household was not permitted. The sample includes 1300 completed interviews. <<>> POLITY WEIGHTS The South African Election Study includes a DEMOGRAPHIC WEIGHT based on the 2011 census data. The weight adjusts the realized to the original sample, by correcting for household size, age and oversampling of the colored and Indian population and respondents in the Northern Cape province. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SOUTH KOREA (2012) --------------------------------------------------------------------------- <<>> STUDY DESIGN This post-election study consists of face-to-face interviews, which were conducted from April 12 to April 29, 2012. Fieldwork started the day after the South Korean legislative election on April 11, 2012, and was realized by the Korean Social Science Data Center. Interviews were conducted in Korean. The sample is meant to be representative of registered South Korean voters aged 19 and above. People living on Cheju Island and other small islands were excluded from the sample frame. This is common for Korean national surveys, as people living in these regions are difficult to contact and make up less than 0.5% of the population. A multistage sampling procedure was used in which electoral districts represented the primary sampling units. 80 electoral districts were randomly selected from a list of all electoral districts. At the second stage, voting districts within each sampled electoral district were randomly sampled. In a final step, a simple random sample of voters could be drawn within each voting district, based on the voters' list provided by the Central Election Management Committee. Prior to the study, a token gift was sent to the sampled individuals. The sample includes 1000 completed interviews. <<>> POLITY WEIGHTS No weights provided. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SWEDEN (2014) --------------------------------------------------------------------------- <<>> STUDY DESIGN This post-election study was administered as a face-to-face survey between September 15 and November 17, 2014. Statistics Sweden, SCB was responsible for fieldwork, which started the day after the Swedish general election on September 14, 2014. Interviews were conducted in Swedish. The study is conceptualized as a two-wave rolling panel design, meaning that respondents are contacted in two successive elections. Thus, while half of the respondents were already interviewed in 2010, the other half was newly sampled in 2014. This latter half of respondents were randomly drawn from a list of all Swedish voters who were eligible to cast their vote in the 2014 Swedish National Election. Therefore, individuals were the primary sampling unit. Eligible Swedish citizens living abroad were excluded from the sample frame (about 2% of the population). The sample is meant to be representative of Swedish citizens living in Sweden who were entitled to vote and between 18 and 80 years old. However, due to the two-wave panel design, the 2014 sample includes some respondents sampled in 2010 who are older than 80. Overall, the sample is composed of 832 completed interviews from respondents of ages 18 to 84. To facilitate participation, chosen respondents were sent letters to inform them about the purpose of the study. <<>> POLITY WEIGHTS No weights provided. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SWITZERLAND (2011) --------------------------------------------------------------------------- <<>> STUDY DESIGN The Swiss interviews were held between October 24 and November 24, 2011, by telephone, in French, German or Italian. Fieldwork was conducted by DemoSCOPE RESEARCH & MARKETING and started the day after the Swiss federal election on October 23, 2011. Telephone interviews were supplemented with a self-administrated mail-back and web survey component, which was sent to a considerable proportion of respondents of the first part of the study. The fielding period for the self-administered interview parts ended on December 12, 2011. The sample is meant to be representative of all Swiss citizens residing in Switzerland who were at least 18 years old at the beginning of October 2011. The Swiss Federal Statistical Office randomly drew the sample, based on a sampling frame that includes all individuals (not households) resident in Switzerland. Thus, individuals were the primary sampling unit. While the sampling frame provides information on respondents' address, marital status, age and sex, telephone numbers were only listed for about two-thirds of the selected addresses. To obtain telephone numbers for the remaining individuals, collaborators considered other databases and searched for them manually. Additionally, a 20CHF incentive was sent to those individuals for whom a phone number could not be found, asking them to provide a phone number where they could be reached. Nevertheless, for 15% of the initially sampled persons, no phone number could be matched to the sampling frame. Therefore, these persons could not be interviewed. This number is close to other estimates of households without a registered landline and/ or mobile phone. Furthermore, Swiss voters living abroad were excluded from the sampling frame (another 2,6% of the total eligible population). The sample size was stratified to include a minimum of 100 respondents per canton in each of the 26 cantons and to have at least 600 respondents in the cantons of Zurich, Geneva, and Tessin. Prior to the study, a letter was sent to sampled individuals. 4391 persons were successfully interviewed. <<>> POLITY WEIGHTS The Swiss Election Study contains a SAMPLE WEIGHT and a POLITICAL WEIGHT. The sample weight controls for an oversampling of persons in the smaller as well as the three largest cantons. The political weight controls for the overall turnout and the vote share of parties, multiplied by the sample weight mentioned above. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - TAIWAN (2012) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES Module was included in a post-election face-to-face survey which was fielded between January 15 and March 6, 2012. Data collection was realized by the Department of Political Science, National Taiwan University and started the day after the Taiwan legislative and presidential elections on January 14, 2012. Interviews were conducted in Chinese. The sample is meant to be representative of qualified voters in Taiwan, i.e. Taiwanese citizens above 20 years. Institutionalized persons were excluded from the sample frame. The sampling of the other respondents was realized by a probability proportional to size (PPS) three-stage systematic sampling. In the first stage, legislative constituencies were sampled according to six major geographic regions. In the second stage, urban villages (li) or rural villages (cun) were selected as sampling units. Finally, respondents were selected from household registration data provided by the Ministry of the Interior. Prior to the study, respondents received a letter announcing the survey. Additionally, they received a gift after completion of the interview. A total of 1,826 interviews were completed. <<>> POLITY WEIGHTS The Taiwanese Election Study includes a DEMOGRAPHIC WEIGHT. To generalize to the population of eligible voters in the 2012 presidential election, the data should be weighted using the demographic weight provided, which accounts for gender, age (5 groups), education (5 groups), and area (6 regions). The method of raking was used to create the weight. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - THAILAND (2011) --------------------------------------------------------------------------- <<>> STUDY DESIGN The study was conceptualized as a post-election face-to-face survey and conducted between July 3 and September 10, 2011. Fieldwork was realized by the King Prajadhipok's Institute and started on July 3, 2011, the day of the Thai general election 2011. Interviews were conducted in Thai, Malay and Tribe language such as Lahu. The multi-stage systematic random sample is meant to be representative of the eligible voting population of Thailand, i.e. citizens aged 18 years or above who are registered to vote. For this survey, the country was divided into five regions corresponding to the four regions of Thailand plus Bangkok. For the first stage of sampling, a list of Amphoe (districts) per region was randomly selected. These districts were the primary sampling units. In the second stage, a number of Tambol (subdistricts) were randomly selected within each Amphoe. The third stage consisted in randomly selecting a number of villages within each Tambol. The fourth stage of sampling determined the number of people to be surveyed within a village. In each stage, the numbers of units to be sampled within a geographic area were all chosen proportional to the size of the population. In the final stage, individual respondents were systematically sampled with information provided by the Community Department of the Ministry of Interior. In Bangkok, a list of sub-districts was randomly selected in the first stage. Afterward, systematic sampling was applied to select individual respondents from voter lists, which were available for the Bangkok region. After the interview, respondents received a book and a pen to thank them for participation. The sample includes 1500 completed interviews. <<>> POLITY WEIGHTS The Thai Election Study includes a SAMPLE WEIGHT, which was constructed to weight the data according to the distribution of the population within the five regions, which were also used for the first sampling stage (see above). --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - TURKEY (2015) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES Module was included in a post-election face-to-face survey which was implemented between July 18 and September 10, 2015. Data collection was realized by Frekans Arastirma and started 41 days after the Turkish general election on June 7, 2015. Interviews were conducted in Turkish. The sample is meant to be representative of the voting age population of Turkey living in urban as well as rural areas – that is, Turkish citizens aged 18 years or older. Institutionalized persons and military personnel were excluded from the sample frame (about 1,59% of the total eligible population). The sampling procedure started with the use of Turkish Statistical Institute's (TUIK) NUTS-2 regions. The target sample was stratified according to each region's share of urban and rural population given by the Address-Based Population Registration System (ADNKS) records as of the end of 2014. Next, TUIK's block data were used with block size set at 400 residents. Address blocks were the primary sampling unit. Twenty voters were targeted to be reached within each block, and no substitution was allowed. The probability proportionate to population size (PPPS) principle was used in distributing the blocks to NUT1 regions. For each of the 20 addresses, up to three visits were carried out with the expectation that approximately 50% of the addresses would result in a completed interview. In rural areas where TUIK was unable to provide addresses, collaborators contacted the village's headman (muhtar) and selected 20 addresses in a systematic random sample from the list of households in the village. To select an individual within the sampled household, names of all reported individuals of 18 years or older were written on cards. Then, a person in the household was asked to randomly select one card with the name of the individual to be interviewed. If for any reason that individual could not participate in the survey at the first visit, the same household was visited up to three times to obtain the interview. If the interview could not be conducted after three trials, the respective household was dropped from the sample without being substituted. However, respondents who were reluctant to be interviewed were shown a letter containing information on the collaborators, the nature of the project and collaborators' privacy commitment. Overall, 1086 interviews were completed. <<>> POLITY WEIGHTS The Turkish Election Study includes a DEMOGRAPHIC WEIGHT. The weight adjusts the sample to known population distributions on gender, six age groups (18-24, 25-34, 35-44, 45-54, 55-64, and 65+), and five education levels. Weights to correct deviations from these 60 cells were multiplied with 12 separately calculated regional weights to correct regional deviations in the sample. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - UNITED STATES (2012) --------------------------------------------------------------------------- <<>> STUDY DESIGN The American National Election Study (ANES) 2012 Time Series has a pre-post-election panel design in which respondents were interviewed twice: once pre-election and once in a post-election survey. The study features a dual-mode design combining the traditional ANES face-to-face interviewing with a separate sample interviewed on the Internet. Although the CSES was included in both modes, the sample provided here only consists of face-to-face interviews, due to the low-response rate in the online component. Post-election face-to-face interviews featuring the CSES module were administered between November 7, 2012, and January 13, 2013, starting the day after the U.S. presidential election on November 6, 2012. Data was collected by Abt SRBI. The questionnaires were administered in English and Spanish. The sample is meant to be representative of U.S. citizens aged 18 or older as of election day. Institutionalized persons, military personnel residing in on-base housing, and the states of Alaska and Hawaii were excluded from the sampling frame. The same holds for addresses where mail is not delivered, as the sampling frame consisted of the Delivery Sequence File (DSF) used by the United States Postal Service for the residential delivery of mail. However, collaborators made further efforts to compensate this. For more information, please consult the design report. An address-based, stratified, multi-stage cluster sample in 125 census tracts was used. In addition to a nationally-representative main sample, there were two oversamples, one of blacks and one of Hispanics. The first stage of the sampling involved stratifying the 48 contiguous states and the District of Columbia into nine regions corresponding to Census Divisions. These Census divisions constitute the study's strata. Within each region, census tracts were then randomly selected, with the number of tracts being proportional to the region's proportion of the U.S. adult population (probability proportional to size). The second stage of sampling consisted of the random selection of residential addresses from the DSF within each tract. Addresses for the black and Hispanic oversamples were selected from tracts with high proportions of one or both of these populations. The third and final stage of sampling was the selection of one eligible person per household. Screening was accomplished by sending an interviewer to the sampled address and completing a brief interview in which eligible respondents were listed. The interviewer's computer then randomly selected one person from the list. No substitutions were acceptable. Prior to the interview, respondents received an advance letter. Furthermore, face-to-face respondents were offered money to complete the interview. Depending on the level of pre-election incentivization, respondents were initially offered $50 or $100. These amounts were increased to $75 and $125, respectively, at the end of the field period. Additionally, eligible respondents reluctant to participate were sent a customized letter addressing their specific concern. For the face-to-face component, panel attrition rate was 7%. The sample consists of 1929 completed interviews. <<>> POLITY WEIGHTS The American National Election Study includes a combined DEMOGRAPHIC AND SAMPLE WEIGHT. The weight is post-stratified to produce estimates that match known population proportions for age, race/ethnicity, educational attainment, martial status, income, census region, home ownership, nation of birth and cross-classifications of age and gender, and of race/ethnicity and educational attainment. /// END OF FILE