=========================================================================== COMPARATIVE STUDY OF ELECTORAL SYSTEMS (CSES) - MODULE 5 (2016-2021) CODEBOOK PART 6: STUDY DESIGNS AND WEIGHTS SECOND ADVANCE RELEASE - MAY 14, 2020 CSES Secretariat www.cses.org =========================================================================== HOW TO CITE THE STUDY: The Comparative Study of Electoral Systems (www.cses.org). CSES MODULE 5 SECOND ADVANCE RELEASE [dataset and documentation]. MAY 14, 2020 version. doi:10.7804/cses.module5.2020-05-14 These materials are based on work supported by the American National Science Foundation (www.nsf.gov) under grant numbers SES-1420973 and SES-1760058, 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 ADVANCE RELEASES ))) OVERVIEW OF "CODEBOOK PART 6: STUDY DESIGNS AND WEIGHTS" ))) HOW TO NAVIGATE THE CSES MODULE 5 CODEBOOK ))) LIST OF TABLES IN CODEBOOK PART 6 ))) OVERVIEW OF STUDY DESIGN & WEIGHTS BY POLITY & ELECTION YEAR >>> POLITY WEIGHTS IN CSES >>> TYPE OF POLITY WEIGHTS BY ELECTION STUDY >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - AUSTRALIA (2019) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - AUSTRIA (2017) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - BRAZIL (2018) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - CHILE (2017) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - FRANCE (2017) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - GERMANY (2017) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - GREECE (2015) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - HONG KONG (2016) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - HUNGARY (2018) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - ICELAND (2016) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - ICELAND (2017) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - IRELAND (2016) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - ITALY (2018) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - LITHUANIA (2016) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - MONTENEGRO (2016) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - NEW ZEALAND (2017) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - NORWAY (2017) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - SOUTH KOREA (2016) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - TAIWAN (2016) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - TURKEY (2018) >>> OVERVIEW OF STUDY DESIGN & WEIGHTS - UNITED STATES (2016) =========================================================================== ))) IMPORTANT NOTE REGARDING ADVANCE RELEASE =========================================================================== This dataset and all accompanying documentation is the "Second Advance Release" of CSES Module 5 (2016-2021). By definition, an Advance Release is a preliminary version of a dataset, and thus lacks some of the checking, cleaning, processing, documentation, data, and variables that are usual to the Full Release of a dataset. Many election studies that will eventually be present in the CSES Module 5 Full Release are not available in this file. Advance Releases are provided as a service to the CSES user community, for those analysts who find it valuable to work with preliminary versions of the dataset. We would appreciate being notified of any errors in the dataset or documentation by email to "cses@umich.edu". Users should expect future changes and improvements to the naming, data, and documentation of variables and election studies that appear in an Advance Release file. If users wish to re-use their programming code on a future release of the file, the code should be written in a way that is flexible and can be accommodating of these future changes. Users of the Advance Release may also wish to monitor the errata for CSES Module 5 on the CSES website, to check for known errors which may impact their analyses. To view errata for CSES Module 5, go to Data Download on the CSES website, navigate to the CSES Module 5 download page, and click on the Errata link in the white box to the right of the page. We hope that until such time as the Full Release of CSES Module 5 is available, users will find this and future CSES Module 5 Advance Releases to be helpful in their work. =========================================================================== ))) 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 5 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 5 Study Page (see https://cses.org/data-download/cses-module-5-2016-2021/). =========================================================================== ))) HOW TO NAVIGATE THE CSES MODULE 5 CODEBOOK =========================================================================== In the CSES Module 5 dataset, all variables begin with the letter "E" (E being the fifth letter of the English alphabet and thus signifying Module 5). 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 5 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 & WEIGHTS BY POLITY & ELECTION YEAR =========================================================================== --------------------------------------------------------------------------- >>> POLITY WEIGHTS IN CSES --------------------------------------------------------------------------- CSES provides users with up to three original weights from each national election study (see variable E1010_) namely: - SAMPLE WEIGHT (variable E1010_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 E1010_2): intended to adjust sample distributions of socio-demographic characteristics to more closely resemble the characteristics of the population - POLITICAL WEIGHT (variable E1010_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 5 Study Page (see https://cses.org/data-download/cses-module-5-2016-2021/). For information on derivative weights calculated for the Cross-National Dataset, users should consult Part 1 and Part 2 (variables E1011-E1014) 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 (2019) X - - | AUSTRIA (2017) - X X | CHILE (2017) - X - | FRANCE (2017) - X X | GERMANY (2017) X X - | GREECE (2015) - X - | HONG KONG (2016) - X - | HUNGARY (2018) - X - | IRELAND (2016) - X - | ITALY (2018) X X X | LITHUANIA (2016) - X - | MONTENEGRO (2016) - X - | NEW ZEALAND (2017) - X - | NORWAY (2017) - X - | TAIWAN (2016) - X - | TURKEY (2018) - X - | UNITED STATES (2016) X X - | ----------------------------------------------------------- | KEY: X = available; - = not available. | | Weights are unavailable for BRAZIL (2018), ICELAND (2016, | 2017) and SOUTH KOREA (2016). --------------------------------------------------------------------------- <<>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - AUSTRALIA (2019): --------------------------------------------------------------------------- <<>> STUDY DESIGN The 2019 Federal election in Australia took place on May 18, 2019. Fieldwork for Australian study began on June 3, 2019, 16 days after the election day. Data collection was completed after 14 days, ending on June 17, 2019. The study, undertaken by Australian National University (ANU) in Canberra, was conducted as part of wave 28 of Social Research Centre’s Life in Australia panel. Members of the panel are Australian residents aged 18 years or more. A total of 2,676 active panel members were invited to take part in the survey, and 2,000 (77.4%) completed the survey. The survey was administered in the English language. The study used a mixed-mode approach, including both online and telephone surveys. A dual-frame random digit dialing (RDD) sample design was employed to undertake recruitment of Life in Australia panel in 2016, with a 30:70 split between the landline RDD sample frame and mobile phone RDD sample frame. For the landline sample, an alternating next/last birthday method was used to randomly select respondents from households where two or more in-scope persons were present. For the mobile sample, the phone answerer was the selected respondent. Only one member per household was invited to join the panel. In May 2018, the panel was refreshed with 267 new panelists. For both the recruitment in 2016 and panel refreshment in 2018, the RDD sample was provided by SamplePages. Panelists receive a small incentive to join the panel and another incentive for each survey they complete. For the Australian (2019) study, all respondents were offered a 10 Australian dollars (AUD) incentive, and they had a chance either to collect the money or to transfer it as a charitable donation to a designated charity. <<>> POLITY WEIGHTS The Australian Election Study provides a SAMPLE WEIGHT. This weight adjusts for the unequal probability of selection. Since some units in the population may not have a chance of selection (for instance, persons without a telephone have no chance of selection for a telephone survey) and there may be different rates of response across unit characteristics, sample weights reduce the extent of any biases introduced through non-coverage. Weights are constructed in the following two steps: 1. Compute a base weight for each respondent as the product of two weights: a. their enrolment weight, accounting for the initial chances of selection and subsequent post-stratification to key demographic benchmarks; b. their response propensity weight, estimated from enrolment information available for both respondents and non-respondents to the present wave. 2. Calibrate the base weights so that they satisfy the latest population benchmarks for several demographic characteristics. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - AUSTRIA (2017) --------------------------------------------------------------------------- <<>> STUDY DESIGN The election was held on October 15, 2017, and fieldwork for the study began four days later, on October 19, 2017. The fieldwork period lasted for 42 days and ended on November 30, 2017. All interviews were conducted by IPR Umfrageforschung via telephone (mobile and landline). The language of survey administration was German. In total, 1,203 interviews were completed and the sample is meant to be representative of all eligible voters in Austria at the day of the election (Austrian citizens, age 16 or older). The primary sampling units were Austrian municipalities, all of which were included in the first stage of the sampling procedure. The sampling procedure included two further stages namely, at the household and respondent level. Households within the municipalities were randomly selected (proportional to the size of the municipality), using a dual frame (50% list frame and 50% random digit dial). Within households, the target person was randomly selected using the next-birthday method. <<>> POLITY WEIGHTS The dataset includes two weights for Austria 2017, a DEMOGRAPHIC WEIGHT and a POLITICAL WEIGHT. The demographic weight (E1010_2) adjusts the sample to match known demographic characteristics in the voting age population (age, gender, education, region, household size, and employment type). The political weight (E1010_3) is a combined weight that adjusts the sample to match the national election results on top of the demographic weight. The combined weight E1010_3, is a winsorized weight in which upper and lower boundaries were determined by the field institute. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - BRAZIL (2018) --------------------------------------------------------------------------- <<>> STUDY DESIGN The Brazilian Election Study was conducted between November, 10 to November 14, 2018. Fieldwork lasted 14 days and started 35 days after the first round and 14 days after the second round of elections. IBOPE Inteligencia, of Sao Paolo, collected data in face-to-face interviews using the electronic questionnaire (CAPI). Interviewing was conducted in the Portuguese language with all Brazilian citizens, 16 years old and older, and registered to vote eligible to be interviewed. The sample for Brazilian study is stratified per state. In the case of states with metropolitan areas, its universe is stratified in metropolitan areas and countryside. The cities are probabilistically selected in each stratum, through systematic PPS (Probability Proportional to Size) method. By this method, 172 cities were selected. In the second stage, census tracts are selected with systematic PPS. In the third stage, a fixed number of respondents are selected according to quotas based on: gender, age, education, line of work, and the number of bathrooms in the household. These quotas were set considering the census update survey from IBGE (Brazilian Institute of Geography and Statistics) 2010, PNADC 2016 (National Survey by Household Sample) and TSE 2018 (Superior Electoral Court). After census tract were chosen, by random walk, in total 2506 interviews were collected. <<>> POLITY WEIGHTS No weights provided. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - CHILE (2017) --------------------------------------------------------------------------- <<>> STUDY DESIGN The General elections in Chile were held on November 19, 2017. These elections included Parliamentary elections (both Lower and Upper house) and the Presidential Election. The second round of Presidential elections was held on December 17, 2017. The fieldwork for the Chilean study started one day after the second round of Presidential elections, on December 18, 2017, and lasted until January 31, 2018. Feedback organization from Santiago, Chile conducted interviewing. The survey was fielded in Spanish, with a face-to-face interviewing using a questionnaire on paper. The Chilean 2017 study is meant to be representative of all citizens of Chile aged 18 years living in Chile, including immigrants with at least five years of residence, which is a requirement to get the right to vote in Chile. The sample for the study was created in four stages. At the first stage, primary sampling units - communas, were selected. These are the smallest administrative units in Chile. They were ordered by size, and every comuna with 50,000 inhabitants or more was automatically selected. Additionally, a sample of 22 comunas (with less of 50,000 inhabitants) was randomly selected. In the second stage blocks inside the communas were selected using stratified random sampling, with a probability of selection proportional to the size of the blocks. In the third stage households within blocks were selected using random systematic sampling procedures. Finally, respondents within households were selected using a Kish table. In total, 2,000 interviews were completed. <<>> POLITY WEIGHTS The 2017 Chilean study includes a DEMOGRAPHIC WEIGHT. This post-stratification weight adjusts the sample to known characteristics of the population. The weight was constructed using Census data and adjusts for gender and age of respondents. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - FRANCE (2017) --------------------------------------------------------------------------- <<>> STUDY DESIGN Interviewing for the French election study began on May 9, 2017, two days after the second round of the 2017 Presidential Election, and continued until May 23, 2017. Kantar Public conducted fieldwork. Interviews were held face-to-face and in French. The sample is meant to be representative of the voting-eligible population, i.e., French citizens who were 18 years or older at the time of the election and registered to vote. Institutionalized persons were excluded from the sample, as were citizens living in the overseas territories due to cost considerations. The sample was realized as a two-stage stratified quota sample. Stratification was based on a matrix cross-tabulating eight regions and four different types of agglomeration (with Paris and its suburbs forming one of these four categories). From the resulting matrix with 29 valid cells, 100 primary sampling units were randomly selected at the electoral district level (circonscriptions). Within the selected circonscriptions, interviewers were instructed to follow random routes starting from a random departure point within the district, including instructions on how to navigate through roads, intersections, city blocks, buildings, and mailboxes. In the second stage, respondents were selected to match predefined quotas for gender (two categories), age (five categories), level of education (four categories), and a combination of occupation (retired or not) and profession (six categories). Quotas were designed to match the latest available data from official statistical sources, with one exception: For education, quotas deliberately underrepresent the lowest education level, reflecting differences in voter registration patterns based on education. For each of the 100 primary sampling units, collaborators aimed for 18 interviews, resulting in 1,830 completed interviews within the deposited sample. <<>> POLITY WEIGHTS The 2017 French Election Study includes two weights: a DEMOGRAPHIC WEIGHT and a combined DEMOGRAPHIC and POLITICAL WEIGHT. The demographic weight corrects distributions in the sample to known socio-demographic population characteristics (gender, age, occupation, region, size of agglomeration, and respondents' diploma). The political weight builds upon the demographic weight and additionally adjusts the sample according to the official election results of the first and the second round of the 2017 presidential election. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - GERMANY (2017) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES module is part of the German post-election cross-sectional study, for which the fieldwork was conducted between September 25 and November 30, 2017, by Infratest dimap Gesellschaft fuer Trend- und Wahlforschung GmbH. Data collection started one day after the federal election on September 24 and lasted for 67 days. The questionnaire was administered in German and in person (CAPI). The sample is designed to be representative of all German citizens with registered residence in the Federal Republic of Germany aged 16 and older, who were (in principle - if underage) eligible to vote in the German federal elections on September 24, 2017. The sample provided to CSES only includes 18-year-olds and above. Residents of islands without land connection and institutionalized persons were excluded from the sampling frame (together less than 0.3% of the eligible population). The survey employed a two-stage stratified random sampling procedure based on resident registers. The primary sampling units were 162 randomly selected sampling points (108 in Western Germany, 54 in Eastern Germany). The sampling points corresponded to a set number of addresses to be selected in the respective municipality during the second stage. Sampling points were allocated to municipalities following a Cox-algorithm. Because of the designed oversampling of Eastern Germans, this procedure was conducted separately for Eastern and Western Germany. The allocation procedure of sampling points to municipalities accounted for stratification along regional criteria. Larger municipalities could be allocated multiple sampling points. After the allocation, a total of 162 sampling points were selected by a systematic drawing procedure with random start for each regional stratum. In case multiple sampling points were drawn in a municipality, the number of addresses selected was increased accordingly. In a second stage, individual respondents were selected from the sampled municipalities. The raw sample of respondents was drawn randomly from the pre-selected municipalities. The selection made use of the state resident register. Again, the procedure used stratification with an age-group x gender-structure matrix. Large municipalities (100,000 inhabitants and above) selected in the first stage were divided into up to 12 spatial clusters. Out of those, four were randomly selected to be included in the individual sampling procedure of the second stage. <<>> POLITY WEIGHTS The 2017 German Election Study provides two weights: a SAMPLE WEIGHT and a DEMOGRAPHIC WEIGHT. The sample weight is designed to account for the oversampling of East Germans. The demographic weight controls for gender, age, educational attainment, functional regional centrality/periphery, and region (East / West Germany). Weights were calculated with an iterative proportional fitting procedure. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - GREECE (2015) --------------------------------------------------------------------------- <<>> STUDY DESIGN The Parliamentary elections in Greece were held on September 20, 2015. These were the second Parliamentary elections in 2015, nine months after the January elections. The fieldwork for the Greece study started 39 days after the election, on October 29, 2015. After 124 days, data collection phase, conducted by the Aristotle University of Thessaloniki, School of Political Sciences, was completed. The survey was fielded in Greek and is meant to be representative of all Greek citizens with voting rights (18 years of age and older). Two interview modes were used: telephone and internet. Respondents were recruited via phone and offered to fill out the questionnaire either online (CAWI) or via telephone (CATI). Because of this recruitment procedure, approximately 18% of households in Greece were excluded, because they do not have a phone. The geographical area was sampled at the first stage, the area code was identified, and the telephone numbers were selected during the second stage by random sampling. Random Digital Dialing (RDD) method was used to select respondents for the study. The individual who answered the phone call was asked to participate in the survey after verifying that they were eligible to vote. If the respondent had an e-mail address, he/she would be encouraged to fill out the questionnaire online. If the respondent had no e-mail address or Internet access, they would be offered a telephone interview. More than 800 respondents, from the study conducted after January 2015 elections, said they are willing to participate again. They were also contacted, either via e-mail or via phone. Up to four follow-up reminders were sent through email to all the respondents, in order to increase the response rate of the survey. In total, 1,078 interviews were completed. <<>> POLITY WEIGHTS The 2015 Greek Election Study includes a DEMOGRAPHIC WEIGHT. The Weight was constructed using the method of raking on gender, age and education. With post-stratification the population was partitioned into subgroups, then the original weights (in Greek case all equal to one) were multiplied by a ratio which was formed by the corresponding population post-stratum size in the nominator and the corresponding sample post-stratum size in the denominator. Raking allows multiple grouping variables to be used by post-stratifying on each variable in turn and the process was repeated until the weights stopped changing. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - HONG KONG (2016) --------------------------------------------------------------------------- <<>> STUDY DESIGN The 2016 Hong Kong election study was conceptualized as a cross-sectional post-election telephone survey (CATI) and conducted between September 6 and September 18, 2016. Fieldwork realized by the Public Governance Programme started two days after the Legislative Council election on September 4. The questionnaire was administered in Chinese (Cantonese and Putonghua). The sample is meant to be representative of registered voters, that is, permanent residents of Hong Kong who are aged 18 or above. Institutionalized persons were excluded from the sample (at most 0.3% of the total eligible population). The same is true for military personnel, who do not have permanent residency in Hong Kong and therefore are not eligible to vote. As fixed-line telephone numbers were called for interviews, other possible exclusions resulted from an unknown percentage of residents without a fixed-line phone connection. The residential fixed-line penetration rate (RFLPR) was used as a proxy measurement and suggests that for every 100 households, there were 94.5 fixed telephone lines in 2016 in Hong Kong. However, 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 5.8 percent. The study was realized as a two-stage random sample. In the first stage, phone numbers were selected randomly from the pool of numbers maintained by collaborators. The last two digits were exchanged with random numbers, to account for numbers not listed in the sampling frame. In the second stage, individual respondents were randomly selected from the households called. The person answering the phone call was asked how many registered voters lived within the household. That number was entered into the CATI system, which would randomly draw one of the eligible persons to be the interviewee. In total, the dataset contains 1,020 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 and sex into account, as indicated by official statistics. For each particular age-sex category, the weight was obtained by dividing the share of persons in the population for that category by the share of respondents in the sample for that category. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - HUNGARY (2018) --------------------------------------------------------------------------- <<>> STUDY DESIGN The parliamentary elections in Hungary were held on April 8, 2018. The interviewing for the Hungarian study started 15 days after the election, on April 23. Median Opinion and Market Research Institute from Budapest, Hungary conducted the fieldwork. The survey was fielded in the Hungarian, using face-to-face interviewing and an electronic questionnaire. All individuals residing in Hungary which are 18 years and older, and have the right to vote were eligible respondents for the study. Institutionalized persons were excluded from the sample. The sample for the study was created using the multi-stage randomized procedure. At the first stage, 120 sampling units representative of the local population were selected based on the census data of the Hungarian Central Statistical Office (KSH). Then 20 addresses from each sampling unit were drawn. In the final stage, at each address, interviewers only interviewed one member of the household determined using the Leslie Kish key. In total, 1,097 interviews were conducted of which 120 were excluded as invalid after having listened to the audio recordings. The sample size achieved through the determined sampling process is 957 respondents. The composition of the sample was heavily distorted regarding gender and age. The distortion of the sample was compensated in an additional round of data collection by determining quotas based on the missing elements from the desired sample in terms of gender and age. The final dataset consists of 1,208 respondents. <<>> POLITY WEIGHTS The Hungarian study provides DEMOGRAPHIC WEIGHT. This weight was designed to compensate for the demographic distortions of the sample and is based on census data. The weight is calculated based on four dimensions: gender, age (five categories), place of residence (three categories) and education. Extreme values on weight variable are recoded to the interval between the minimal (0.3) and maximal (3.5) values. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ICELAND (2016) --------------------------------------------------------------------------- <<>> STUDY DESIGN The election was held on October 29, 2016, and fieldwork for the study began one day later, on October 30, 2016. The fieldwork period lasted for 87 days and was completed on January 25, 2017. All interviews were conducted by the Social Science Research Institute of the University of Iceland via telephone (landline and mobile phones). The language of survey administration was Icelandic. In total, 1,295 interviews were completed and the dataset is meant to be representative of the eligible voting population in Iceland (age 18 or older). The primary sampling unit was individual voters on the Iceland National Register. Individuals were selected from the register, using a simple random sample. This method yielded 2,600 potential respondents in the final sample. Approximately 9.9% of the population was excluded from the initial sampling procedure because they were listed as “Do Not Call” in the National Iceland Register. <<>> POLITY WEIGHTS No weights provided. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ICELAND (2017) --------------------------------------------------------------------------- <<>> STUDY DESIGN The election was held on October 28, 2017, and fieldwork for the study began two days later, on October 30, 2017. The fieldwork period lasted for 83 days and was completed on February 2, 2018. All interviews were conducted by the Social Science Research Institute of the University of Iceland via telephone (landline and mobile phone). The language of survey administration was Icelandic. In total, 2,037 interviews were completed and the dataset is meant to be representative of the eligible voting population in Iceland (age 18 or older). The primary sampling unit was individual voters on the Iceland National Register. Individuals were selected from the register, using a simple random sample. The generated sample included 4,000 people who were sent a letter informing them about the study. Approximately 9.9% of the population was excluded from the initial sampling procedure because they were listed as “Do Not Call” in the National Iceland Register. <<>> POLITY WEIGHTS No weights provided. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - IRELAND (2016) --------------------------------------------------------------------------- <<>> STUDY DESIGN The 2016 Irish general election took place on Friday, February 26, 2016. Fieldwork began six days after the elections were held. All interviews were carried out by RED C Research & Marketing Ltd between March 1 and March 6, 2016. The survey was administered in English via telephone. To be eligible for inclusion, respondents had to be Irish citizens and be at least 18 years old. Institutionalized persons were excluded from the sample. The sample was drawn in two stages. In the first stage of the sampling procedure, the field institute used the random digit dial (RDD) method to randomly select households. These respondents were interviewed in twelve pre-election polls leading up to the election. Half of the sample was interviewed using an RDD landline, the other half using an RDD mobile phone line. Approximately 65% of the respondents interviewed in the pre-election polls agreed to be re-contacted for the post-election study in which the CSES questionnaire was administered. This led to a sample poll of 8,000 respondents. A random contact approach was conducted among this set of respondents, using both a landline approach and a mobile approach. The sample was stratified along quotas so that it would accurately represent the electorate. If necessary to fulfill the quotas, new interviews from the initial RDD were conducted. In total, 1,000 interviews were completed. <<>> POLITY WEIGHTS The Irish National Election Study provides a provides a DEMOGRAPHIC WEIGHT. The weight adjusts the sample to match known demographic characteristics in the voting age population (age, class, gender, and region). --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - ITALY (2018) --------------------------------------------------------------------------- <<>> STUDY DESIGN The election was held on March 4, 2018, and fieldwork for the study began four days later, on March 8, 2018. The fieldwork period lasted for 55 days and ended on May 2, 2018. All surveys were administered by Demetra Opinioni.net via telephone, mobile phone, and the Internet. The language of survey administration was Italian. In total, 2,001 surveys were completed, and the data is meant to be representative of the Italian voting age population according to the Lower House Electoral Rules (Italian residents, age 18 or older; the voting age for Upper House Elections, also included in the study, is 25 years). The initial sample was drawn in two steps. In a first step, the collaborators randomly drew electoral districts within seven macro areas, stratified by the population in of those macro areas and excluding nine large cities (Bari, Bologna, Florence, Genoa, Milan, Naples, Palermo, Rome, and Turin). The region Valle d’Aosta/Vallee d’Aoste and the province Bolzano/Bozen were excluded from the sample frame because voters living in the region voted under different electoral rules than the rest of Italy. This resulted in the exclusion of approximately 1.53% of the eligible voting population. In the second step of the sampling procedure, the field institute sampled respondents for an online component of the study, a phone component, and a mobile phone component of the study in the following ways. For the online component, panelists of an existing Demetra Opinioni.net web panel (total N=13,734; initial recruitment by Computer Assisted Telephone Interviewing, CATI) were randomly selected within the selected districts from the first sampling step as well as the nine larger cities. The chosen panelists were invited to participate in the election study via email and provided with the opportunity to fill out the election study questionnaire online. Invited panelists were offered a telephone top-up or Amazon voucher worth 1EUR upon completion. 59.63% of invited panelists completed the survey. For the second set of respondents (N=1,501), respondents were selected using a dual frame (66% list frame and 34% Random Digit Dialing) through random digit dialing based on the national phone book and random digit dialing for mobile phone interviews. Respondents selected from the national phone book were interviewed using Computer Assisted Telephone Interviewing (CATI; N of completed interviews=1,000) and respondents selected by random digit dialing for mobile interviews were interviewed using Computer Assisted Mobile Interviewing (CAMI; N of completed interviews=500). Respondents chosen for the mobile phone interviews were asked at the beginning of the interview about the city and neighborhood they resided in. If they did not live in a city or district sampled in the first stage, the interviews were canceled and the respondents were flagged as ineligible. In total, the two stage sampling procedure resulted in approximately 20 interviews per electoral district (range: 14 to 26 interviews) except for Milan (73 interviews), Naples (38 interviews), Rome (102 interviews), and Turin (47 interviews) where more respondents were interviewed. <<>> POLITY WEIGHTS The dataset includes three weights for ITALY 2018, a SAMPLE WEIGHT, a DEMOGRAPHIC WEIGHT, and a POLITICAL WEIGHT each of which was created for the whole dataset. The sample weight (E1010_1) adjusts the sample to correct for the different selection probabilities of clusters (i.e. constituencies and large cities) within different macro-areas as well as for the different probabilities of selection of individuals within clusters. The demographic weight (E1010_2) adjusts the sample to match known demographic characteristics namely gender (male, female), age group (18-44, 45-59, 60+), geographical area (north, center, south), and education (low, middle, high). The political weight (E1010_3) adjusts the sample to the population frequencies by vote shares of the major coalitions and parties (center-right and center-left) in the lower house election and abstainers at the regional level. Because the voting age differs between the lower house (18 years and older) and the upper house (25 years and older), users should be aware that applying any of the three weights when analyzing upper house elections may result in non-representative results. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - LITHUANIA (2016) --------------------------------------------------------------------------- <<>> STUDY DESIGN The first round of Parliamentary elections in Lithuania was held on October 9, 2016. The second round of elections, in constituencies where no candidate won required majority to be elected, was held on October 23, 2016. The interviewing for the Lithuanian study started 33 days after the first round and 19 days after the second round of elections, on November 11. Baltic Surveys Ltd from Vilnius, Lithuania conducted the fieldwork. The survey was fielded in Lithuanian, using face-to-face interviews with a paper questionnaire. The Lithuanian study is meant to be representative of all citizens of Lithuania aged 18 years or older who are registered to vote, with only the homeless people being excluded from the sample frame. The sample for the study was created using the multi-stage stratified proportional sampling. At the first stage, based on Nomenclature of Territorial Units as used in the official EU statistics, the Republic of Lithuania was divided into ten administrative units (counties). Then, the total number of interviews in each survey region (county) was allocated to six strata in proportion to the population of each stratum. The stratum is defined by a settlement size. In the third stage, primary sampling units were randomly selected from each stratum (county and settlement size combination cell) list. After that, households were selected using the random route from the starting point, and respondents within households were selected by using “last birthday rule” among all household members aged 18 years and older. In total, 1,500 interviews were completed. <<>> POLITY WEIGHTS The Lithuanian study provides a DEMOGRAPHIC WEIGHT. This weight was designed to make the data representative of the total population by gender, age and urban/rural population. Weights were calculated by dividing counts based on official statistics with those achieved in the survey sample. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - MONTENEGRO (2016) --------------------------------------------------------------------------- <<>> STUDY DESIGN The 2016 Montenegrin cross-sectional post-election study was conducted as a face-to-face survey, from December 8, 2016, to January 16, 2017. Data collection was realized by the De Facto Consultancy, starting 53 days after the parliamentary election on October 16, 2016. Interviews were conducted in Montenegrin. The sample is meant to be representative of eligible Montenegrin citizens, aged 18 years or older and registered to vote. Institutionalized persons (people in prison or hospitals) were excluded from the sample (less than 1 percent of the eligible population). The sampling was conceptualized as a stratified multistage random sample, with three stages in total. In the first stage, the population was divided into three regions, North, Center and South. These regions were the primary sampling unit and chosen based on differences in population size, the distribution of ethnic groups within a region, age groups, and the region's geography, economy and history. Second, inside of each regional stratum, polling stations were identified, which constitute the smallest municipal unit which is at the same time a unit for voting. These polling stations formed the second level of stratification and were divided into three groups according to their size. Polling stations were selected randomly from each group. In the third and final stage, a random procedure based on 'step-and-go' was employed inside polling stations to choose the household. The last person to have had their birthday within the household was interviewed. In total, 1,213 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 E1010_2 is a multiplication of the two individual weights. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - NEW ZEALAND (2017) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES module release is from persons freshly sampled for the New Zealand Election Study. Those persons forming a Panel back to the 2014 election were also given the module questions, but are not included in the CSES release. Fieldwork was conducted between September 26, 2017 and February 28, 2018. Hence, data collection started three days after the general election and lasted for 156 days. The questionnaire was administered in English and could be undertaken via mail or the Internet. The sample consists of all eligible citizens that registered on the electoral roll on the writ day then aged between 18 and older. On the writ day, the electoral roll contained about 88% of those eligible. Enrollment continued after writ day until the day before the election, with the final roll including just over 92% of the eligible population. Those in prison were excluded from the sample but as they are not eligible to enroll or vote, this does not influence the sampling frame. Military personnel were included in the sample. The primary mode was mail self-completion and citizens were given the option of completing it by the Internet. Therefore, members of the population without access to the Internet were not excluded. The survey employed a random sampling procedure based on the electoral rolls as of August 23. The primary sampling units were four groups where the roll was segmented into. Those in general electorates 18-30 years, those in general electorates 31 and above, those in Maori electorates 18-30 and those in Maori electorates 31 and above. The young and those in The Maori electorates were over sampled to adjust for lower response rates. The primary sampling units were not randomly selected as they contained together the whole roll for sampling. Further stages of selection were the individuals on the electoral rolls within the four segments. At those further stages, units were randomly selected. The sample design included clustering and quota sampling in the sense that it was sampled within the four segments. The survey used a draw for four small monetary prizes as participation incentives. <<>> POLITY WEIGHTS The 2017 New Zealand Election Study provides a DEMOGRAPHIC WEIGHT. The weight controls for gender and age as some groups were under or over-represented even after re-sampling the oversamples. Therefore, the weight makes the survey representative of the electoral roll from which it was sampled (about 92 per cent of the voting population). For the weight construction, respondent's age and gender were used. It was generated from a table breaking down the entire roll by age and gender. --------------------------------------------------------------------------- <<>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - NORWAY (2017): --------------------------------------------------------------------------- <<>> STUDY DESIGN The study is a post-election survey that was fielded between September 20, 2017, and October 16, 2017. Data collection was realized by Statistics Norway, starting nine days after the Norwegian parliamentary election. Sampling frame for the study is a two-stage design. In the first stage, the country was divided into 363 primary sampling units (psus), based on the local municipalities. 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. 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. Interviews were conducted in Norwegian via the Internet. The sample is meant to be representative of the Norwegian citizens aged between 18 and 91. As the interviews were conducted via the Internet, Norwegians without Internet access were excluded from the sampling frame (2 % of the eligible population). Respondents received both an e-mail and a text message. Both the email and the text message contained a hyperactive link which, when activated, led the respondents straight to the survey, no need for user ID or password. It also included general information about the survey and its theme. The respondents did not get any incentives for their participation in the study. Additional reminders were sent to those who did not participate in the survey after the initial contact. The CSES module was previously a part of the Norwegian National Election Study. However, the CSES module was taken out of the 2017 election survey and conducted separately. Thus, none of the respondents was a part of a previous panel. The sample consisted of 5000 persons over the age of 18 which was pulled randomly from the National registry covering all Norwegian inhabitants. As they pulled individuals and not household, more than one person could have been interviewed from a single household. From the sample if 5000 persons, 1782 participants completed the interviews. Therefore, the study received a response rate of 36.5 %. <<>> POLITY WEIGHTS The 2017 Norwegian Election Study contains a DEMOGRAPHIC WEIGHT. The weight was constructed after collaborators found that over- and under representation in the survey was related to particular demographic characteristics. Respondents with higher education were over represented compared to those with lower education. Those in the age group 45-66 years were over represented as well, whilst the oldest in the sample were underrepresented. The weight adjusts for gender, age, and education, dividing the sample into 18 different strata. Gender contained two categories: male and female. Age was divided into three categories: 30 years or under, 31 to 59 years, and 60 years and over. Education was also divided into three categories: first no education/ primary school or unspecified, second high school, and third university/ college. The stratified weight is sat then as equal to the number of persons in the strata in the population divided by the number of persons in the strata in the net sample. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - SOUTH KOREA (2016) --------------------------------------------------------------------------- <<>> STUDY DESIGN The 2016 South Korean cross-sectional post-election study consists of face-to-face interviews, which were conducted from April 14 to April 23, 2016. Fieldwork started the day after the South Korean legislative election on April 13, 2016, 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 one percent of the population. The sample was realized as a multistage random sample, for which electoral districts represented the primary sampling units. In the first stage, 80 electoral districts were randomly sampled from a list of all electoral districts in South Korea. At the second stage, voting districts within each sampled electoral district were randomly sampled. In the third and final stage, a simple random sample of voters was drawn within each voting district, based on electoral registers provided by the Central Election Management Committee. When contacting a respondent’s household, interviewers asked for the name of the respondent, and identified her or him to be the right person listed. Prior to the study, a token gift (folding umbrella) was sent to the sampled individuals. The final dataset includes 1,199 completed interviews. <<>> POLITY WEIGHTS No weights were provided. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - TAIWAN (2016) --------------------------------------------------------------------------- <<>> STUDY DESIGN General elections in Taiwan were held on January 16, 2016. These included Taiwan presidential and legislative elections. Fieldwork for the Taiwanese study started one day after the elections and lasted until April 28, 2016, in total 103 days. Interviews were conducted by the Election Study Center at the National Chengchi University. The survey was administered in Chinese and respondents were interviewed face-to-face using an electronic questionnaire (CAPI). Throughout the whole questionnaire for Taiwan (2016) study additional categories "it depends" and "no opinion" were used apart from "refuse," "don't know" and "missing." Since these two categories do not match CSES standard, they were recoded to missing for all variables. The sample is meant to be representative of all qualified voters in Taiwan. These are all Taiwan citizens who are 20 years or older. The survey used a probability proportional to size (PPS) three-stage systematic sampling. In the first stage, legislative constituencies were sampled according to major geographic regions. In the second stage, urban villages or rural villages were selected as sampling units. In the third stage, respondents were selected from household registration data provided by the Ministry of the Interior. In total, 7,601 citizens were sampled, out of which 1,690 completed interviews have been collected. All respondents received a gift as an incentive to participate in the study. <<>> POLITY WEIGHTS Taiwan (2016) study provides DEMOGRAPHIC WEIGHT. The provided data weights are constructed by raking with gender, age (5 groups), education (5 groups), and area (6 regions), to fit the sample to eligible voters in 2016 presidential and legislative elections. The demographic characteristics are based on “2015 Taiwan-Fuchien Demographic Fact Book, Republic of China”, published by the Ministry of Interior, Republic of China. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - TURKEY (2018) --------------------------------------------------------------------------- <<>> STUDY DESIGN The CSES module is part of the Turkish post-election cross-sectional study, for which the fieldwork was conducted between July 23 and September 9, 2018, by Frekans Arastirma. Data collection started 29 days after the Turkish General Election on June 24 and lasted for 49 days. The questionnaire was administered in Turkish and in person (PAPI). The sample is designed to be representative of the Turkish voting-age population living in urban as well as rural areas – that is, Turkish citizens aged 18 years or older. Institutionalized persons in prisons or hospitals and military personnel were excluded from the sampling frame (about 1% of the voting-eligible population). The sample was realized as a stratified multistage random sample. NUTS-2 regions, as given by the Turkish Statistical Institute (TUIK), formed the basis for stratification. In the first step, the target sample was stratified according to each region's share of the urban and rural population following the Address-Based Population Registration System (ADNKS) records as of the end of 2017. Next, TUIK's address 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. Based on the probability proportionate to population size (PPPS) principle, blocks were distributed to NUT1 regions. In a third step, for each of the twenty addresses within each block, 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 twenty addresses in a systematic random sample from the list of households in the village. To select individuals within sampled households, names of all reported eligible household members 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, the selected 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 realized after three visits, the respective household was dropped from the sample without being substituted. Overall, the sample includes 1,069 completed interviews. <<>> POLITY WEIGHTS The 2018 Turkish election study includes a DEMOGRAPHIC WEIGHT. The weight adjusts the sample to the following known population distributions of gender, age, and educational attainment: Males and Females, six age groups (18-24, 25-34, 35-44, 45-54, 55-64 and 65+-year-olds), and five education groups (below primary school, primary school graduates, secondary school graduates, high school graduates, University graduates and above). The resulting 60 cells (2 x 6 x 5) were filled with the respective respondents from within the sample to calculate the weight and to obtain the national aggregates. Because there are minor variations in the geographic planned distribution of observations and realized interviews, collaborators calculated a separate geographic weight. The demographic weight included in CSES is the product of this geographic weight and the demographic weight described above. --------------------------------------------------------------------------- >>> OVERVIEW OF STUDY DESIGN AND WEIGHTS - UNITED STATES (2016) --------------------------------------------------------------------------- <<>> STUDY DESIGN The American National Election Study (ANES) 2016 Time Series has a pre- post-election panel design in which respondents were interviewed twice: Once prior to the election and once after the election. The CSES sample includes only respondents that took part in both rounds of interviews. The study features a dual-mode design combining face-to-face interviewing with a separate sample interviewed on the Internet. Respondents were assigned to modes and did not switch modes between interviews. Post-election interviewing featuring the CSES module started on November 9, 2016 - the day after the General election on November 8 - and continued until January 9, 2017. The study was administered by Westat, Inc. in both English and Spanish. Probability samples were drawn independently for face-to-face and web interviews. Both samples are meant to be representative of U.S. citizens aged 18 or older at the time of recruitment. As the sampling frame for both modes was composed of lists of residential addresses provided by the U.S. Postal Service, a respondent had to reside at the sampled address to be eligible. Addresses where mail is not delivered were excluded from the sampling frame. Residents from the states of Alaska and Hawaii were not sampled for face-to-face interviews (less than one percent of the study population). Furthermore, addresses associated with more than one dwelling unit (“drop point addresses”) were excluded from the sampling frame for web-surveys (2.9 percent of residential addresses). The face-to-face sample was realized as a multi-stage stratified cluster sample. Primary sampling units (PSUs) were counties, or a combination of counties, with a minimum population of 50,000 people. The Los Angeles County was divided into two PSUs due to its large size. For the face-to-face mode, 60 PSUs were selected with probability proportional to the number of adult citizens, previously stratified by Census region, the prevalence of poverty, members of minority groups and population size. The five largest PSUs were selected with certainty. In a second stage, within each PSU, four Census Block groups were drawn as the secondary sampling units. Within these areas, households were sampled randomly from addresses included in the U.S. Postal Service’s computerized delivery sequence file (DSF). Interviewers were sent to the sampled address and conducted a brief screening interview in which eligible respondents from the household were listed. One person was randomly selected from this list. During the last two weeks of pre-election fieldwork, half of the remaining eligible cases were sub sampled out, to concentrate resources on a smaller set of cases (adaptive design). Depending on the level of pre-election incentives, respondents were initially offered 25, 50 or 100 USD for completing the post-election interview. These amounts were increased to 50 or 100 USD, respectively, at the end of the fielding period. The web-sample is based on a simple random sample from the list from the DSF-file, excluding drop point addresses and addresses contacted for the ANES pretest. Selected addresses were sent letters, asking one household member to go online for completing a survey. That survey contained a screening instrument to randomly select one person from all eligible household members for the final interview. For post-election web interviews, incentives were set to the amount paid for the pre-election interview, that is, either 40 or 80 USD. Panel attrition rates were 10 percent for the face-to-face component, and 16 percent for the web component. The sample consists of 1,058 and 2,590 completed face-to-face and web interviews, respectively. <<>> POLITY WEIGHTS The American National Election Study includes a combined DEMOGRAPHIC and SAMPLE WEIGHT. The weight adjusts for unequal probability of selection, non response, age by gender, race/ethnicity by educational attainment, marital status by gender, race/ethnicity by census region, nation of birth and home tenure by metropolitan status. The weight deposited with CSES is targeted to the complete U.S. 2016 dataset (including both web and face-to-face modes). Users interested in conducting analyses with either face-to-face or web interviews only are referred to the ANES 2016 Time Series Study, which provides individual weights for each mode. /// END OF FILE