iassist quarterly - 11 The Swedish Election Studies: studies of 30 years of Swedish electoral behavior bv Iris Alfredson 1 Before using data for secondary analysis, the researcher must be aware of the principles used in compiling the material. The aim of this paper is to give a short description of the Swedish election studies, and discuss some of the problems which may arise if one is lo study changes over time using these data.. Quite a few variables are common lo all ten studies, but are they comparable? There may have been slight changes in question wording, or the coding of the variable may have changed over lime. These are some of the changes thai affect direct comparisons. Additional complications arise when making international comparisons, but such problems are outside the scope of this paper. '•Presented at the International Association for Social Science Information Service and Technology (IASSIST) Conference held in Vancouver, British Columbia, Canada on Mav 19-22, 1987 / hope that this presentation will act as an introduction to the extensive material available on Swedish elections. For those whose requirements are more comprehensive, a bibliography of books, reports and papers on the subject is provided. Background In conjunction with the local elections of 1954, Jorgen Westerstahl and Bo Sarlvik of the Department of Poliucal Science, University of Gothenburg, conducted the first scientific field survey of Swedish voting behavior. The survey was a local election study conducted in Gothenourg and the countryside around Boras. The survey was inspired by that conducted by Lazarsfeld, Berelson and Gaudet in Eire County. The questions concerned mainly social background, political party sympathies and motives, previous political sympathies, stated and actual vote behavior, reasons for change in sympathies, public reaction to content of media coverage, political sympathies of friends, and knowledge of and poliucal opinions on specific elecuon campaign issues. The 1954 survey was a pilot test for a larger national study. Two years later, in conjunction with the parliamentary election of 1956, the first nation wide survey on party choices, participation and poliucal opinions of the Swedish electorate was conducted. This was the real start of the Swedish elecuon research program. Since then, similar surveys have been carried out at all elections. Further, studies have been conducted in conjunction with the two referenda that have taken place since then, the referendum on the general supplementary pension scheme (ATP) in 1957, and a referendum on nuclear power in 1980. The Swedish election studies are, together with those carried out in the United States, Norway, Summer 1987 12 - iassist quarterly France and West Germany, the only academic election studies that extend as far back as the 1950s. The Swedish parliamentary election studies have been carried out at every election since 1956, making them one of the most comprehensive sources on voting behavior. Financing The local elecuon study of 1954 and the national election study of 1956 were funded by the Swedish Social Science and Legal Research Council, the Foundation for Research in Sociology at the University of Gothenburg, the Swedish Broadcasting Cooperation and the political parties. Since 1960, the election studies have been financed through government grants and carried out as a part or the election statistics program by the Ceniral Bureau of Statistics (Staususka Central byran). Survey Design Since the mid-1950s, there has been a political behavior research program at the Department of Poliucal Science, University of Gothenburg. The elecuon studies have, with the exception of the 1976 study, come into being through a close collaboration between the Department of Poliucal Science at the University of Gothenburg and the Swedish Central Bureau of Stausucs (SCB). The 1976 study was conducted b\ the Department of Political Science in Uppsala in collaboration with SCB. The Survey Research Center of the Central Bureau of Statistics is responsible for the sampling for the election studies, their permanent interview organization performs the field work, and they also collect additional data from public registers. The research project at the Department of Poliucal Science is responsible for the general planning of the studies, the construction of the questionnaires, and the analysis and presentation of data. The basic principles of the studies, the questionnaires, and the coding scheme were originally designed by Bo Sarlvik. The surveys were initiated by professor JOrgen Westerstahl and directed by Bo Sarlvik (1956, 1960, 1964, 1968, 1970, 1973), Olof Petersson (1973, 1976), and SOren Holmberg (1979, 1982, 1985). The sample represents the resident, enfranchised population. The samples for the earlier election studies were drawn from the Survey Research Center's sampling framework which consisted of a nationwide set of primary sampling units which provide the framework for a 'general purpose' two-stage population sample. Since 1973, another method has been applied. The samples are now drawn directly from the SCB register over the total population (RTB), by- means of the SCB standard program for random samples. Respondents not included in the target population because of ineligibility to vote are excluded from the sample. The target population even includes Swedes resident abroad, but these are not included in the sample. People above 80 years of age at the time of the study are excluded in order to avoid the difficulties encountered in interviewing very old people : the 1968 survey was an exception, the age limit was 84). The proportion excluded by this age limit is about 3% of the total sample. In 1956, respondents were interviewed twice, once before and once after election day. From the 1960 study onwards, with the exception of the 1970 study, field work has been carried out in two stages. The total sample is split into two subsamples of equal size. One subsample is contacted for personal interviews during the field work stage preceding the election. Respondents in this subsample are contacted again after elecuon day through a short mail Summer 19S7 iassisl quarterly - 13 questionnaire. The primary purpose of this mail questionnaire is to obtain information about the final vote decision of these respondents. The second subsample is contacted for personal interview during the weeks immediately after the election. Most of the interviews are held between mid-August and mid-October. The 1970 election study differs from the others because the entire survey was carried out after election day. Different techniques were used in 1970: about 1/3 of the sample were interviewed in their homes, and additional 1/3 through telephone interviews, and the remainder received a short mail questionnaire. In the other election surveys, respondents were interviewed in their homes. Some busy respondents were interviewed by telephone. Respondents interviewed before the election received, immediately after election day, a short mail questionnaire which mainly contained questions on final voting decision. panel, in which half of the 1973 sample was reinterviewed in 1976. The "new" respondents in 1976 were reinterviewed in 1979, and so on. In this way, all respondents are interviewed twice. At present, there are four panels: 1973-1976, 1976-1979, 1979-1982 and 1982-1985. In every survey, a supplementary- sample of first-time voters, who have become entitled to vote since the last election, is also drawn. In conjunction with the 1957 referendum, respondents were interviewed three times; this design was also applied to the 1980 referendum. The questionnaire which was sent out to respondents in the 1980 referendum sample was also sent out to respondents belonging to the 1979 election study sample. In this way, two long term panels were created, one for the years 1976-1979-1980 and one for the years 1979-1980-1982; this makes it possible to do analyses over a longer time span. Panels In order to facilitate the study of variation and constancy in voting behavior, a panel design has, with one exception, been used. The panel technique used has varied over the years. In 1956, respondents were interviewed twice, once before and once after election day. In 1960, no panel was used. In 1964, a three-stage panel was started, in which the same respondents were interviewed at the elections of 1964, 1968, and 1970. A major problem with a panel extending over such a long period of time is that the sample loss has a tendency to increase at every stage. In order to facilitate the study of individual changes, but at the same time avoid too large a sample loss, a new type of panel was introduced in 1973. This was a kind of "rolling" two-stage Sample Loss During the first decades in which Swedish field surveys were conducted, sample loss rates were low, about 5% - 7%. In the 1960s and the beginning of the 1970s, there was a dramatic rise, the sample loss increasing to about 20% before stabilizing at that level. Sample loss in the Swedish election studies have also followed this pattern: the 1965 sample loss was 5%. during the 1960s it slowly increased (8%, 8% and 12%) until it reached 14% in the 1970 survey. Then there was a sudden increase, from 18% in the 1973 survey to 26% in the 1976 survey. In the latest election studies, sample loss has been reduced to less than 20%, mainly through the use of shorter interviews. The main portion of the sample loss was due to refusals to participate, and through shorter interviews with respondents who are unwilling Summer 1987 14 - iassist quarterly or are pressed for time, sample loss can be reduced. The shortened interviews take about half the time of a normal interview. Some respondents answered an extremely short interview, containing only those questions on final vote decision. The framing of the questions The Swedish election studies series now extends over a period of thirty years. At present, the series consists of ten surveys conducted in conjunction with parliamentary elections, and two conducted in conjunction with referenda. During this period, a great number of questions have been asked. Some questions are repeated in all surveys, which makes it possible to study changes over a period of thirty years, and there are also questions specific to one or two surveys. Some surveys have a large number of questions on media, while others (1970 and 1973) don't touch on the subject at all. In the 1973 and 1976 surveys, there were many questions on international politics and events, an area not covered in the other surveys at all. The 1968 election study was almost twice as large as the others, largely because of the goal of the survey which was to cover the system of representation. Comparable questions were also asked in an interview survey of Members of the Lower House of Parliament 1 . Questions asked on current topics include: questions on public representation on bank- boards (1968, 1970), possible Swedish membership in the European Economic Community (1968, 1970, 1973), quesuons on nuclear power (1976, 1979), and wage-earners" investment funds (1979, 1982). : RepresentationsundersOkningen, by Sarivik et al.) Bo However, the central questions in an election study are always those on party preference. Therefore, all surveys contain questions on: respondent voting habits, party preferences, and voting in both current and previous elections. Voter participation is also checked using electoral registers. Information on father's political sympathies is also available in most of the surveys. Other recurring questions cover political interest and party identification. Social background factors such as information about the respondents' date of birth, sex, marital status, education, occupation and trade union affiliation are also available. Information on occupation, education, trade union affiliation and political party membership in the 1970 survey (the third stage of the 1964-1968-1970 panel) was extracted from the 1968 survey, for all respondents except those lost from the 1968 sample and those added in the 1970 supplementary sample. For these respondents, information from 1970 was used. The Swedish Social Science Data Service is at present compiling a continuity guide to all questions and variables used in the Swedish election studies. Problems of comparability A series of surveys extending over thirty years could be a gold mine for researchers wishing to study changes over time. Such comparisons are not, however, always problem-free; a number of factors make comparison difficult, such as changes in question wording between surveys. In addition, questions may not address the same groups, the variable coding may have changed, the source of background variables may change from registers to interviews, or vice versa. Changes in society, such as an increasing proportion of women in the work force, and Summer 1987 lassist quarterly 1? new education systems, also make direct comparisons difficult In order to save time and trouble, it is necessary' to study the construction of the variables carefully before starting analysis. The following are problems that have been identified in the Swedish election studies. Marital status: Information on the respondent's marital status has alternately been gathered through questionnaires and from registers. In 1956, the question asked was: "Are you married, unmarried, widow/widower, divorced?". In 1960-1976. this information was extracted from a population register. In the most recent surveys, the information has again been collected in the interview. The question in 1979/1982 was worded: "Which of the following alternatives best describes your marital status (Married or co- habiting, unmarried or divorced, widow/widower)? The advantage of this method is that the data collected are more up-to-date. The number of code categories used has also varied over time; from 1956 to 1968, four categories were used: married, widow/widower, divorced or unmarried. Since the end of the 1960s, it became more common that people live together without being married; this is reflected in the election studies. In the 1970 survey, the category 'unmarried but cohabiting' was introduced. Since the 1976 survey, the number of categories has been reduced to three: married/cohabiting couple, unmarried or divorced, widow/widower. Education: Changes have been made to the variables on education over the years, because of changes in society. New school systems have been introduced, and the general level of education has risen dramatically. The coding scheme has also changed: the election studies conducted from 1968 to 1976 used a very detailed coding for education. Question wording has remained fairly constant over the years: Have you ar.y education above 'folkskol' level? (IF YES:) What education do you have?" (1956-1964) "Do you have any practical or theoretical education above 'folkskol' level? What education? Have you any other practical or theoretic education?" (1968-1973) "What education do you have? Have you any other practical or theoretic education?" (1976-1982). The early election studies (1956-1964) have the same education code categories, except that category 2 (folkhOgskola, yrkesskola) in 1956 was divided into two categories in the following two surveys. Categories in the earlier surveys remain in the more recent surveys (1979-1982), but changes in the school system are easy to track. Categories 3 and 4 in the 1960 and 1964 election surveys have been combined into a new category 3. This category includes 'grundskola', a level which did not exist earlier. There is a new category, '4', which includes education in 2-year 'gymnasie' courses; this is also a new level of education, as compared with earlier studies. During the period 1968-1976, a very detailed coding scheme was used for education. The question on education was split with a variable for general basic education, followed by addiuonal variables for other education. Education is coded with a three digit cod, of which the first digit denotes the main group, the second the level of education, and the third the type of education (degree). Occupation: Major problems of comparability occur in the definitions of work and class. There are problems in the classification of married women and students, and a new classification system for occupation has been introduced. The variable 'occupation group' is present in all surveys. In the 1956 survey, it had 12 categories; since 1960, it has had over 30 categories. Married, non-working women have been included in different categories over the years: in the 1956 survey, all married women were classified according to husband's occupation. In 1960 and 1964, working married women were coded according to their own occupation, while non-working married women were coded according to husband's occupation. In 1968, 1970 and 1973, a special code was used for these women, and since 1976. they have been classified according to previous occupauon. Summer J 987 Id - iassist quarterly Students have also been treated differently over the years. Since 1970 they have had a separate code category, but before that, they were coded according to the social status or occupation of the family head. Since 1968, a detailed, 3—digit code has been used for occupation. The first two digits denote area of occupation, the third digit status of occupation. Place of residence: All surveys, with the exception of the 1970 and 1973 studies, include information on respondent's place of residence. The categories have changed from survey to survey. With the exception of the 1982 survey, it is possible, by combining categories, to extract three comparable categories: large cities (Stockholm, Goteborg and MalmO), other towns, and rural areas and villages. Party preference: The question "Which party did you vote for?" is asked in all surveys. Respondents interviewed before the election, with exception of the 1964 survey, received a questionnaire immediately after the election. Respondents in the 1964 pre-election sample were asked "Which party do you like best?" Information on who actually voted in the election is also available. In most surveys (1956 to 1964 and 1976 to 1982), one variable is used to summarize the contents of the two variables on voting and election participation. Respondents who stated that they voted for a certain party, but who, according to the election register, didn't vote, are excluded. With the exception of the 1956 study, all surveys have contained a question on how the respondent voted in earlier elections: "Did you vote in the election 19..? (IF YES:) Which party did you vote for?" Similar data can, from the 1956 survey, be extracted by combining the answers to the following questions: "Did you vote for the same party at earlier elections?" and for those who answered 'no': "Which party did you vote for? (1956-1964) there is only one question: "Some people are strongly convinced adherents of their party. Others are not so strongly convinced. Do you yourself belong to the strongly convinced adherents of your party? In the later surveys, party identification is measured in three questions: "Many feel strongly for a particular party, whilst others do not feel the same allegiance towards any of the parties. How do you see yourself, as a liberal or social democrat or moderate or centrist or communist? Or don't you have this attitude towoards any of the parties?" The respondents who consider themselves party adherents are asked the following question: "Which party do you like best?" and finally, those who named a party were asked: "Some people are strongly convinced adherents ...". Newspapers: With the exception of the 1970 and 1973 surveys, questions on which newspapers respondents read have been asked. But one must be carefull with these data. In the earlier studies (1956 to 1964), respondents were asked which newspapers they read daily, while in the remaining surveys (1968, 1976-1982), respondents were asked which papers they read regularly. In the 1976 survey, 'regularly' was defined as at least 4 times/week for daily newspapers and at least every second week for weekly papers, while in the 1979-1982 surveys, 'regularly' was defined as at least once/week for all papers. In 1968, only one-half the sample were asked which papers they read regularly. This question was asked in the mail quesuonnaire which was sent to pre-election respondents, and 'regularly' was not defined. The coding of the variable in the earlier surveys makes direct comparison with later surveys impossible. The code categories included a variety of combinations of information on subscriptions, political affiliation of the newspapers, type of newspaper, etc. Since 1968, the names of the newspapers have been coded. Party identification: All studies include questions on party identification. In the earlier surveys A project to recode the oldest surveys, is currently ongoing at the Department of Political Summer 1987 iassisl quarterly - 17 Science, Gothenburg University. This will, hopefully, solve many of the problems of non-comparability among the surveys. Swedish election studies at the Swedish: Social Science Data Service ((SSD)) All election studies through 1982 and the 1980 referendum survey have been deposited in the Swedish Social Science Data Service (SSD). With the excepdon of the referendum study, they have all been documented with the aid of the GIDO-system, which produces a data file and machine-readable codebook in OSIRIS format These OSIRIS-format files can be converted to other formats. The 1980 referendum is currently being documented, and we hope that, in the near future, the 1985 elecdon survey will also be available. It is also possible that the referendum study of 1957 will be documented. The following is a short summary of the Swedish election studies currendy available from the Swedish Social Science Data Service (SSD): SWEDISH ELECTION STUDY, 1956 (SSD 0020) Principal investigators: JOrgen Westerstahl and Bo Sarlvik, Department of Polidcal Science, University of Gothenburg. Total sample: 1,146 Number of respondents: 1,088 Sample loss: 58 (4.9%) Number of variables: 220 Weight: Persons born 1876-1885 were sampled using 1/2 probability. These are represented by two cards in the data-file. The total number of respondents is therefore 1.131 (43 duplicates). Method: Interview in home. Panel: Respondents were interviewed twice, once before and once after elecdon day. Format: OSIRIS, SPSS-x, machine-readable codebook. SWEDISH ELECTION STUDY, 1960 (SSD 0001) Principal investigator: Bo Sarlvik, Department of Polidcal Science, University of Gothenburg. Total sample: 1,603 Number of respondents: 1,466 Sample loss: 137 (8.5%) Number of variables: 215 Method: Interview in home. One half of the sample was interviewed before elecdon day, the other half after. Pre-elecdon respondents also answered a short mail quesdonnaire after the election, which mainly contained quesdons on final vote decision. Panel: None. Format: OSIRIS, SPSS-x, machine-readable codebook. SWEDISH ELECTION STUDY. 1964 (SSD 0007) Principal investigator: Bo Sarlvik, Department of Polidcal Science, University of Gothenburg. Total sample: 3,109 Number of respondents: 2,849 Sample loss: 260 (8.4%) Number of variables: 219 Method: Interview in home. One-half of the sample was interviewed before elecdon day, the other half after. Pre-elecdon respondents also answered a short mail questionnaire after the elecdon, which mainly contained quesdons on final vote decision. Panel: The first stage of a three-stage panel study in which the sample was reinterviewed in conjuncdon with the parliamentary elecdons of 1968 and 1970. Format: OSIRIS, SPSS-x, machine-readable codebook. SWEDISH ELECTION STUDY, 0039) 1968 (SSD Summer 1987 18 - iassist quarterly Principal investigator: Bo Sarlvik, Department of Political Science, University of Gothenburg. Total sample: 3,356 Number of respondents: 2,943 Sample loss: 413 (12.3%) Number of variables: 532 Method: Interview in home. One-half of the sample was interviewed before election day, the other half after. Pre-election respondents also answered a short mail questionnaire after the election, which mainly contained questions on final vote decision. Panel: The second stage of the three-stage 1964-1968-1970 panel study. Format: OSIRIS, SPSS-x, machine-readable codebook. SWEDISH ELECTION STUDY, 1970 (SSD 0047) Principal investigator: Bo Sarlvik, Department of Political Science, University of Gothenburg. Total sample: 4,815 - Interview in home 1,602 - Telephone interview 1,580 - Mail questionnaire 1,633 Number of respondents: 4,130 - Interview in home 1,355 - Telephone interview 1,407 - Mail questionnaire 1,368 Sample loss: 685 (14.2%) - Interview in home 247 (15.4%) - Telephone interview 173 (10.9%) - Mail questionnaire 265 (16.2%) Number of variables: 223 Method: Three types of interviews: in-home interviews, a somewhat shorter telephone interview, and a mail questionnaire with only a small number of questions. All interviews were conducted after the elecuon. Panel: The study represents stage three in the 1964-1968-1970 panel. Format: OSIRIS, SPSS-x, machine-readable codebook. SWEDISH ELECTION STUDY, 1973 (SSD (HMO) Principal investigators: Bo Sarlvik and Olof Petersson, Department of Political Science, University of Gothenburg. Total sample: 3,179 Number of respondents: 2,596 Sample loss: 583 (18.3%) Number of variables: 239 Method: Interview in home. One-half of the sample was interviewed before elecuon day, the other half after. Pre-election respondents also answered a short mail questionnaire after the elecuon, which mainly contained questions on final vote decision. Panel: The study represents stage one in the 1973-1976 panel. Format: OSIRIS, SPSS-x, machine-readable codebook. SWEDISH ELECTION STUDY, 1976 (SSD 0008) Principal investigator: Olof Petersson, Department of Political Science, University of Uppsala. Total sample: 3,580 Number of respondents: 2,652 Sample loss: 928 (25.9%) Number of variables: 290 Weight: Respondents belonging to the 1973-1976 panel who did not respond in 1973 had sample probability halved in 1976; when processing the data, these were weighted by a factor of 2. Method: Interview in home. One-half of the sample was interviewed before election day, the other half after. Pre-election respondents also answered a short mail questionnaire after the election, which mainly contained questions on final vote decision. Panel: The study represents stage two in the 1973-1976 panel and stage one in the 1976-1979 panel. Format: OSIRIS, SPSS-x, machine-readable codebook. SWEDISH ELECTION STUDY, 1979 (SSD 0089) Summer 1987 lassist quarterly - 19 Principal investigator: SOren Holmberg, Department of Political Science, University of Gothenburg. Total sample: 3,498 Number of respondents: 2,816 Sample loss: 682 (19.5%) Number of variables: 303 Weight: Respondents belonging to the 1976-1979 panel who did not respond in 1976 had sample probability halved in 1979. These are represented by two records in the data file. The number of respondents is therefore 2,905 and the sample loss is 853. Method: Interview in home. One-half of the sample was interviewed before election day, the other half after. Pre-election respondents also answered a short mail questionnaire after the election, which mainly contained questions on final vote decision. Panel: The study represents stage two in the 1976-1979 panel and stage one in the 1979-1982 panel. Format: OSIRIS,' SPSS-x, machine-readable codebook. SWEDISH ELECTION STUDY, 1982 (SSD 0157) Principal investigator: SOren Holmberg, Department of Political Science, University of Gothenburg. Total sample: 3,597 Number of respondents: 2,943 Sample loss: 654 (18.2%) Number of variables: 303 Weight: Respondents belonging to the 1979-1982 panel who did not respond in 1979 had sample probability halved in 1982. These are represented by two records in the data file. The number of respondents is therefore 2,980 and the sample loss is 744. Method: Interview in home. One-half of the sample was interviewed before election day, the other half after. Pre-election respondents also answered a short mail questionnaire after the election, which mainly contained questions on final vote decision. Panel: The study represents stage two in the 1979-1982 panel and stage one in the 1982-1985 panel. Format: OSIRIS, SPSS-x, machine-readable codebook. Publications A number of books, reports, and papers have been published based on the results of the election studies. The following list does not claim to be comprehensive.n Bibliography Asp, Kent Journalisternas inflytande i valkampen. Nordicom-Information ni. 3-4, 1983. Asp, Kent, Per Hedberg, and Peter Esaiasson. MassmedieundersOkningen. FolkomrOstningen 1980. Teknisk rapport [TS] GOteborg: Statsvetenskapliga instituuonen, GOteborgs universitet 1984. Asp, Kent Rikspoliiik och kommunalpolitik. ValrOrelsernas utrvmme i svensk dagspress 1956-1985. (Statens offentliga utredningar, 1987:6) Stockholm, 1987. Asp, Kent. The struggle for the agenda - party agenda, media agenda and voter agenda in the 1979 Swedish election campaign. Communication research 10(3):333-355, 1983. Asp, Kent Sveriges Radio och 1982 ars valrOrelse. [TS] GOteborg: Statsvetenskapliga institutionen, GOteborgs universitet 1983. Asp, Kent V'aljarna och massmediernas partiskhet in: Valiare parti er massmedia . Stockholm: LiberfOrlag. 1982. Clausen, Aage, SOren Holmberg and L. Dehavens. Contextual factors in the accuracy of leader perception of constituents' views. Summer 1987 20 - iassist quarterly Journal of politics 45(2):449-472, 1983. Esaiasson, Peter. Partiledarna infor valiarna . (Forskningsrapport 1985:4) Goteborg: Statsvetenskapliga institutionen, Goteborgs universitet, 1985. Esaiasson, Peter and Mikael Gilljam. V'aljarna och de vilda strejkerna. Statsvetenskaplig Udskrift 86(3): 197-207, 1983. Georgsson, Ake. Den unge v'aljaren och poliUken. Lie. diss., Statsvetenskapliga institutionen, Goteborgs universitet, 1973. Gilljam, Mikael and Lennart Nilsson. Svenska folket och den offenUiga sektorn . (Rapportserien 1984:2) Goteborg: Statsvetenskapliga institutionen, Goteborgs universitet, 1984. Gilljam, Mikael and Lennart Nilsson. Svenska folkets asikter om den offentliga sektorn - tva fOrklaringsansatser. Statsvetenskaplig udskrift 88(2): 123-139, 1985. Gilljam, Mikael. V'aljarna och v'arldspoliUken. Statsvetenskaplig Udskrift 87(2), 1984. Granberg, Donald and SOren Holmberg. Modeling the relaUonships among preference, expectauons and voUng behavior . (Rapportserien 1983:3) Goteborg: Statsvetenskapliga insUtuUonen, Goteborgs universitet, 1983. Granberg. Donald and Soren Holmberg. Prior behavior, recalled behavior, and the predicUon of subsequent voUng in Sweden and the United States. Human relations 39(2): 135-148, 1986. Holmberg, SOren. Facklig akuvitet bland kvinnor och man. in: Jamstalldhetsperspektiv i forskningen . Rapport fran ett symposium arrangerat av Riksbankens Jubileumsfond. April 1980 (RJ 1980:4) Holmberg, Soren, Hans Nordlof and Mikael Gilljam. FolkomrOstningsundersOkningen 1980. Teknisk rapport . (ValundersOkningar, Rapport 7) Stockholm: LiberFOrlag, 1984. Holmberg, SOrcn, and Kent Asp. Kampen om karnkraften. En bok om valiare, massmedier och folkomrOstningen 1980 . (ValundersOkningar, Rapport 6) Stockholm: LiberFOrlag, 1984. Holmberg, SOren. Karnkraften och 1980-talets poliUska konflikter. in: Moderna Uder . Rapport fran en symposium om teknik, poliUk och samhallsdebatt arrangerat av Riksbankens Jubileumsfond, mars 1979. (RJ 1979:4) Holmberg, SOren and Olof Petersson. Marxismen och folkbegreppel [TS] Goteborg: Statsvetenskapliga insUtuUonen, Goteborgs universitet, 1972. Holmberg, SOren and Olof Petersson. Nagra anteckningar om klassbegreppet. [TS] Goteborg: Statsvetenskapliga insUtuUonen, Goteborgs universitet, 1971. Holmberg, SOren. PoliUska sakfragor och den representaUva demokraUn. [TS] Goteborg: Statsvetenskapliga insUtuUonen, Goteborgs universitet, 1971. Holmberg, SOren. 'Riksdagen representerar svenska folket'. Empiriska studier i representaUv demokraU. diss., Statsvetenskapliga insUtuUonen, Goteborgs universitet (Lund: StudenUitteratur) 1974. Holmberg, SOren. Riksdagsm'an och valiare . Stockholm: Forum for samhallsdebatt, 1974. Holmberg, SOren. Svenska valiare . (ValundersOkningar, Rapport 4) Stockholm: LiberFOrlag, 1981. Holmberg, SOren. Valet 1979. in: Allmanna valen 1979 . Part 3. (Sveriges officiella staUsUk) Stockholm: StaUsuska Centralbyran, 1981. Holmberg, SOren and Hans Nordlof. ValundersOkning 1979. Teknisk rapport. (ValundersOkningar, Rapport 5) Stockholm: LiberFOrlag, 1981. Holmberg, SOren and Mikael Gilljam. ValundersOkning 1982. Teknisk rapport . (ValundersOkningar, Rapport 9) Stockholm: LiberFOrlag, 1985. Holmberg, SOren, Mikael Gilljam and Maria Oskarson. ValundersOkning 1985. Teknisk rapport. (ValundersOkningar, Rapport 11) Stockholm: LiberFOrlag, 1987. Holmberg, SOren. Valiare i fOrandring . (ValundersOkningar, Rapport 8) Stockholm: LiberFOrlag, 1984. Summer 1987 [assist quarterly - 21 Holmberg, SOren and Mikael Gilljam. Valiare och val i Sverige . (Valundersokningar, Rapport 10) Stockholm: Bonniers [sic], 1987. Holmberg, Sbren. Valjarna och lOntagarfonderna. in: Valiare partier massmedia . Stockholm: LiberForlag, 1982. Oscarsson, VilgoL Politiskt deltagande. En jamfOrande studie av politisk participation i Sverige, USA, Storbritannien och Vasttyskland. Lie. diss., Statsvetenskapliga institutionen, Goteborgs Universitet, 1973. Oskarson, Maria. Yrkes- och socialgrappsklassificeringar i valundersOkningama 1956-1985. [TS] Goteborg: Statsvetenskapliga institutionen, Goteborgs universitet, 1986. Petersson, Olof. Klassidentifikation. in: Valiare partier massmedia . Stockholm: LiberForlag, 1982. Petersson, Olof. New trends in the Swedish electorate: a focus on the 1976 election, paper presented at the ECPR workshop "Social structure and political change". Grenoble, April 1978 Petersson, Olof. Social class and electoral change: Sweden. 1956-1973. [TS] Uppsala: Statsvetenskapliga institutionen, Uppsala universitet. 1975. Petersson, Olof. Stormaktspolitik i svensk opinion. Internationella studier 1975(4). Petersson, Olof. Stormaktspolitik i svensk opinion 1973-1976. Internationella studier (1):9-12, 1978. Petersson, Olof. ValundersOkningen 1976. Teknisk rapport (Valundersokningar, Rapport 3) Stockholm: LiberForlag, 1978. Petersson, Olof and Bo Sarlvik. Valet 1973. in: Allmanna valen 1973 . (Sveriges officiella statistik) Stockholm: Statistiska Centralbyran, 1975. Petersson, Olof. Valet 1976 in: Allmanna valen 1976 . Part 3. (Sveriges officiella statistik) Stockholm: Statistiska Centralbyran, 1978. Petersson, Olof. Valiarna och valet 1976 . (Valundersokningar, Rapport 2) Stockholm: LiberForlag, 1977. Petersson, Olof. Valiarna och varldspolitiken . (Statsvetenskapliga fOreningen i Uppsala, nr. 90) Stockholm: Norstedt & SOners fOrlag, 1982. Petersson, Olof. Valjarnas Okade osakerheL [TS] Uppsala: Statsvetenskapliga institutionen, Uppsala universitet, 1978. Petersson, Olof. V'anster-hOger dimensionen och variationer i politisk participation. [TS] GOteborg: Statsvetenskapliga institutionen, Goteborgs universitet, 1973. Petersson, Olof. The 1973 general election in Sweden. Scandinavian political studies 9:219-228, 1974. Sainsbury. Diane. Class voting and left voting in Scandinavia. [TS] Stockholm: Statsvetenskapliga institutinen. Stockholms universitet, 1986. Smith, L, Aage Clausen, and SOren Holmberg. Perceptual accuracy as a function of legislative attitudes and the distribution of constituency views. Paper presented at the 1981 Annual Meeting of the American Political Science Association, New York, September 1981. Sarlvik, Bengl Stabila valjare och partibytare. Statsvetenskapli e tidskrift (3):121-151, 1975. Sarlvik, Bo. Det kommunala sambandet och valjaropinionen. En studie rOrande valjarnas information och attityder i fOrfattningsfragor. [TS] Goteborg: Statsvetenskapliga institutionen, Goteborgs universitet, 1965. Sarlvik. Bo. Electoral behavior in the Swedish multiparty system, diss., Statsvetenskapliga institutionen, Goteborgs universitet, 1970. Sarlvik, Bo. IntervjuundersOkningen. in: Allmanna val. Riksdagsmannavalen 1959-1960 . Part 2. (Sveriges officiella statistik) Stockholm, Stockholm: Statistiska Centralbyran, 1961. Sarlvik. Bo. IntervjuundersOkningen. in: Allmanna val, Riksdagsmannavalen 1961-1964 . Part 2. (Sveriges officiella statistik) Stockholm: Statistiska Centralbvran, 1965. Sarlvik, Bo. lnteryjuundersOkning rOrande valet till andra kammaxen 1968. in: Allmanna val. Riksdagsmannavalen 1965-1968. Part 2. Summer 1987 22 - iassist quarterly (Sveriges officiella statislik) Stockholm: Statistiska Centralbyran, 1970. Sarlvik, Bo. Qpinionsbildningen vid folkomrOstningen. 1957 . (Statens ofTentliga utredningar, 1959:10) Stockholm, 1959. Sarlvik, Bo. Paitibyten som matt pa avstimd och dimensioner i partisystemet. Sociologisk forsknine 5(l):35-80, 1968. Sarlvik, Bo. Party politics and electoral opinion formation: a study of issues in Swedish politics 1956-1960. Scandinavian political studies vol. 2, 1967. Sarlvik, Bo. Party profiles in the Swedish electorate. A collection of data drawn from the 1964 elecuon study. [TS] Paper presented at the Conference on Comparability in Voting Studies, Loch Lomond, Scotland, Julv 1968. Sarlvik, Bo. Political stability and change in the Swedish electorate. Scandinavian political studies (l):188-222, 1966. Sarlvik, Bo. Politisk rOrlighet och stabilitet i valmanskaren. Statsvetenskaplig tidskrift 67(4): 185-219, 1964. Sarlvik, Bo. The relation of partisan orientation to political engagement in the Swedish muluparty system. [TS] Paper presented at the International Conference on Comparative Bectoral Behavior, University of Michigan, Ann Arbor, Mich., April 1967. Sarlvik. Bo and Olof Petersson. Rikspolitik och iokalpolitik i valet 1973 . (ValundersOkningar, Rapport 1) Stockholm: LiberFOrlag, 1975. Sarlvik, Bo. Skiljelinjer i valmanskaren, Statsvetenskaplie tidskrift 68(2-3): 141-183. 1965. Sarlvik, Bo. Socioeconomic determinants of voting behavior in the Swedish electorate. Comparative political studies 2(1):9 C>-135, 1969. Sarlvik, Bo. Socioeconomic position, religious behavior, and voting in the Swedish electorate - application of computerized classification techniques. Quality and quantity 4(1):95-116, 1970. Sarlvik, Bo. Socio-economic predictors of voting behavior: research notes from a study of political behavior in Sweden. [TS] Paper presented at the Third International Conference of the ISA Research Committee on Political Sociology, Berlin, January 1968. Sarlvik, Bo. Sweden: the social bases of the parties in a developmental perspective, in: Richard Rose (ed.)/ Electoral behavior: a comparative handbook . New York, NY: Free Press, 1974. Sarlvik, Bo. The Swedish elecuon study in 1968: a note on survey design & draft questionnaire. [TS] Paper presented at the Conference on Comparability' in Voting Studies, Loch Lomond, Scotland, July 1968. Sarlvik. Bo. The Swedish party system in a developmental perspective. [TS] GOteborg: Statsvetenskapliga institutionen, Gbteborgs universitet, 1971. Sarlvik, Bo. Valet 1970. in: Allmanna valen 1970 . Part 3. (Sveriges officiella statistik) Stockholm: Statistiska Centralbyran, 1973. Sarlvik, Bo. Valrorelse och valjarna. Tiden 47(6):334-340, June 1955. Sarlvik, Bo. Voting behavior in shifting 'election winds'. An overview of the Swedish elections. 1964-1968. Scandinavian political studies 5:241-283, 1970. Westerstahl, JOrgen, Bo Sarlvik, and EsbjOrn Janson. An experiment with information pamphlets on civil defense. Public opinion quarterly 25:236-248, Summer 1961. Westerstahl. JOrgen, Bo Sarlvik, and EsbjOrn Janson. Fragor rOrande civilforsvar och psykologiskt fOrsvar. [TS] Beredskapsnamnden for psvkologiskt fOrsvar. 1958. Westerstahl. JOrgen and Bo Sarlvik. Svensk hjalp till mindre utvecklade omr'aden. En attitydundersOkning. [TS] Centralkommitten for hjalp till mindre utvecklade omraden, 1957. Westerstahl, JOrgen and Bo Sarlvik. Svensk valrorelse 1954. Tva lokala studier. Arbetsrapport I. [TS] GOteborg: Statsvetenskapliga institutionen, Goteborgs universitet. 1955. Westerstahl. JOrgen and Bo Sarlvik. Svensk Summer 1987 iassist quarterly — 23 valrOrelse 1954. lcke-rOstning - mediastudier. Arbetsrapport II. [TS] GOteborg: Statsvetenskapliga institutionen, GOteborgs universitet, 1956. Westerstahl, Jbrgen and Bo Sarlvik. Svensk valrorelse 1956. Arbetsrapport I: IntervjuundersOkningen. [TS] Goteborg: Statsvetenskapliga institutionen, Goteborgs universitet, 1957. Summer 1987