Name: The Culture and Leisure Time study (Fritidsundersökningen)
Period(s): 1985 (SMS) Code book: III

Information
This text is primarily based on earlier texts from the Project Metropolitan by Carl-Gunnar Janson. The sections concerning the compilation of this codebook are by Therese Sehlstedt. The texts by Janson from 1986 have been translated from Swedish.

When Project Metropolitan started in 1964 the original plan was to follow the cohort until the age of 30 (1983), and then finish with an interview or questionnaire study (Janson, 1984). The concluding survey was delayed until spring 1985, when it was carried out by the researchers working on the project from the Department of Sociology at Stockholm University and the Listener and Programme Research Department of Swedish Radio (PUB)1. The field work for the Culture and Leisure Survey was carried out in March and April, 1985 (Janson, 1986a). The questions were concerned with cultural and leisure habits such as watching television, listening to the radio, reading newspapers, recreational activities, household composition etc.

The cohort studied in Project Metropolitan consists of those who were born in 1953 and were registered as living in the Stockholm Metropolitan area on 1 November, 1963. The total cohort consists of 15,117 persons, 7,398 females and 7,719 males (Janson, 1980). The Culture and Leisure Survey was sent out to a selection of the cohort members, using the same weighted sample as for the family study that was carried out in 1968 (Janson, 1986a). This sample of 4,021 was drawn from the cohort in 1968. To create the sample for the family study the cohort was divided into five strata. Firstly, the population members still in the metropolitan area on 1 November 1967 were listed. Some 525 members had been lost since 1 November 1963. Of those who remained, those who had complete mental test scores from the school study were distributed into three strata (High, Medium, and Low) according to score. The ‘high’ stratum comprised those with the five per cent highest scores, and the ‘low’ stratum held those with the five per cent lowest scores. The ‘medium’ stratum covered the intermediate 90 percent. The cutting points on the score line were fixed separately for girls and boys and duly corrected for those few who had done the test in 1967 instead of in 1966. Both end strata became a little larger than 5 percent. Those who had not participated in the school study were put in a fourth, non-response stratum. Finally, a small fifth stratum contained those with incomplete test scores. In the high and low strata all members were included in the sample. In each of the other strata every fifth was drawn. The total sample comprised 4,021 cohort members, 1,972 girls and 2,049 boys (Janson, 1980).
As the sample used for this study had been created in 1968, for the purpose of the Family Study, some further investigation was carried outmade by the researchers before sending out the surveys. Of the total sample (4,021 persons) 21 had left Greater Stockholm. Moreover, 3,939 of them where registered as living in Sweden in February 1985; 69 of these individuals had no specified address. Thus, 3,870 cohort members remained (1,975 males and 1,895 females), for the researchers to try to get in contact with for the survey. The stratified sample from the Family Survey, with cohort members who had high and low mental test scores from the school study overrepresented, corresponded in 1968 to 96.5 percent of the cohort members after re-evaluation. The 3,870 remaining in 1985 represented 90.9 percent of the cohort (Janson, 1986a).

As mentioned earlier, the field work was carried out in March and April 1985. The study took the form of a mail questionnaire, with complementary telephone interviews with some of those who had not answered after two reminders. The questionnaire, the letter of introduction and two reminder letters were formulated in the autumn of 1984 and in January and February 1985. The field work (sending out and receiving the questionnaires, carrying out the telephone interviews) was performed by the Listener and Programme Research Department of Swedish Radio (PUB), while the other parts of the study were performed by the researchers working on the project (Janson, 1986a). The mail questionnaire was sent out on 18 March, and a reminder was sent to everyone on 22 March, followed by a second reminder to those who had still not answered on the 29th. After three weeks, 2,390 identified respondents had answered the survey completely or partially, of whom 1,228 were female and 1,162 male. The researchers then began identifying remaining respondents for telephone interviews. Another 469 questionnaires were answered in this way (208 women and 261 men). This gave a total of 2,859 identified respondents, of whom 1,436 were female and 1,423 male. (Janson, 1986a; 1986b).

The response rate is accounted for by a variable wherein the first category represents those who had answered the mail questionnaire and the second category those who had been interviewed via telephone. The original version of this variable shows that 2,377 respondents answered the survey by mail and 465 by telephone. The response rate totals 2,843, which does not correspond to the numbers documented in 1986. This matter has been investigated further when working on this code book. One observation has been deleted, as the person was not included in the original sample of 4,021, and must have been sent the mail questionnaire by mistake. This left us with 2,377 and 465, adding up to 2,842. When the various kinds of nonresponse are added we get 3,923, which leaves us with 99 persons not accounted for by the variable svar. As those 99 should have been accounted for by various kinds of non-response we then used variables that measure deaths and residence to address the discrepancy. The findings resulted in some changes in the original variable. The number of missing responses was reduced from 99 to 39, and the changes are best understood by comparing the old and the new variable.

When available data and the response-rate reported in 1986 are compared there is a total discrepancy of 17. The most likely explanation for this is that at some point (for reasons unfortunately not found in the documents) it was decided that these people should be deleted. This could have been because surveys were only partially answered or because telephone interviews were interrupted, as was recorded in the report in 1986 (Janson, 1986a). Irrespective of the cause, this introduces the question of validity and reliability. A comparison was accordingly carried out with the frequency tables reported in 1986 (Janson, 1986b). The result of this evaluation indicates that the quality of the available data remains good; the weighted frequency table corresponds well with those reported in 1986. The comparison was performed by looking at the frequency in percentage with one decimal (as it was reported in 1986). Most variables, and their categories, demonstrate exact accuracy. In some cases numbers deviate but only by one decimal, with the exception of two categories at different variables which deviate at most by .5 decimals. This high rate of correspondence indicates that the missing 18 people are not a problem when using the available data for analysis. Some of the questions in the survey are open questions, such as those about which television programs, radio programs, films and books the respondent likes. We do not, unfortunately, have any detailed information about how some of these questions were dealt with in the original coding work. However, the distributions of the answers are partly accounted for in the report that was sent out to the respondents after the survey. The questions of an open character are accounted for by reporting frequencies over 0.5 percent, and were in some cases categorised together in various subjects, i.e. not all answers or codes were reported. It has to some extent been possible to ascertain the codes for these particular questions (numbers 2, 4 and 9) by looking at the reported percentages and matching them with the variables in the data. The data was weighted by the variable fam8 before this was done, to take the selective sample for the study to account. Frequencies exceeding 0.5 percent were confirmed by this method. For Question 11 “Mention a GOOD book you have read in your spare time in the last 12 months”, descriptive statistics are presented instead of frequencies as the method described above did not prove useful; the compared percentages did not correspond with each other. As the sample used for this study is the stratified sample from the Family Survey, in which birth cohort members with high and low mental test scores from the school study are overrepresented, the data needs to be weighted. Using the variable “fam8” accounts for the overrepresentation, and gives a new weighted sample proportional to the original total sample of 15 117 individuals, i e the answers of interviewees in the intermediary category and the two residual categories have been multiplied by 5. Using the variable “vikt1” in the analyses also accounts for the overrepresentation in mental test scores, but gives a new weighted sample proportional to the size of the subsample.

References

Janson, C-G. (1980), The Family Study – A Code Book, Project Metropolitan. Research Report No 4, Department of Sociology. Stockholm: Stockholm University.

Janson, C-G. (1986a), Projekt Metropolit: Årsrapport för 1985. [Project Metropolit: Annual Report for 1985]. Not published.

Janson, C-G. (1986b), Sammanställning av frekvenstabuleringar samt introduktionsbrev till utskicket sänt till deltagare i undersökningen som önskade ta del av svarsfördelningarna för de olika frågorna. [Compilation of frequencies and introductory letter sent to those taking part in the survey who requested information about the results for the different questions]. Not published.