Nordic adolescents responding to demanding survey scales in boring contexts : Examining straightlining

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY-NC-ND 4.0 https://creativecommons.org/licenses/by-nc-nd/4.0/ Nordic adolescents responding to demanding survey scales in boring contexts : Examining straightlining © 2022 the Authors Published version Aarø, Leif Edvard; Fismen, Anne‐Siri; Wold, Bente; Skogen, Jens Christoffer; Torsheim, Torbjørn; Arnarsson, Ársæll Már; Lyyra, Nelli; Löfstedt, Petra; Eriksson, Charli Aarø, L. E., Fismen, A., Wold, B., Skogen, J. C., Torsheim, T., Arnarsson, Á. M., Lyyra, N., Löfstedt, P., & Eriksson, C. (2022). Nordic adolescents responding to demanding survey scales in boring contexts : Examining straightlining. Journal of Adolescence, 94(6), 829-843. https://doi.org/10.1002/jad.12066 2022 Received: 5 April 2022 | Accepted: 6 June 2022 DOI: 10.1002/jad.12066 RESEARCH ARTICLE Nordic adolescents responding to demanding survey scales in boring contexts: Examining straightlining Leif Edvard Aarø 1,2 |Anne‐Siri Fismen 1,3 |Bente Wold 2 | Jens Christoffer Skogen 1,3,4 |Torbjørn Torsheim 5 |Ársæll Már Arnarsson 6 | Nelli Lyyra 7 |Petra Löfstedt 8 |Charli Eriksson 9 1 Department of Health Promotion, Norwegian Institute of Public Health, Bergen, Norway 2 Department of Health Promotion and Development, University of Bergen, Bergen, Norway 3 Centre for Evaluation of Public Health Measures, Norwegian Institute of Public Health, Oslo, Norway 4 Alcohol and Drug Research Western Norway, Stavanger University Hospital, Stavanger, Norway 5 Department of Psychosocial Science, University of Bergen, Bergen, Norway 6 Faculty of Health Promotion, Sport and Leisure Studies, University of Iceland, Reykjavík, Iceland 7 Faculty of Sport and Health Sciences, University of Jyväskylä, Jyväskylä, Finland 8 Department of Public Health and Community Medicine, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden 9 Department of Learning, Informatics, Management, and Ethics, Karolinska Institute, Stockholm, Sweden Correspondence Leif Edvard Aarø, Department of Health Promotion, Norwegian Institute of Public Health, Zander Kaaes gt. 7, 5015 Bergen, Norway. Email: [email protected] Funding information Folkhälsomyndigheten Abstract Introduction: Straightlining, or identical responses across all items within a multi‐ item scale, is often taken as an indication that responses to all items in a questionnaire are of poor quality. The purpose of this study was to examine straightlining on two scales: The Sense of Unity Scale (SUS) and the short version of the Warwick‐ Edinburgh Mental Well‐being Scale (SWEMWBS). Methods: Data stem from the 2017–2018 data collections in four Nordic countries of the Health Behaviour in School‐children study (HBSC) (15‐year‐old students only; 50.9% girls; n= 5928). Data were weighted to adjust for oversampling of Swedish‐ speaking Finnish students and to equalize sample size across countries. The main analyses were done with general linear modeling with adjustments for cluster effects (school classes). Results: The proportion with straightlining on SUS was 22.8%, varying from 5.8% among Swedish girls to 46.4% among Finnish boys. The proportion with straightlining on SWEMWBS was 18.4%, varying from 5.2% among Norwegian girls to 46.0% among Finnish boys. Straightlining on one of the scales correlated with straightlining on the other one. Straightlining tended to inflate Cronbach's αvalues and reduce number of factors in factor analyses. Associations between the two scales and external variables tended to be lower among straightlining students. Associations between external variables (other than SUS/SWEMWBS) are on average slightly weaker among straightliners. Straightlining students obtained more favorable scores on several resource‐related variables. Conclusion: Although some problems have been identified, straightlining does not serve well as a general indicator of poor data quality. KEYWORDS adolescents, Nordic countries, questionnaires, scales, straightlining, surveys 1|INTRODUCTION Research on the development of adolescent mental health and quality of life is challenging and studies using high‐quality data are crucial. This has been the emphasis in a collaboration among researchers from the Nordic countries, utilizing data from the 2017/2018 data collection of the Health Behaviour in School‐aged Children study (HBSC) (Eriksson et al., 2019). Journal of Adolescence. 2022;1–15. wileyonlinelibrary.com/journal/jad | 1 This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non‐commercial and no modifications or adaptations are made. © 2022 The Authors. Journal of Adolescence published by Wiley Periodicals LLC on behalf of Foundation for Professionals in Services to Adolescents. This paper focuses on a kind of response set called “straightlining”in relation to two measures of positive mental health: the Sense of Unity Scale (SUS) and the short version of Warwick‐Edinburgh Mental Well‐being Scale (SWEMWBS). “Response bias”refers to factors or conditions that take place during the process of responding to survey questions and which affect the responses and thereby lead to a nonrandom deviation of the answers from their true value (Villar, 2008, p. 751). A subcategory of response biases is “response styles,”defined as “a respondent's tendency to respond to survey questions in certain ways regardless of the content”(Van Vaerenbergh & Thomas, 2013). The definition of “response set”is strikingly similar. There are several kinds of response sets described in the scientific literature, for instance, the acquiescence bias, the bias of extreme responding, bias to the middle, random responses, and straightlining. The occurrence of response bias is related to the length of the questionnaire and types of questions and response categories. If respondents find the questions irrelevant or boring, this might produce less thoughtful answers. This is particularly so if the questions come towards the end of a long questionnaire (Herzog & Bachman, 1981). Adolescence is a period of life characterized by a high level of impatience and impulsivity (Steinberg et al., 2009). Fatigue and boredom when filling out questionnaires may therefore be particularly challenging in surveys among adolescents. A simple and broadly accepted remedy for at least some types of response sets was suggested by Moser and Kalton. “Variation between positive and negative items forces the respondent to consider each item carefully, rather than to respond automatically to them all in the same way”(Moser & Kalton, 1971, p. 362). Changing the order of response categories across a set of items introduces new problems. One challenge is that it becomes more demanding and time consuming to respond to the questions. In many, perhaps most multi‐item scales, the order of response categories is therefore the same for all items. Instead of investing the necessary efforts into providing thorough answers, study participants may simply provide answers that appear to be adequate. Krosnick (1991) coined the term “satisficing”to characterize this kind of responses. Krosnick pointed at six different strategies of “satisficing”: (i) selecting the first response alternative that seems reasonable, (ii) agreeing with assertions, (iii) endorsing the status quo (not particularly relevant in our context), (iv) non‐differentiation in using rating scales, (v) saying “Don't know,”and (vi) mental coin‐flipping. Krosnick hypothesized that the risk of satisficing increases with the difficulty of the scale and decreases with study participants' ability and motivation. To the extent that the scale is cognitively demanding, and study participants are unmotivated and lack cognitive abilities, satisficing responses will be more common. “Non‐differentiation”or “straightlining”in using rating scales is defined as selecting a response category on the first item, and then using the same response category for all remaining items (Krosnick, 1991; Loosveldt & Beullens, 2017). Straightlining has been found to be more common among respondents with low levels of education and among respondents who have low levels of verbal abilities, and straightlining is more common towards the end of questionnaires (Herzog & Bachman, 1981; Krosnick, 1991). Some researchers, for instance Yan (2008), regard straightlining to be a source of measurement error which contributes to decreasing quality of data. Reuning and Plutzer (2020) are, however, critical of the practice of regarding the occurrence of straightlining as an indicator of low study quality. They are also critical of the practice of removing respondents identified as straightliners from data based on the assumption that their answers are meaningless. They distinguish between valid versus invalid straightlining: “Valid straightlining occurs when two conditions are satisfied: A respondent is motivated to carefully read/listen to questions and answer them truthfully, and provides identical (non‐differentiated) responses to a sequence of questions”(Reuning & Plutzer, 2020, p. 440). Invalid straightlining occurs when items are not examined carefully, and responses are provided without sufficient attention and cognitive processing. The likelihood that a study participant will provide identical responses to all items decreases with increasing numbers of items and response alternatives. In the 2017/2018 data collection of the WHO Collaborative Study on Health Behaviour in School‐aged Children (the HBSC Study), there were two scales that we will use for a closer inspection of straightlining—a seven items version of the Warwick‐Edinburgh Mental Well‐being Scale (SWEMWBS) (Ringdal et al., 2017; Stewart‐Brown et al., 2009) and the Sense of Unity Scale (SUS) which comprises eight items (Eriksson et al., 2019; Samdal et al., 2016). All items in both scales have five response categories. The response categories and their order are identical within scales. Both scales are relatively abstract and cognitively demanding. Since identical scales were used across four countries, the total number of observations was sufficient for the analyses carried out in this study. The data at hand are therefore rather ideal in a study of straightlining. As mentioned above, with reference to Reuning and Plutzer (2020), identical responses across all items are not necessarily an indication of bias. A young person may examine each item of the SWEMWBS scale items carefully and come to the exact same conclusion for each item, for instance that the answer is always “Some of the time,”which is the third response category out of five possible. If the proportion of students with this kind of uniform answers across items is very high, or varies strongly across groups of students, this may indicate bias at the group level. Associations with third variables which deviate strongly from similar associations among those who have not practiced any straightlining at all may be another indication of response bias. If straightlining means that responses are arbitrary (characterized by some degree of randomness), lower correlations between scores on the scale where straightlining has taken place and other variables can be expected. 2 | AARØ ET AL. The purposes of the present paper are to (i) examine the occurrence of straightlining among students in four Nordic countries (Finland, Iceland, Norway, and Sweden) and (ii) differences in the prevalence of straightlining across genders and countries. We will also (iii) examine effects of straightlining on the dimensionality and internal consistency of scales, and (iv) examine differences in associations with external variables among straightliners and non‐straightliners. Finally, we want to (v) describe differences between straightliners and non‐straightliners on several resource‐related variables and (vi) compare intercorrelations and associations among external variables among straightliners and non‐straightliners. For researchers using survey methodology straightlining is an important issue. If straightlining is a strong indicator of poor quality of survey responses, steps should be taken to reduce the risk of straightlining when new data are collected. And when data are analyzed, examining effects of straightlining on data quality may turn out to be imperative. 2|MATERIALS AND METHODS 2.1 |The HBSC study This study is based on data from the HBSC, a World Health Organization (WHO) collaborative cross‐national survey with an overall aim to generate increased understanding of health, health behavior, and their contexts in the lives of young people aged 11, 13, and 15 years. The present study includes samples of 15 years old Finnish (n= 1146), Icelandic (n= 2195), Norwegian (n= 874), and Swedish (n= 1606) students participating in the HBSC study for the survey year 2017/2018. 2.2 |Samples and data collection In Finland, Norway, and Sweden, samples were drawn randomly by cluster sampling within strata defined by, for instance, grade, geographical area, and language. School classes were used as sampling units, and stratification was proportional. The recommended national sample size per grade was 1500, and the mean age should be 11.5, 13.5, and 15.5 years, where 90% of the sample should fall between +/−6 months of the mean age. Iceland invited all schools in the country and therefore had a larger sample compared to the other countries. As mentioned above, the present study is based on data from the oldest age group (15 years) only. In the HBSC study the English version of the questionnaires is the primary one. Procedures for translation and adaption to other languages have been established. This includes back‐translations and piloting. Questionnaires were made available in Finnish and Swedish in Finland, Swedish in Sweden, in the two versions of Norwegian—Norwegian “bokmål”and Norwegian “nynorsk”—in Norway, and Icelandic in Iceland. The questionnaire was not administered in any of the immigrant group languages or in the Sami language. The Swedish language version of the questionnaire in Finland was not identical to the Swedish language questionnaire used in Sweden, but simply a Swedish language version of the Finnish language questionnaire. The two Norwegian languages are too similar to represent any challenge with regard to translations. The students answered the questionnaire at school, after receiving instructions from their teacher. Oral and written information on the confidentiality of their responses were provided and participation was confidential and voluntary. In Finland, Iceland, and Norway, the questionnaire was web‐based. In Sweden they used a printed questionnaire for data collection. Data collections took place in class during school hours. Data collections could take place any day of the week, from Monday to Friday, and during any school hour, but the day and hour had to be the same for all students within one school class. A standardized international research protocol was followed to ensure consistency in survey instruments, data collection, and processing procedures. Response rates for schools/students varied across the countries: Finland 47/60%, Norway 15/71%, and Sweden 47/89%. In Iceland the response rate for the whole student population was 49%, but as the largest municipality in the country, Reykjavik, did not allow tracking individual schools, it was impossible to estimate a separate response rate for schools (Inchley et al., 2020). Schools or classes that declined to participate, as well as students absent on the day the survey was carried out, were the two main sources of nonresponse and were not followed up. The items used in the present study were part of optional HBSC packages, included by Nordic countries and part of a Nordic HBSC collaboration on mental health (Eriksson et al., 2019). The HBSC Data Management Centre, located at the University of Bergen, Norway, checked the quality of the data collected, performed appropriate cleaning of the data, and merged national data sets into an international data file. Detailed information about the study is available at http://www.hbsc.org/. The methodology for data collection is described in the HBSC protocol (Inchley et al., 2018), which prescribes consistency in sampling plans, survey instruments, and data collection. JOURNAL OF ADOLESCENCE | 3 2.3 |Measures The Sense of Unity Scale was developed within the HBSC study and refers to the positive feeling of being part of a larger social structure (Wahlström et al., 2021). The measure was piloted in Norway in 2016 and included as part of an optional package in the international survey in 2017/2018. The pilot study reports a high internal consistency (Cronbach's α=.89; Samdal et al., 2016) and a qualitative study in the form of a classroom discussion (28 students aged 13–14 years) about all the items indicated high validity and good understanding of the items. The Sense of Unity scale comprises eight items: “I feel…”;“a strong sense of togetherness”;“responsibility for others”; “that it is good to be part of a community”;“that I contribute without expecting anything in return”;“that I mean a lot to others”;“that others mean a lot to me”;“that I trust most people”; and “that we rely upon each other.”The response alternatives were “Not at all true for me,”“Not very true for me,”“Neither true nor untrue for me,”“Somewhat true for me,”and “Really true for me.”The wording of the response categories varied slightly across the four countries. Categories were coded 1 to 5, and these scores were summed to an index ranging from 8 to 40, higher values indicating higher sense of unity. In the analyses presented in this publication, a mean score was used instead, the sum score was divided by the number of items (range 1–5). In the context of this study, it is important to notice that in three countries (Finland, Iceland, and Norway), the SUS had a rather wordy introduction: “Everyone belongs to different social groups and communities such as school classes, neighborhoods, sports teams, towns, countries, and wider society. Rate how true each statement regarding how you feel about belonging to such groups and communities is for you ranging from ‘Not at all true for me’to ‘Really true for me’. Mark the box under the rating that best applies to you. I feel …”. In the Swedish questionnaire most of this introduction was removed. Only the words “I feel …” (“Jag känner …”) remained. In the HBSC data collections in the countries included in this study, the seven‐item version of the Warwick‐Edinburgh Mental Well‐being Scale (SWEMWBS) was used: “I've been feeling optimistic about the future,”“I've been feeling useful,”“I've been feeling relaxed,”“I've been dealing with problems well,”“I've been thinking clearly,”“I've been feeling close to other people,”and “I've been able to make up my own mind about things.”The response alternatives were: “None of the time,”“Rarely,”“Some of the time,”“Often,”and “All of the time”(Tennant et al., 2007). Categories were coded 1–5, and a mean score with higher values indicating higher levels of mental well‐being was constructed. Straightlining was in this study defined as giving the same response across all items within one scale. This means that straightlining could only be detected among those who had valid answers on all scale items. A number of questions and scales served as external variables which were analyzed against SUS and SWEMWBS. Analyses of associations among these external variables were also carried out. Most of these variables belong to the domains of mental health, quality of life, or interpersonal relationships. For each multi‐item scale described below, mean scores were constructed. Life Satisfaction was a single item measure: “Here is a picture of a ladder. The top of the ladder ‘10’is the best possible life for you and the bottom ‘0’is the worst possible life for you. In general, where on the ladder do you feel you stand at the moment? Tick the box next to the number that best describes where you stand.”The students were instructed to select one of the boxes which were drawn between the steps of a ladder with zero at the bottom and ten at the top (Cantril, 1965). Previous research has shown good test–retest reliability and validity of this scale among adolescents (Jovanović,2016; Levin & Currie, 2014). Self‐esteem was measured with a scale based on Rosenberg's general self‐esteem model. It has been used in the Danish Methodological Study (Nielsen et al., 2016). Three items are used to measure students' positive conceptions of self and their conceptions of what others think about them: “I like myself,”“ I am good enough as I am,”and “Others my age like me.”Five response categories were used: “Strongly agree,”“Agree,”“Neither agree nor disagree,”“Disagree,”and “Strongly disagree.”Categories were coded 1 to 5 with higher values reflecting high self‐esteem. General Self‐efficacy was a two‐items measure: “How often do you find a solution to a problem if you try hard enough?” and “How often do you manage to do the things that you decide to do?”. Response categories were: “Never,”“Rarely,”“Sometimes,”“Most of the time,”“Always,”coded 1 to 5. (Schwarzer, 1992). This short version has been used in a series of studies (Meilstrup et al., 2020; Nielsen et al., 2016). Loneliness was assessed using a single item on global loneliness. In all countries other than Sweden, loneliness was assessed by the question ‘Do you feel lonely?’with four response categories: 1 “Yes, very often,”2“Yes, quite often,” 3“Sometimes,”and 4 “No.”In Sweden, adolescents were asked ‘How often do you feel lonely?’with five response categories: 1“Always,”2“Most of the time,”3“Sometimes,”4“Rarely,”and 5 “Never.”For Swedish data, the response categories 4 and 5 were combined to better match the response categories used in other Nordic countries. The direct, single‐item measure of loneliness is highly correlated with the multi‐item UCLA Loneliness Scale, which is an indirect measure of loneliness (Lyyra et al., 2021). The response categories were recoded in order to let high values indicate high levels of loneliness. Health complaints were measured with the HBSC Symptoms Checklist. “In the last 6 months: how often have you had the following?”The list of complaints included Headache, Stomachache, Backache, Feeling low, Irritability or bad temper, 4 | AARØ ET AL. Feeling nervous, Difficulties in getting to sleep, and Feeling dizzy. The response alternatives were: “Rarely or never,”“About every month,”“About every week,”“More than once a week,”“About every day”(Hetland et al., 2002). The categories were coded 1 to 5 with higher numbers indicating higher levels of complaints. Self‐rated global health was measured with one item: “Would you say your health is …? Four response alternatives were provided: “Poor,”“Fair,”“Good,”“Excellent.”The categories were coded 1 to 4 with higher numbers indicating better health. Liking school was measured with one item: “How do you feel about school at present?”Response categories were: “I like it a lot,”“I like it a bit,”“I don't like it very much,”and “I don't like it at all.”The categories were coded 1 to 4 with higher numbers indicating stronger liking of school. Relationship with teachers was measured with three items: (i) I feel that my teachers accept me as I am, (ii) I feel that my teachers care about me as a person, and (iii) I feel a lot of trust in my teachers. Response categories were: “Strongly agree,”“Agree,”“Neither agree nor disagree,”“Disagree,”and “Strongly disagree.”The categories were coded 1 to 5, with higher numbers indicating stronger agreement. The cross‐country reliability and validity of the teacher support scale has been demonstrated previously (Torsheim et al., 2000,2012). Family support and Peer support were measured using the family and friend subscales of the Multidimensional Scale of Perceived Social Support (MSPSS) (Zimet et al., 1988). The scale is reliable and valid, as has been shown in studies among various adolescent samples from different countries (Canty‐Mitchell & Zimet, 2000; Edwards, 2010). Peer support was measured with four items: (i) “My friends really try to help me,”(ii) “I can count on my friends when things go wrong,”(iii) ”I have friends with whom I can share my joys and sorrows,”and (iv) “I can talk about my problems with my friends.”Responses were indicated on a scale from 1 to 7 with labels on the two endpoints. “Very strongly disagree” was indicated by the number 1 and “Very strongly agree”by the number 7. Family support:“We are interested in how you feel about the following statements. Please show how much you agree or disagree with each one. Please tick one box for each line.”The following four items were listed: “My family really tries to help me,”“I get the emotional help and support I need from my family,”“I can talk about my problems with my family,”and “My family is willing to help me make decisions.”Study participants could indicate on a seven‐point scale how much they agreed or disagreed with “Very strongly disagree”at one end (coded 1) and “Very strongly agree”at the other end (coded 7) (Zimet et al., 1988). Family affluence Scale III comprised six items: “Does your family own a car, van or truck?”(“No,”“Yes, one,”“Yes, two or more”). Do you have your own bedroom for yourself?”(“No,”“yes”), “How many computers do your family own (including laptops and tablets, not including game consoles and smartphones)”(“None,”“One,”“Two,”“More than two”). “How many bathrooms (room with a bath/shower or both) are in your home?”(“None,”“One,”“Two,”“More than two”). “Does your family have a dishwasher at home?”(“No,”“Yes”). “How many times did you and your family travel out of [insert country here] for a holiday/vacation last year?”(“Not at all,”“Once,”“Twice,”“More than twice”) (Currie et al., 1997; Hartley et al., 2016). A ridit transformation (conversion to cumulative probabilities) by age and sex was carried out on the sum score of the Family Affluence Scale III items. 2.4 |Statistical analyses Data were analyzed with SPSS Version 27. Since school classes were sampling units, adjustments for cluster effects were necessary. Analyses of two‐way frequency tables were carried out in SPSS Complex. Analyses of dimensionality were done with factor analysis with principal axis factoring, pairwise deletion of cases and direct oblimin rotation of factors. General linear modeling (GLM) in SPSS Complex was used for analyses with metric outcome variables (sum scores as well as single‐ item variables) and categorical or metric predictors. In Finland, students from the Swedish‐speaking parts of the country were oversampled. These students constitute 17.8% of the Finnish sample of students in the grade used in this study (the grade corresponding to the age group 15 years). An additional challenge was the strong variation in number of study participants across countries, that is, Iceland, with a small population (about 340,000) has the largest sample (n= 2244), while Norway, with more than 5.5 million inhabitants, has a much smaller sample (n= 876). The data were therefore weighted in such a way that the number of participants for each country is set equal, the proportion of students from the Swedish‐speaking part of Finland is reduced, and the number of students in the total sample remains unchanged. When applying these weights to the data, the number of participants from each of the four Nordic countries included in this study equals 1482, and the total number is 5928. The largest weight is for Norway (1.6918) and the smallest weight is for Swedish‐speaking students in Finland (0.4352). In addition to weighting, adjustments for the cluster effect were necessary. On this point we ran into a challenge. The school authorities of Reykjavik, the capital of Iceland, did not permit registration of schools and classes. Reykjavik therefore had to constitute one large cluster. Adjustments for the cluster effect were therefore only partly possible in the data from Iceland. This limitation of the study may have had consequences at a more detailed level of analysis and some of the comparisons across countries. For analyses based on all data combined, it is unlikely that conclusions would change much JOURNAL OF ADOLESCENCE | 5 even in the absence of this clustering problem. Intraclass correlations for the SWEMWBS and SUS scales with school class as the clustering unit for Iceland did not deviate much from the averages for the Nordic countries combined. The proportion of students with missing observations on at least one of the items for each scale was 6.3% on the Warwick‐Edinburgh Mental Wellbeing scale and 12.4% on the Sense of Unity Scale (weighted numbers). In some analyses students with missing values were included as a separate category. In other analyses, students with missing were excluded. 3|RESULTS 3.1 |Descriptives As shown in Table 1, the distribution across gender is approximately equal in all countries combined with 49.1% boys and 50.9% girls (weighted numbers). There was no significant difference in gender distributions across countries (χ 2 = 0.461; F adjusted = 0.130; df: 2 and 711; p= .928). Percentage distributions for all items of the SUS are shown in Table 2. Most items are moderately left‐skewed with skewness values ranging from −0.59 to −1.18 (weighted). The number of missing answers varied from 439 (7.4%) to 498 (8.4%) (weighted). The proportion with responses to all items on the SUS scale was 87.6%, and 12.4% were with missing on at least one item. The number of study participants who had missing on all items of the SUS was 371 (6.3%) (all numbers weighted). Percentage distributions for all items of the SWEMWBS scale are shown in Table 3. Most items are moderately left‐ skewed with skewness values ranging from −0.39 to −0.77 (weighted). The number of missing answers varied from 177 (3.0%) to 221 (3.7%) (weighted). The proportion with valid answers on all items on the SWEMWBS scale was 93.7%, while the proportion with missing on at least one item on the SWEMWBS scale was 6.3%. The number of study participants who had missing on all items was 147 (2.5%). The mean of the SUS mean score was 3.78 (range 1–5) and the skewness was −0.88. The SWEMWBS mean score (range 1–5) had a mean of 3.63 and the skewness was −0.58. 3.2 |Straightlining The overall proportion of straightliners on the SUS was 22.8% varying from 5.8% among Swedish girls to 46.4% among Finnish boys (Table 4). The variation across countries was significant for both genders (boys: χ 2 PEARSON = 153.354; F ADJUSTED = 37.097; df: 2.798 and 836.604; p< .001). Girls: χ 2 PEARSON = 92.492; F ADJUSTED = 18.945; df: 2.907 and 892.551; p< .001). The proportions of straightliners on the SUS were higher among boys than among girls in all the four countries (Finland: χ 2 PEARSON = 50.849; F ADJUSTED = 39.386; df: 1 and 73; p< .001. Iceland: χ 2 PEARSON = 32.310; F ADJUSTED = 25.412; df: 1 and 100; p< .001. Norway: χ 2 PEARSON = 15.733; F ADJUSTED = 15.694; df: 1 and 64; p< .001. Sweden: χ 2 PEARSON = 32.124; F ADJUSTED = 19.412; df: 1 and 77; p< .001). The overall proportion of straightliners on the SWEMWBS scale was 18.4%, varying from 5.2% among Norwegian girls to 46.0% among (again) Finnish boys (Table 5). The variation across countries was significant for both genders TABLE 1 Respondents by gender and country. Uweighted a . Country Boys Girls Total %%% N unweighted N weighted b Finland 49.5 50.5 100.0 1146 1473 Iceland 49.4 50.6 100.0 2195 1450 Norway 49.0 51.0 100.0 874 1479 Sweden 48.4 51.6 100.0 1606 1438 Total 49.1 50.9 100.0 5821 5839 Note:X 2 Pearson = 0.461; F adjusted = 0.130; df: 2 and 711; p= .928. a Weighting changes the number of observations for each country, totals across countries, percentages for Finland, and significance testing. Weighting has no effect on percentages for Iceland, Norway, and Sweden. b For all countries the total weighted number of participants is 1482. Due to some missing on gender, the numbers in this table are slightly lower and with some variation across countries. 6 | AARØ ET AL. (Boys: χ 2 PEARSON = 227.433; F ADJUSTED = 50.179; df: 2.787 and 833.188; p< .001. Girls: χ 2 PEARSON = 93.131; F ADJUSTED = 22.962; df: 2.822 and 866.460; p< .001). The proportions of straightliners on the SWEMWBS were higher among boys than among girls in all the four countries (Finland: χ 2 PEARSON = 74.339; F ADJUSTED = 74.327; df: 1 and 73; p< .001. Iceland: χ 2 PEARSON = 40.492; F ADJUSTED = 48.343; df: 1 and 100; p< .001. Norway: χ 2 PEARSON = 9.238; F ADJUSTED = 8.101; df: 1 and 64; p= .006. Sweden: χ 2 PEARSON = 52.019; F ADJUSTED = 42.848; df: 1 and 77; p< .001). Among students who straightlined on the SUS, 45.5% straightlined on the SWEMWBS. Among those who did not practice straightlining on the SUS, the proportion practicing straightlining on the SWEMWBS was 10.7%. The difference was significant (χ 2 PEARSON = 726.640; F ADJUSTED = 486.281; df: 1 and 317; p< .001). The ˠcoefficient was 0.750 and ϕwas 0.376. For the SUS, the Cronbach's αvalue was .92, highest for Finnish boys (.96) and lowest among Swedish girls (.84). After exclusion of all straightliners, the overall αfor all groups combined was lower (.86), but still high. In all subgroups defined by gender and country, the αvalue became smaller after exclusion of straightliners. The effect of excluding all straightliners for the SWEMWBS is the same as shown for the SUS with a reduction in Cronbach's αfor all groups combined is from .90 to .84. TABLE 2 Sense of Unity indicators—descriptives and percentage distributions, weighted. Not at all true for me Not very true for me Neither true nor untrue for me Some‐ what true for me Really true for me Total Missing observations Mean SD Skewness % % % % % % N a N a A strong sense of togetherness 3.67 1.01 −0.64 4.2 5.9 29.8 39.1 21.0 100.0 5430 498 Responsibility for others 3.68 0.98 −0.68 3.7 6.0 28.4 42.4 19.4 100.0 5478 450 It is good to be part of a community 3.96 0.98 −0.95 3.1 4.0 19.9 39.9 33.1 100.0 5473 455 I contribute without expecting anything in return 3.70 0.99 −0.65 3.6 6.1 28.2 40.4 21.7 100.0 5472 456 I mean a lot to others 3.73 1.03 −0.71 3.9 7.4 24.5 40.3 23.9 100.0 5481 447 Others mean a lot to me 4.12 0.97 −1.18 2.7 3.5 15.0 36.8 41.9 100.0 5488 440 I trust most people 3.58 1.10 −0.59 5.4 11.0 24.7 38.1 20.8 100.0 5489 439 We rely upon each other 3.75 1.01 −0.71 3.7 5.8 26.2 40.0 24.3 100.0 5458 470 a Numbers of observations weighted. TABLE 3 Warwick scale indicators—descriptives and percentage distributions weighted. Never Rarely Every now and then Frequently All the time Total Missing observations Mean SD Skewness % % % % % % n a n a I have been feeling optimistic about the future 3.60 1.02 −0.55 3.8 9.8 27.9 39.3 19.1 100.0 5751 177 I have been feeling useful 3.53 0.98 −0.47 3.3 10.3 31.3 40.1 15.0 100.0 5728 200 I have been feeling relaxed 3.44 1.03 −0.39 4.1 13.6 30.8 37.0 14.5 100.0 5738 190 I have been dealing with problems well 3.56 0.98 −0.47 3.1 9.8 30.9 40.1 16.1 100.0 5724 204 I have been thinking clearly 3.68 0.97 −0.57 2.7 8.6 26.9 42.2 19.6 100.0 5707 221 I have been feeling close to other people 3.76 1.01 −0.65 2.8 8.3 23.8 40.2 24.9 100.0 5721 207 I have been able to make up my mind about things 3.84 1.00 −0.77 2.9 6.7 21.9 40.5 27.9 100.0 5712 216 a Numbers of observations weighted. JOURNAL OF ADOLESCENCE | 7 3.3 |Scale distributions by straightlining To examine distributions more closely, the mean scores for the two scales (Sense of Unity and Warwick‐Edinburgh Mental Wellbeing Scale) were categorized. All values were recoded to the closest of the integers 1 to 5. The categorized variables by straightlining (No/Yes) are shown in Figures 1and 2. On both scales, the most extreme scores (1 and 5) are more frequent among straightliners. In both distributions, there are lower percentages around value 4 among straightliners. On the Sense of Unity score, the means were 3.73 and 3.79 (Wald F= 2.381; df: 1 and 308; p= .124), SDs 1.09 and 0.68, and skewness −0.75 and −0.81 for those with and without straightlining. On the SWEMWBS, the corresponding means were 3.81 and 3.59 (Wald F= 29,625; df: 1 and 315; p< .001), SDs 1.06 and 0.69, and skewness −0.94 and −0.49 for those with and without straightlining. TABLE 4 Straightlining (identical responses across all eight items) on the Sense of Unity scale by gender and country, weighted and adjusted for cluster effects (school classes). Only students with valid values on all items are included. Straightlining Yes No Total Significance Country Gender % % % NN weighted Gender Country Finland Boys 46.4 53.6 100.0 503 642 χ 2 PEARSON = 50.849 F ADJUSTED = 39.386 df: 1; 73. p< .001 Boys: χ 2 PEARSON = 153.354 F ADJUSTED = 37.097 df: 2.798; 836.604 p< .001 Girls 25.0 75.0 100.0 518 675 Iceland Boys 27.3 72.7 100.0 1008 666 χ 2 PEARSON = 32.310 F ADJUSTED = 25.412 df: 1; 100. p< .001 Girls a 16.9 83.1 100.0 1068 705 Norway Boys 29.6 70.4 100.0 334 565 χ 2 PEARSON = 15.733 F ADJUSTED = 15.694 df: 1; 64. p< .001 Girls: χ 2 PEARSON = 92.492 F ADJUSTED = 18.945 df: 2.907; 892.551 p< .001 Girls a 16.9 83.1 100.0 355 601 Sweden Boys 15.0 85.0 100.0 682 611 χ 2 PEARSON = 32.124 F ADJUSTED = 19.412 df: 1; 77. p< .001 Girls 5.8 94.2 100.0 737 660 Total 22.8 77.2 100.0 5205 5124 b a By coincidence, the percentage distributions for Icelandic girls and Norwegian girls are identical. b The sum total of observations in the table is 5124. Due to rounding errors in the weighted numbers, the sum of number of observations across subgroups is 5125. TABLE 5 Straightlining (identical responses across all seven items) on the Warwick scale indicators by gender and country. Weighted and adjusted for cluster effects (school classes). Only students with valid values on all items are included. Country Gender Straightlining Yes No Total Significance %%%NN weighted Gender Countries Finland Boys 46.0 54.0 100.0 521 663 χ 2 PEARSON = 74.339 F ADJUSTED = 74.327 df: 1; 73. p< .001 Boys: χ 2 PEARSON = 227.433 F ADJUSTED = 50.179 df: 2.787; 833.188 p< .001 Girls 20.9 79.1 100.0 529 693 Iceland Boys 23.2 76.8 100.0 1045 690 χ 2 PEARSON = 40.492 F ADJUSTED = 48.343 df: 1; 100. p< .001 Girls 12.6 87.4 100.0 1099 726 Norway Boys 11.0 89.0 100.0 390 660 χ 2 PEARSON = 9.238 F ADJUSTED = 8.101 df: 1; 64. p= .006 Girls: χ 2 PEARSON = 93.131 F ADJUSTED = 22.962 df: 2.822; 866.460 p< .001 Girls 5.2 94.8 100.0 421 712 Sweden Boys 21.7 78.3 100.0 715 640 χ 2 PEARSON = 52.019 F ADJUSTED = 42.848 df: 1; 77. p< .001 Girls 8.4 91.6 100.0 782 700 Total 18.4 81.6 100.0 5502 5485* *The sum total of observations in the table is 5485. Due to rounding errors in the weighted numbers, the sum of number of observations across subgroups is 5484. 8 | AARØ ET AL. Jovanović, V. (2016). The validity of the satisfaction with life scale in adolescents and a comparison with single‐item life satisfaction measures: A preliminary study. Quality of Life Research,25, 3173–3180. https://doi.org/10.1007/s11136-016-1331-5 Kim, Y., Dykema, J., Stephenson, J., Black, P., & Moberg, D. P. (2018). Straightlining: overview of measurement, comparison of indicators, and effects in mail–web mixed‐mode surveys. Social Science Computer Review,37(2), 214–233. https://doi.org/10.1177/0894439317752406 Krosnick, J. A. (1991). Response strategies for coping with the cognitive demands of attitude measures in surveys. Applied Cognitive Psychology,5, 213–236. Levin, K. A., & Currie, C. (2014). Reliability and validity of an adapted version of the Cantril Ladder for use with adolescent samples. Social Indicators Research,119, 1047–1063. Loosveldt, G., & Beullens, K. (2017). Interviewer effects on non‐differentiation and straightlining in the European Social Survey. Journal of Official Statistics, 33(2), 409–426. Lyyra, N., Thorsteinsson, E. B., Eriksson, C., Löfstedt, P., Madsen, K., Tolvanen, A., & Välimaa, R. (2021). The association between loneliness, mental well‐ being, and self‐esteem among adolescents in four Nordic countries. International Journal of Environmental Research and Public Health,18(14), 7405. https://doi.org/10.3390/ijerph18147405 Meilstrup, C., Holstein, B. E., Nielsen, L., Due, P., & Koushede, V. (2020). Self‐efficacy and social competence reduce socioeconomic inequality in emotional symptoms: A social epidemiological study of schoolchildren. European Journal of Public Health,30(1), 80–85. https://doi.org/10.1093/eurpub/ckz058 Moser, C. A., & Kalton, G. (1971). Survey methods in social investigation. Heinemann. Nielsen, L., Stewart‐Brown, S., Vinther‐Larsen, M., Meilstrup, C., Holstein, B. E., & Koushede, V. (2016). High and low levels of positive mental health: are there socioeconomic differences among adolescents? Journal of Public Mental Health,15,37–49. Omrani, A., Wakefield‐Scurr, J., Smith, J., & Brown, N. (2019). Survey development for adolescents aged 11‐16 years: A developmental science based guide. Adolescent Research Review,4, 329–340. Reuning, K., & Plutzer, E. (2020). Valid vs. invalid straightlining: The complex relationship between straightlining and data quality. Survey Research Methods,14(5), 439–459. Ringdal, R., Eilertsen, M.‐E. B., Bjørnsen, H. N., Espenes, G. A., & Moksnes, U. K. (2017). Validation of two versions of the Warwick‐Edinburgh Mental Well‐Being Scale among Norwegian adolescents. Scandinavian Journal of Public Health,46(7), 718–725. Samdal, O., Moreno, C., Morgan, A., Matos, M. G., Banan, A., & Wold, B. (2016). Stimulating Adolescent Life skills through Unity and Drive (SALUD). Poster at HBSC meeting, Luxembourg, Nov. 3, 2016. Schleicher, A. (2018). PISA 2018. Insights and interpretations. OECD. https://www.oecd.org/pisa/PISA%202018%20Insights%20and%20Interpretations% 20FINAL%20PDF.pdf Schwarzer, R. (1992). Self‐efficacy in the adoption and maintenance of health behaviors: Theoretical approaches and a new model. In R. Schwarzer (Ed.), Self‐efficacy: Thought control of action (pp. 217–243). Hemisphere. Steinberg, L., Graham, S., O'Brian, L., Woolard, J., Cauffman, E., & Banich, M. (2009). Age differences in future orientation and discounting. Child Development,80(1), 28–44. Stewart‐Brown, S., Tennant, A., Tennant, R., Platt, S., Parkinson, J., & Weich, S. (2009). Internal construct validity of the Warwick‐Edinburgh Mental Well‐ being Scale (WEMWBS): A Rasch analysis using data from the Scottish Health Education Population Survey. Health and Quality of Life Outcomes,7, 15. https://doi.org/10.1186/1477-7525-7-15 Tennant, R., Hiller, L., Fishwick, R., Platt, S., Joseph, S., & Scott Weich, J. (2007). The Warwick‐Edinburgh Mental Well‐being Scale (WEMWBS): Development and UK validation. Health and Quality of Life Outcomes, 20075, 63. https://doi.org/10.1186/1477-7525-5-63 Torsheim, T., Samdal, O., Rasmussen, M., Freeman, J., Griebler, R., & Dur, W. (2012). Cross‐national measurement invariance of the Teacher and Classmate Support Scale. Social Indicator Research,105(1), 145–160. Torsheim, T., Wold, B., & Samdal, O. (2000). The Teacher and Classmate Support scale—factor structure, test‐retest reliability and validity in samples of 13‐and 15‐year‐old adolescents. School Psychology International,21(2), 195–212. Van Vaerenbergh, Y., & Thomas, T. D. (2013). Response styles in survey research: A literature review of antecedents, consequences, and remedies. International Journal of Public Opinion Research,25(2), 195–217. Villar, A. (2008). Response bias. In P. J. Lavrakas (Ed.), Encyclopedia of survey research methods (Vol. 2, pp. 751–753). Sage. Wahlström, J., Modin, B., Svensson, J., Löfstedt, P., & Brolin Låftman, S. (2021). Sense of unity and self‐reported health among 15‐year‐olds: Findings from the Swedish 2017/18 Health Behavior in school‐aged Children Study, International Journal of Public Health,66(36), 621964. https://doi.org/10.3389/ ijph.2021.621964 Yan, T. (2008). Nondifferentiation. In P. J. Lavrakas (Ed.), Encyclopedia of Survey Research Methods (Vol. 2, pp. 520–521). Sage. Zhang, C., & Conrad, F. (2014). Speeding in web surveys: the tendency to answer very fast and its association with straightlining. Survey Research Methods, 8(2), 127–135. https://doi.org/10.18148/srm/2014.v8i2.5453 Zimet, G., & Grodaon, K. (1988). The multidimensional scale of perceived social support. Journal of Personality Assessment,52(1), 30–41. Zimet, G. D., Dahlem, N. W., Zimet, S. G., & Farley, G. K. (1988). The multidimensional scale of perceived social support. Journal of Personality Assessment, 52(1), 30–41. SUPPORTING INFORMATION Additional supporting information can be found online in the Supporting Information section at the end of this article. How to cite this article: Aarø, L. E., Fismen, A.‐S., Wold, B., Skogen, J. C., Torsheim, T., Arnarsson, Á. M., Lyyra, N., Löfstedt, P., & Eriksson, C. (2022). Nordic adolescents responding to demanding survey scales in boring contexts: Examining straightlining. Journal of Adolescence,1–15. https://doi.org/10.1002/jad.12066 JOURNAL OF ADOLESCENCE | 15