Skip to main content
Download PDF

Measuring social desirability bias in a multi-ethnic cohort sample: its relationship with self-reported physical activity, dietary habits, and factor structure

BMC Public Health volume 23, Article number: 415 (2023) Cite this article

Abstract

Background

Social desirability bias is one of the oldest forms of response bias studied in social sciences. While individuals may feel the need to fake good or bad answers in response to sensitive or intrusive questions, it remains unclear how rampant such a bias is in epidemiological research pertaining to self-reported lifestyle indicators in a multicultural Asian context. The main purpose of the current study is, therefore, to examine the sociodemographic correlates and impact of social desirability responding on self-reported physical activity and dietary habits at an epidemiological scale in a non-western multi-cultural Asian setting.

Methods

Prior to the main analyses, confirmatory and exploratory factor analyses were conducted to determine the factorial validity of a western derived concept of social desirability. Multiple regression analyses were conducted on cross-sectional data (n = 2995) extracted from a nationwide survey conducted between 2019 and 2020.

Results

A unique factor structure of social desirability was found and was therefore used for subsequent analyses. Multiple regression analyses revealed older age groups, the Indian ethnic group, those with past or present marriages, and having no income, had a significantly greater tendency to act on the bias.

Conclusion

The construct of social desirability bias was fundamentally different in a multicultural context than previously understood. Only a small proportion of variance of self-report lifestyle scores was explained by social desirability, thus providing support for data integrity.

Peer Review reports

Introduction

One of the oldest and most common forms of response bias studied in social science, the social desirability bias, is defined as an individual’s propensity to respond in a way that is viewed favourably by society. Alike many forms of response biases, social desirability bias causes misreports, either by faking good or bad responses, making the scientific inquiry of any social phenomenon challenging [1].

Self-report surveys, which are commonly used in attitudinal research, are particularly vulnerable to such bias. Surveys that ask of sensitive or intrusive topics increases an individual’s tendency to misreport for fear of social disapproval or embarrassment. The tendency to misreport depends greatly on the sensitivity of the context, and the magnitude of its confounding effect therefore varies. While there are alternatives to self-report measures that can improve the precision of data, such as dietary or physical activity surveillance used in health research, these alternatives incur larger administrative costs, participant burden, and intrusion [2, 3]. Self-report thus remains a popular method due to its ease of administration, minimal costs, and greater nonmaleficence.

What is considered to be socially acceptable depends largely on culture. Since such a bias is greatly attributed by sociocultural norms, past research that are typically based on homogeneous and individualistic cultures may not be generalizable to a multi-cultural society in Asia. Collectivistic societies that strive for cooperativeness and sensitivity may exercise more of such a bias than individualistic societies [4]. However, empirical data on the effects of social desirability bias is lacking in collectivistic societies. Similarly, it is also unclear if western-derived constructs of social desirability can be generalized to a non-western population that is also multi-cultural [5].

Social desirability bias has been largely neglected in aspects of attitudinal research where there is a greater cause for concern of misreporting. Misreporting on lifestyle issues, such as smoking, physical exercise, or dietary habits, may in turn confound precision greatly needed to inform health policy [6,7,8]. Research has shown that social desirability bias is readily exercised by children and adults alike on self-reporting dietary intake for instance [9, 10]. Individuals with a less than ideal height or weight tend to self-report within the range deemed to be socially acceptable [11]. Similarly, there is also a tendency to over-report physical activity levels to a significant degree [12].

In Singapore, epidemiological research on physical and mental health are becoming increasingly apparent. To name a few: the national population health survey (NPSH) conducted by the Ministry of Health that began in 1998 [13], the national youth survey (NYS) commenced in 2002 [14], the Well-being of the Singapore Elderly (WiSE) population study which involved older persons above the age of 60 started its first iteration in 2010 [15], the Singapore Mental Health Study (SMHS) which focused on nation-wide mental health surveillance, was initiated in 2009 [16], and finally, the Knowledge, Attitudes and Practices (KAP) survey on diabetes commenced in 2019 [17]; all of which are ongoing epidemiological research with a central aim of providing up-to-date representative data on the health and well-being of Singapore’s population. It is crucial to note that none of the existing local health research have studied the impact of social desirability bias. At this juncture, not enough is understood of the effects of such a bias on self-reported physical health and lifestyle concerns—indicators that are typically captured in existing local epidemiological research. If such effects exist, how much of self-report data is attributed to such a bias and which groups have a greater tendency to exercise such a bias? Understanding the direction and magnitude of the bias is thus necessary so as to address it.

Singapore is a multi-ethnic and multi-cultural city-state in Southeast Asia with a population of approximately 4 million residents (citizens and permanent residents; based on 2021 data from the Singapore Department of Statistics). To the best of knowledge, this is the first study to report the sociodemographic correlates of social desirability bias and its impact on self-reported lifestyle behaviours—dietary habits and physical activity specifically. Results of the study will be beneficial for understanding the current level of bias, if any, in a multi-ethnic Asian population.

Methods

Data were drawn from a large cross-sectional study called the Knowledge, Attitudes, and Practices (KAP) study, which is a nationwide survey conducted between 2019 and 2020 to understand the KAP of diabetes in relation to lifestyle factors in Singapore. The full details of the study’s protocol and design have been published elsewhere [17]. Briefly, Singapore residents were randomly selected from a national administrative registry and were subsequently approached by trained interviewers at their households to complete a set of questionnaires. An additional hundred diabetes cases were conveniently sampled. All participants completed written informed consent and were compensated an amount for their participation. An overall response rate of 66.2% was achieved, which was expected result [17, 18]. Chi-square analyses between responders and non-responders were conducted to ascertain response bias. Based on the original sample size of 2895, younger age groups (χ2(3) = 20.70, p < .001), those of Malay and Indian ethnicities (χ2(3) = 57.5110, p < .001), were more likely to respond to the survey. The study was approved by the relevant local ethics committees: Institutional Research Review Committee (IRRC) and the National Healthcare Group Domain Specific Review Board (DSRB; Ref no. 2018/00430).

Participants

The final study sample included individuals from all ethnic groups (Chinese, Malay, Indian, or other), who were 18 years of age and above, and were literate in any of Singapore’s national languages: English, Chinese, Malay, or Tamil. The study excluded individuals who were not living in Singapore, hospitalised or institutionalised during the study period. Data from a total of 2995 individuals (including 100 cases with diabetes) were extracted for the purpose of this investigation.

Measures

Marlowe–Crowne Social Desirability Scale (MCSDS) by Crowne and Marlowe (1960) measures the level of social desirability bias [19]. Participants were asked to make True/False responses to 33 items. Individual item responses range from 1 to 2. However, item 1, “Before voting I thoroughly investigate the qualifications of all the candidates,” was dropped from the questionnaire after participants expressed concerns that it assessed their political preferences and may be used for other purposes. ‘False’ responses were denoted as a 2, and ‘True’ responses were given the value 1. The overall internal consistency of 32 items was found to be good (α = 0.75). The short-form versions (type A, B, C, XX, X1, X2) were extracted from the original 33-item. Items 1, 2, 4, 7, 8, 13, 16, 17, 18, 20, 21, 24, 25, 26, 27, 29, 31, and 33 were reversed scored. Higher scores denote a greater bias.

Global Physical Activity Questionnaire (GPAQ), developed by the World Health Organisation (WHO, 2006), measures self-reported levels of physical activity (moderate or vigorous intensity levels) on average at work, travelling to and from places, during recreational activities on average daily. It also records self-reported minutes spent on sedentary activities per day, such as sitting or reclining while awake [20]. Metabolic Equivalents (METs) are calculated by the ratio of working to resting metabolic rate in minutes/week for each domain. An overall global MET score (GPAQ grand MET score) was subsequently calculated by the sum of METs in each domain except the sedentary domain, which was calculated separately (GPAQ sedentary score).

A locally validated diet screener was utilised to record the level of consumption of typical food and beverages in the past one year [21]. Respondents self-reported the frequency of food/beverage intake on a 10-point scale, ranging from “never/rarely” to “6 or more times per day”. A Dietary Approaches to Stop Hypertension (DASH) score—a sensitive measure to test against health-related outcomes [21]—was subsequently calculated by the amount of intake of vegetables, nuts or legumes, fruit, low fat dairy, processed or red meat (reversed scored), whole grains, and sweetened beverages (reversed scored) as a whole (based on standard servicing sizes). Participants eventually received a score of 1 to 5 on each of the seven categories depending on the quintile level of intake [22, 23], which were eventually totalled as an overall DASH score. Higher DASH scores thus represent a healthier diet.

A sociodemographic questionnaire comprising of age, gender, ethnicity, personal monthly income (in Singapore dollars (SGD), highest education level, height and weight information were also completed by the participants.

Statistical analysis

As the MCSDS scale was originally conceptualized in the west, it is unclear if the original factorial structure is valid in a multicultural Asian setting, as factors of social desirability depend heavily on cultural norms that determine what constitutes a desirable response. Past research showed that the MCSDS scale, while internally valid and reliable, do not have a consistent factorial structure across countries. While the original one-factor solution has been supported in the Icelandic population [24], various two-factor solutions have gained greater support, such as in Africa, Switzerland, Canada, and in Malaysia [25,26,27]. Additionally, abbreviated versions of the scale (e.g., type A, B, C, XX, X1, X2) that contain 10 to 20 of the original 33 items [28, 29], and that were created out of a need to address the lack of factorial fit and lengthiness of the scale, have also been widely validated [30, 31]. Generally, the assumed single-dimensionality of alternative forms too have received mixed support [24, 32]. Thus, prior to the main analyses, confirmatory and exploratory factor analyses were conducted to ascertain the factorial fit of MCSDS in the local sample.

Confirmatory factor analysis (CFA)

Missing data accounted for 2.33% (7/2995) of all data and were deleted listwise. All analyses were conducted based on the final sample size of n = 2988. To address the dichotomous True/False format of the MCSDS, CFA was performed using unweighted least squares estimator (ULSMV), set at a default mode of 1000 iterations and delta parameterization. A fixed factor method for factor scaling was used by fixing the model factor variance to 1 and constraints were freed on the first item (on item 2, because item 1 was dropped) of each factor [33]. Model fit was assessed using absolute and relative fit indices, i.e., model chi-square χ2, root mean square error of approximation (RMSEA), comparative fit index (CFI), and Tucker–Lewis index (TLI). This study utilised Marsh et al.’s recommendations for good fit as the model evaluation criteria [34]: RMSEA < 0.08; CFI > 0.90; TLI > 0.90.

Exploratory factor analysis (EFA)

EFA was subsequently performed on one split-half of the sample (n = 1494) to extract the factor structure of MCSDS that gave the best fit of data. Similarly, EFA was performed using unweighted least square mean and variance (ULSMV), with a default setting of oblique geomin rotation. The final model derived from EFA was tested using CFA on the second split-half sample, and eventually, on the overall sample. Communality coefficients (h2) were calculated by the formula: 1 - estimated residual variances. CFA and EFA were performed on Mplus 8.3. by Muthén and Muthén [35].

Multiple linear regression analyses

To address the main aim of the study, multiple linear regression analysis was performed on 2895 data points (with 100 cases with diabetes removed) between sociodemographic categorical variables, such as age, gender, ethnicity, personal income, and education levels, and the total continuous scores of the 21-item adapted version of the MCSDS. Subsequent hierarchical linear regression analyses using the enter method were performed to ascertain the unique contribution of MCSDS in predicting dietary habits, physical activity, and sedentary behaviours. Sociodemographic and BMI covariates were entered in the first step to control for confounds, followed by the modified MCSDS which was added in the second step. Weights were applied to the survey data for all multiple linear regression analysis for representativeness. All missing values were deleted listwise. The analyses were conducted on IBM SPSS statistics software version 23.0.

Results

Goodness of fit of original, popular short forms, and the modified version

After dropping the first item of the scale due to participants’ concerns, CFA was performed on the remaining 32 items, and type A, B, C, XX, X1, X2 short-form versions, to determine the factorial appropriateness of existing popular forms. Model parameters of all full and abbreviated forms of the MCSDS derived from the CFA were unsatisfactory (Table 1). This led the team to conduct EFA on the full 32-item scale (with item 1 dropped) using a split-half sample. The number of factors extracted was determined by an inspection of the scree plot [36], which suggested an optimal extraction of three factors (Fig. 1). Items that had weak rotated factor coefficients (< 0.04) were removed (see Table 2). This was proceeded by additional EFA iterations to eliminate problematic items (i.e., low factor loadings and cross-loadings). Afterwards, it was realised that the third factor could not be identified due to insufficient component saturation and common meaning between its two items, leading author WLT to reconsider a two-factor solution. The three-factor solution was thus rejected and multiple EFAs were subsequently performed to eliminate problematic items in the two-factor solution. The degree of stability of the final two-factor model was confirmed by performing CFA on the second split-half of the sample [37]. Factor coefficients with greater than 0.40 item loadings were retained, resulting in the final retention of 21 items (Table 2). The final 21-item two-factor solution was assessed with CFA on the second split-half and eventually on the full sample (Table 1). Parameter estimates of the final 21-item two-factor solution showed vast improvements to the model fit and represented the best fit of data.

Fig. 1
figure 1

Extracted eigen values from Exploratory Factor Analysis of the Marlowe-Crowne Social Desirability Scale (MCSDS; the first item of the 33 items was removed due to participant feedback)

Full size image
Table 1 Exploratory (EFA) and Confirmatory Factor Analysis (CFA) derived Absolute and Comparative fit measures of original and short form versions of the Marlowe-Crowne Social Desirability Scale (MCSDS)
Full size table
Table 2 Final 21-item modified Marlowe-Crowne Social Desirability Scale (MCSDS) Factor Structure Matrix Rotated to the Geomin Criterion
Full size table

The modified 21-item version (modified MCSDS)

The final adapted MCSDS comprised two factors with 21 items (Table 2). The first factor—Civil and Lawful Impression—centralises on giving an impression of being seen as non-hostile (to an authority) or law-abiding. The second factor—Integrous Image boosting—centralises on presenting oneself with an honest, upright, or refined image. The 21-item MCSDS is hereafter referred to as the modified MCSDS. The internal consistency of the modified MCSDS was found to be acceptable, α = 0.74. The correlation between the full MCSDS and the modified version was r = .938, with r2 = 0.890.

Sociodemographic correlates of the modified MCSDS

The overall model was significant, F(19, 2729) = 14.36, p < .001, R2 = 0.168. Multiple linear regression analyses revealed that older age (35 years and above) as compared to younger age (below 35 years), Indian ethnicity as compared to the Chinese ethnicity, marriage history (current and past) as compared to single/never married, and having no income as compared to below 2k and 10k and above income brackets, were significantly associated with higher social desirability bias. Gender, education levels, Malay, and other ethnicities, did not emerge as significant sociodemographic factors of the bias. The results of the multiple weighted linear aggression analyses are summarized in Table 3.

Table 3 Sociodemographic determinants of the modified Marlowe-Crowne Social Desirability Scale (MCSDS).
Full size table

The association between modified MCSDS, dietary habits, and physical activity

Results from hierarchical linear regression analyses, summarized in Table 4, revealed that higher social desirability bias was significantly associated with healthier self-reported dietary habits (DASH scores), b = 0.16, t = 4.69, p < .001, 95% CI [0.09,0.23], greater physical activity (GPAQ global MET scores), b = 73.27, t = 1.98, p=. 0.048, 95% CI [0.81, 145.73], and shorter sedentary time (GPAQ sedentary score), b= -5.48, t= -3.76, p < .001, 95% CI [-8.34, -2.62], after accounting for sociodemographic and BMI covariates. Additionally, the modified MCSDS explained significant but small proportions of variance in dietary scores (F(1, 2556) = 22.00, R2 = 0.1502, step 2 ΔR2 = 0.015, p < .001), physical activity scores (F(1, 2556) = 3.93, R2 = 0.0882, step 2 ΔR2 = 0.0024, p = .0475), and sedentary scores, (F(1, 2556) = 14.13, R2 = 0.1413, step 2 ΔR2 = 0.0089, p = .0002).

Table 4 Summary of hierarchical regression analysis for variables predicting self-reported diet scores (DASH scores), self-reported physical activity scores (GPAQ grand MET score), self-reported sedentary level scores (GPAQ sedentary score)
Full size table

Discussion

The main purpose of the study is to ascertain the association between social desirability, sociodemographic, and self-reported lifestyle factors. Prior to the main investigation, the factor structures of the original 33-item MCSDS scale [19] and its popular abbreviated versions [25, 28] were evaluated in the context of a large, representative, multi-ethnic non-western population. A total of nine CFAs were conducted, and the results showed that the full and short-form versions did not achieve adequate fit of data. EFA was subsequently conducted, and a 21-item two-factor solution was ultimately derived. Additional iterations of CFA on the full and split half samples revealed an excellent fit of data for the modified 21-item version. As the dimensions of social desirability are uniquely different in Singapore as compared to elsewhere, it is therefore necessary to discuss its factorial properties in this section.

Overall, the results supported a two-factor structure [26, 30,31,32]. Like past studies, various items on the full scale had weak rotated factor coefficients (< 0.04), implying that not all items of the original scale had adequately captured the factors [24]. Unique to our investigation, item 1 of the scale, “Before voting in an election, I thoroughly investigate the qualifications of all the candidates”, was removed due to significant concerns raised by participants during the survey, but not due to weak factor loadings. While future research that use this scale may include the first item for scale completeness, users need to be aware that participants may feel uncomfortable answering this question face-to-face and may have to contend with missing values.

The present results differed from Millham’s (1974) components of “attribution” (i.e., tendency to endorse characteristics that are socially desirable; items 1,4,13,17,25,26,27,29) and “denial” (i.e., tendency to deny characteristics that are socially undesirable; items 3,5,9,10,11,23,28)[38, 39], or the two-factor solution of “achievement” and “interpersonal relationship” derived by Verardi and colleagues [27]. Instead, the two factors—Civil and Lawful Impression (CLI) and Integrous Image Boosting (IIB)—appeared to strike a chord with Paulhus’s two components of the Balanced Inventory of Desirable Responding (BDIR; [25]—the concepts of Impression Management (IM), defined as the “conscious dissimulation of test responses designed to create a favourable impression”, and Self-deceptive Enhancement (SDE), defined as “any positively biased response that the respondent actually does believe to be true”. Items that load on the CLI factor corresponded to a mix of IM and SDE items, whereas the items that load on the IIB factor comprised a majority of SDE items with a small overlap of IM items. The highly overlapping constructs of IIB and SDE factors could imply that they were measuring a highly similar construct — an esteem boosting of self-image, whereas the lower degree of overlap between the CLI and the IM factor (or a mixed overlap with both IM and SDE factors) could signify a culturally nuanced form of social self-control that is unique to the present population; one that represents a form of social self-control against the system or an authority. Future research could replicate and extend our findings further by investigating the stability and generalizability of the current factor structure in more stigmatizing fields, such as mental illness perceptions in Singapore [40, 41].

It was discovered that social desirability explained a significant but small proportion of variances of lifestyle factors (physical activity, sedentary time, diet scores), which corroborated past reports that were based in western communities or unrelated to health-based research [12, 42]. The small effect of the modified MCSDS on lifestyle constructs could be due to the study’s emphasis on anonymity and protection of personal data during consent taking, despite it being conducted face-to-face, which could have brought about confidentiality confidence in participants. Older age groups, those of Indian ethnicity, those who were ever married and those without income (as compared to those earning below 2k or at least 10k and above), had a significantly greater tendency to respond in a way that is socially desired. This could be due to unconscious social motivations underlying the need to boost self-image, such as the need for social approval, to avoid negative evaluation, or to be stereotyped against [19], as the surveys were conducted face-to-face. It is thus important for future epidemiology research to take into account these minority and/or vulnerable groups. Improving the level of anonymity, such as by the use of online self-administered questionnaires [43], can further mitigate this bias across groups of individuals who have a greater inclination to give socially desirable answers thus leading to under-estimation of the true extent of the problem in these communities. Gender was not a significant predictor of social desirability scores in the present sample which corroborated the majority of past reports, regardless of MCSDS item makeup [19, 27, 30, 32, 44].

Though a small percentage of variance of dietary intake, physical activity, and sedentary behaviour, were attributed to social desirability bias, not at all accounting for the bias is discouraged. On the contrary, such a bias should be measured and accounted for, ideally, in all health-related research that uses the self-report methodology. Though face-to-face interviewing may contribute to social desirability bias to a greater extent than online-based surveys, this mode of data collection is essential to engage particularly hard-to-reach demographics, such as older respondents who may not be as internet-savvy; therefore, having a mix of online and offline modes of data collection should be encouraged in epidemiological research that requires population representation [45, 46].

Strengths and limitations

Several caveats have to be taken into consideration before conclusions can be made. This study uses cross-sectional data, therefore, causal or temporal inferences cannot be made. In addition, we did not use objective measures to determine the level of deviation from self -reported physical activity, dietary intake, or sedentary behaviours, and thus, we could not establish the levels of discrepancies in self-reports attributed by the bias. While the response rate of approximately 66% is acceptable, [47], the results may not be representative of non-responders. As this study uses face-to-face surveys, social desirability bias may be larger than other surveying modes that provide better levels of anonymity. However, due to an ageing society, face-to-face interviews were necessary to engage hard-to-reach groups such as older adults. This was the first study that investigated social desirability bias in a multi-cultural Asian context. Results are based on a large cohort sample that is representative of Singapore’s resident population.

Conclusion

Due to Singapore’s unique cultural makeup, pre-existing popular abbreviated scales had failed to provide an acceptable factorial fit of local data. Instead, a unique 21-item modified scale had produced the best fit. Additionally, certain groups of participants were identified to be more likely to exercise this bias, which could be due to unconscious social motivations that interact with face-to-face research. Nonetheless, only a small proportion of variance of scores on self-report lifestyle scores could be explained by social desirability bias, which is reassuring as it suggests that participants were not greatly influenced by social expectations to respond in a socially desirable way despite a face-to-face survey, which has a lower level of anonymity than online surveys. In conclusion, measuring social desirability bias in health-based research reliant on self-report methodology is highly encouraged so as to account for its effects.

Data Availability

The data that support the findings of this study are available but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission; data requests can be submitted to mythily@imh.com.sg.

Abbreviations

KAP:

Knowledge, Attitudes, and Practices

IRRC:

Institutional Research Review Committee

DSRB:

Domain Specific Review Board

MCSDS:

Marlowe–Crowne Social Desirability Scale

GPAQ:

Global Physical Activity Questionnaire

WHO:

World Health Organisation

METs:

Metabolic Equivalents

GPAQ grand MET score:

Global MET score

GPAQ sedentary score:

Sedentary MET score

DASH:

Dietary Approaches to Stop Hypertension

RMSEA:

Root mean square error of approximation

CFI:

Comparative fit index

TLI:

Tucker–Lewis index

EFA:

Exploratory factor analysis

CFA:

Confirmatory factor analysis

ULSMV:

Unweighted least square mean and variance

References

  1. Nederhof AJ. Methods of coping with social desirability bias: a review. Eur J Soc Psychol. 1985;15(3):263–80.

    Article  Google Scholar 

  2. Foster E, Bradley J. Methodological considerations and future insights for 24-hour dietary recall assessment in children. Nutr Res. 2018;51:1–11.

    Article  CAS  PubMed  Google Scholar 

  3. Henriksen A, Haugen Mikalsen M, Woldaregay AZ, Muzny M, Hartvigsen G, Hopstock LA, et al. Using fitness trackers and smartwatches to measure physical activity in research: analysis of consumer wrist-worn wearables. J Med Internet Res. 2018;20(3):e110.

    Article  PubMed  PubMed Central  Google Scholar 

  4. Ryan AM, Bradburn J, Bhatia S, Beals E, Boyce AS, Martin N, et al. In the eye of the beholder: considering culture in assessing the social desirability of personality. J Appl Psychol. 2021;106(3):452.

    Article  PubMed  Google Scholar 

  5. Nurumov K, Hernández-Torrano D, Ait Si Mhamed A, Ospanova U. Measuring social desirability in collectivist countries: a psychometric study in a representative sample from Kazakhstan. Front Psychol. 2022;1449.

  6. King BM, Cespedes VM, Burden GK, Brady SK, Clement LR, Abbott EM, et al. Extreme under-reporting of body weight by young adults with obesity: relation to social desirability. Obes Sci Pract. 2018;4(2):129–33.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Latkin CA, Edwards C, Davey-Rothwell MA, Tobin KE. The relationship between social desirability bias and self-reports of health, substance use, and social network factors among urban substance users in Baltimore, Maryland. Addict Behav. 2017;73:133–6.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Kirkpatrick SI, Baranowski T, Subar AF, Tooze JA, Frongillo EA. Best practices for conducting and interpreting studies to validate self-report dietary assessment methods. J Acad Nutr Diet. 2019;119(11):1801–16.

    Article  PubMed  Google Scholar 

  9. Hebert JR, Clemow L, Pbert L, Ockene IS, Ockene JK. Social desirability bias in dietary self-report may compromise the validity of dietary intake measures. Int J Epidemiol. 1995;24(2):389–98.

    Article  CAS  PubMed  Google Scholar 

  10. Klesges LM, Baranowski T, Beech B, Cullen K, Murray DM, Rochon J, et al. Social desirability bias in self-reported dietary, physical activity and weight concerns measures in 8-to 10-year-old african-american girls: results from the Girls Health Enrichment Multisite Studies (GEMS). Prev Med (Baltim). 2004;38:78–87.

    Article  Google Scholar 

  11. Burke MA, Carman KG. You can be too thin (but not too tall): Social desirability bias in self-reports of weight and height. Econ Hum Biol. 2017;27:198–222.

    Article  PubMed  Google Scholar 

  12. Adams SA, Matthews CE, Ebbeling CB, Moore CG, Cunningham JE, Fulton J, et al. The effect of social desirability and social approval on self-reports of physical activity. Am J Epidemiol. 2005;161(4):389–98.

    Article  PubMed  Google Scholar 

  13. Epidemiology & Disease Control Division and Policy R& SG, Ministry of Health and Health Promotion Board S. National Population Health Survey 2020 [Internet]. Singapore. ; 2020. Available from: https://www.moh.gov.sg/docs/librariesprovider5/default-document-library/nphs-2020-survey-report.pdf

  14. National Youth Council S. National Youth Survey [Internet]. [cited 2023 Jan 29]. Available from: https://www.nyc.gov.sg/en/initiatives/resources/national-youth-survey/

  15. Subramaniam M, Abdin E, Vaingankar JA, Sambasivam R, Seow E, Picco L, et al. Successful ageing in Singapore: prevalence and correlates from a national survey of older adults. Singap Med J. 2019;60(1):22.

    Article  Google Scholar 

  16. Subramaniam M, Vaingankar J, Heng D, Kwok KW, Lim YW, Yap M et al. The Singapore Mental Health Study: an overview of the methodology. Int J Methods Psychiatr Res [Internet]. 2012 Jun 1 [cited 2020 Apr 27];21(2):149–57. Available from: https://doi.org/10.1002/mpr.1351

  17. AshaRani PV, Abdin E, Kumarasan R, Kumar FDS, Shafie S, Jeyagurunathan A, et al. Study protocol for a nationwide knowledge, Attitudes and Practices (KAP) survey on diabetes in Singapore’s general population. BMJ Open. 2020;10(6):e037125.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  18. Wee HL, Ho HK, Li SC. Public awareness of diabetes mellitus in Singapore. Singap Med J. 2002;43(3):128–34.

    CAS  Google Scholar 

  19. Crowne DP, Marlowe D. A new scale of social desirability independent of psychopathology. J Consult Psychol. 1960;24(4):349.

    Article  CAS  PubMed  Google Scholar 

  20. Armstrong T, Bull F. Development of the world health organization global physical activity questionnaire (GPAQ). J Public Health (Bangkok). 2006;14(2):66–70.

    Article  Google Scholar 

  21. Whitton C, Ho JCY, Rebello SA, van Dam RM. Relative validity and reproducibility of dietary quality scores from a short diet screener in a multi-ethnic asian population. Public Health Nutr. 2018;21(15):2735–43.

    Article  PubMed  Google Scholar 

  22. Fung TT, Chiuve SE, McCullough ML, Rexrode KM, Logroscino G, Hu FB. Adherence to a DASH-style diet and risk of coronary heart disease and stroke in women. Arch Intern Med. 2008;168(7):713–20.

    Article  PubMed  Google Scholar 

  23. Lee YY, Lau JH, Seet V, Whitton C, Asharani PV, Kumar FDS, et al. Dietary intake of persons with depressive and psychotic disorders in Singapore. Ann Acad Med Singapore. 2021;50(5):379–89.

    Article  PubMed  Google Scholar 

  24. Vésteinsdóttir V, Reips U-D, Joinson A, Thorsdottir F. Psychometric properties of measurements obtained with the Marlowe–Crowne Social Desirability Scale in an icelandic probability based Internet sample. Comput Hum Behav. 2015;49:608–14.

    Article  Google Scholar 

  25. Paulhus DL. Two-component models of socially desirable responding. J Pers Soc Psychol. 1984;46(3):598.

    Article  Google Scholar 

  26. Ventimiglia M, MacDonald DA. An examination of the factorial dimensionality of the Marlowe Crowne Social Desirability Scale. Pers Individ Dif. 2012;52(4):487–91.

    Article  Google Scholar 

  27. Verardi S, Dahourou D, Ah-Kion J, Bhowon U, Tseung CN, Amoussou-Yeye D, et al. Psychometric properties of the Marlowe-Crowne social desirability scale in eight african countries and Switzerland. J Cross Cult Psychol. 2010;41(1):19–34.

    Article  Google Scholar 

  28. Reynolds WM. Development of reliable and valid short forms of the Marlowe-Crowne Social Desirability Scale. J Clin Psychol. 1982;38(1):119–25.

    Article  Google Scholar 

  29. Strahan R, Gerbasi KC. Short, homogeneous versions of the Marlowe-Crowne social desirability scale. J Clin Psychol. 1972

  30. Loo R, Thorpe K. Confirmatory factor analyses of the full and short versions of the Marlowe-Crowne Social Desirability Scale. J Soc Psychol. 2000;140(5):628–35.

    Article  CAS  PubMed  Google Scholar 

  31. Sârbescu P, Costea I, Rusu S. Psychometric properties of the Marlowe-Crowne social desirability scale in a romanian sample. Procedia-Social Behav Sci. 2012;33:707–11.

    Article  Google Scholar 

  32. Barger SD. The Marlowe-Crowne affair: short forms, psychometric structure, and social desirability. J Pers Assess. 2002;79(2):286–305.

    Article  PubMed  Google Scholar 

  33. Wang J, Wang X. Structural equation modeling: applications using Mplus. John Wiley & Sons; 2019.

  34. Marsh HW, Hau K-T, Wen Z. In search of golden rules: Comment on hypothesis-testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing Hu and Bentler’s (1999) findings. Struct Equ Model. 2004;11(3):320–41.

  35. Muthén B, Muthén L, Mplus. Chapman and Hall/CRC; 2017.

  36. Cattell RB. The scree test for the number of factors. Multivar Behav Res. 1966;1(2):245–76.

    Article  CAS  Google Scholar 

  37. Brown TA. Confirmatory factor analysis for applied research. Guilford publications; 2015.

  38. Ramanaiah NV, Martin HJ. On the two-dimensional nature of the Marlowe-Crowne Social Desirability Scale. J Pers Assess. 1980;44(5):507–14.

    Article  CAS  PubMed  Google Scholar 

  39. Kurz AS, Drescher CF, Chin EG, Johnson LR. Measuring social desirability across language and sex: a comparison of Marlowe–Crowne Social Desirability Scale factor structures in English and Mandarin Chinese in Malaysia. PsyCh J. 2016;5(2):92–100.

    Article  PubMed  Google Scholar 

  40. Picco L, Chang S, Abdin E, Chua BY, Yuan Q, Vaingankar JA, et al. Associative stigma among mental health professionals in Singapore: a cross-sectional study. BMJ Open. 2019;9(7):e028179.

    Article  PubMed  PubMed Central  Google Scholar 

  41. Chang S, Ong HL, Seow E, Chua BY, Abdin E, Samari E, et al. Stigma towards mental illness among medical and nursing students in Singapore: a cross-sectional study. BMJ Open. 2017;7(12):e018099.

    Article  PubMed  PubMed Central  Google Scholar 

  42. Vesely S, Klöckner CA. Social desirability in environmental psychology research: three meta-analyses. Front Psychol. 2020;11.

  43. Tan HC, Ho JA, Kumarusamy R, Sambasivan M. Measuring social desirability bias: do the full and short versions of the Marlowe-Crowne Social Desirability scale matter? J Empir Res Hum Res Ethics. 2022;17(3):382–400.

    Article  PubMed  Google Scholar 

  44. Larson KE, Bradshaw CP. Cultural competence and social desirability among practitioners: a systematic review of the literature. Child Youth Serv Rev. 2017;76:100–11.

    Article  Google Scholar 

  45. Koivula A, Räsänen P, Sarpila O. Examining social desirability bias in online and offline surveys. In: Human-Computer Interaction Perspectives on Design: Thematic Area, HCI 2019, Held as Part of the 21st HCI International Conference, HCII 2019, Orlando, FL, USA, July 26–31, 2019, Proceedings, Part I 21. Springer; 2019. p. 145–58.

  46. Heerwegh D. Mode differences between face-to-face and web surveys: an experimental investigation of data quality and social desirability effects. Int J Public Opin Res. 2009;21(1):111–21.

    Article  Google Scholar 

  47. Fincham JE. Response rates and responsiveness for surveys, standards, and the Journal. Am J Pharm Educ. 2008;72(2).

Download references

Acknowledgements

NA.

Funding

This study is funded by the National Medical Research Council of Singapore (NMRC/HSRG/0085/2018).

Author information

Authors and Affiliations

  1. Research Division, Institute of Mental Health, 10 Buangkok View, Singapore, Singapore

    Wen Lin Teh, Edimansyah Abdin, Asharani P.V., Fiona Devi Siva Kumar, Kumarasan Roystonn, Peizhi Wang, Saleha Shafie, Sherilyn Chang, Anitha Jeyagurunathan, Janhavi Ajit Vaingankar & Mythily Subramaniam

  2. Admiralty Medical Centre, Khoo Teck Puat Hospital, 676 Woodlands Drive 71, Singapore, Singapore

    Chee Fang Sum

  3. National Healthcare Group Polyclinics, 3 Fusionopolis Link. Nexus@One-North, Singapore, Singapore

    Eng Sing Lee

  4. Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore, Singapore

    Eng Sing Lee

  5. Saw Swee Hock School of Public Health, National University of Singapore, 12 Science Drive 2, Singapore, Singapore

    Rob M. van Dam

  6. Departments of Exercise and Nutrition Sciences and Epidemiology, Milken Institute School of Public Health, The George Washington University, 950 New Hampshire Ave NW, Washington, DC, United States

    Rob M. van Dam

Authors
  1. Wen Lin Teh
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Edimansyah Abdin
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Asharani P.V.
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Fiona Devi Siva Kumar
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Kumarasan Roystonn
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Peizhi Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Saleha Shafie
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Sherilyn Chang
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Anitha Jeyagurunathan
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Janhavi Ajit Vaingankar
    View author publications

    You can also search for this author in PubMed Google Scholar

  11. Chee Fang Sum
    View author publications

    You can also search for this author in PubMed Google Scholar

  12. Eng Sing Lee
    View author publications

    You can also search for this author in PubMed Google Scholar

  13. Rob M. van Dam
    View author publications

    You can also search for this author in PubMed Google Scholar

  14. Mythily Subramaniam
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

WLT wrote the manuscript and analysed data, EA analysed data, AJ KR PW SS SC APV FD conducted literature reviews and were involved in data collection. APV JAV CFS ESL RVD and MS were involved in the conceptualization, grant applications, and/or had provided intellectual input on the overall design and methodology of the study.

Corresponding author

Correspondence to Wen Lin Teh.

Ethics declarations

Ethics approval and consent to participate

All participants provided written informed consent to participate. All methods were performed in accordance with the relevant guidelines and regulations. The study was approved by the relevant local ethics committees: Institutional Research Review Committee (IRRC) and the National Healthcare Group Domain Specific Review Board (DSRB; Ref no. 2018/00430).

Consent for publication

NA.

Competing interests

None.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Teh, W.L., Abdin, E., P.V., A. et al. Measuring social desirability bias in a multi-ethnic cohort sample: its relationship with self-reported physical activity, dietary habits, and factor structure. BMC Public Health 23, 415 (2023). https://doi.org/10.1186/s12889-023-15309-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12889-023-15309-3

Keywords

BMC Public Health

ISSN: 1471-2458

Contact us