Ideally, studies should control for social desirability bias (the tendency for responses to reflect what is presumed to be desired, rather than the truth) through overall design. Specifically, the survey questions and design should not reveal the purpose of the study or, in the case of animal or vegetarian/vegan intervention studies, the fact that it is being conducted by an animal advocacy group, prior to asking questions believed to be affected by social desirability bias. In the event that such control is not possible, we propose controlling for social desirability bias through scores on the following short form of the Marlowe-Crowne scale (Reynolds’s Form C), also presented on our complete list of recommended questions. Of the validated social desirability scales available, we feel this offers the best balance of length and desirable statistical characteristics.
Marlowe-Crowne Scale (Reynolds’s Form C)
Listed below are a number of statements concerning personal attitudes and traits. Read each item and decide whether the statement is True or False as it pertains to you personally.
Note: Response options of True and False should be provided for each statement
- It is sometimes hard for me to go on with my work if I am not encouraged.
- I sometimes feel resentful when I don’t get my way.
- On a few occasions, I have given up doing something because I thought too little of my ability.
- There have been times when I felt like rebelling against people in authority even though I knew they were right.
- No matter who I’m talking to, I’m always a good listener.
- There have been occasions when I took advantage of someone.
- I’m always willing to admit it when I make a mistake.
- I sometimes try to get even rather than forgive and forget.
- I am always courteous, even to people who are disagreeable.
- I have never been irked when people expressed ideas very different from my own.
- There have been times when I was quite jealous of the good fortune of others.
- I am sometimes irritated by people who ask favors of me.
- I have never deliberately said something that hurt someone’s feelings.
Assign each respondent a social desirability score based on their answers to the questions on the scale.
- Add 1 point to the score for each “True” response to statements 5, 7, 9, 10, and 13. Add 0 points to the score for each “False” response to these statements.
- Add 1 point to the score for each “False” response to statements 1,2,3,4,6, 8, 11, and 12. Add 0 points to the score for each “True” response to these statements.
Each respondent should now have a social desirability score between 0 and 13. These scores are intended to measure how likely the respondent is to give answers that sound good instead of answers that are true. While most people will answer in the socially desirable way to some questions, and some people really will have more of the “good” traits than others, those respondents with especially high scores may have obtained them by answering in ways that exaggerate their good qualities and minimize their bad ones. This is the same behavior we would be worried about leading to under-reporting of animal product consumption if the respondents know the survey is being carried out by an animal advocacy group.
Detecting Social Desirability Bias in Responses
Social desirability scores cannot be used to identify participants whose answers to other questions should or should not be trusted, because any individual respondent may have a high score either by actually having many socially desirable traits or by exhibiting social desirability bias in their answers. Instead, responses to other important questions (or other derived outcome measures) should be checked for correlation with social desirability scores. Questions for which the responses are strongly correlated with respondents’ social desirability scores are more likely to be affected by social desirability bias than questions where the correlation is weak, although strong correlation does not guarantee that responses were inaccurate and weak correlation does not guarantee that responses were accurate. This is a population-level phenomenon; if everyone who responds to a question in a certain way has a high social desirability score, the results of that question will be flagged as potentially influenced by social desirability. If there is no group trend relating high social desirability scores to certain answers, each respondent’s answer will be deemed trustworthy, regardless of their social desirability score.
Testing for correlation is a little more complicated than figuring out what percentage of respondents chose which answer. If you need to, you can get help with it from a group like Statistics Without Borders. If you choose to analyze your data yourself, the steps of testing for correlation are:
1. Make sure responses to the question of interest are on a numeric scale, with an order that makes sense. For instance, if considering an item on a food frequency scale, use the daily frequency of consumption or daily amount of consumption for each respondent. If the question of interest has only two answers, assign the value 0 to one answer and 1 to the other. If the question of interest has answers that do not make sense on a numeric scale (such as “Cat, Dog, Other, None”), it is necessary to treat each answer as a separate item, assigning a value of 1 when it is chosen and 0 when it is not.
2. Use a spreadsheet program or statistical software to find the correlation coefficient between the responses to the item in question and the respondents’ social desirability scores. Most software will by default find Pearson’s correlation coefficient, which ranges between -1 and 1.
3. Assess whether the correlation coefficient causes concern. If the answers to the question are not related to respondents’ social desirability scores, the correlation coefficient should be near 0. A coefficient near either 1 or -1 indicates a strong linear relationship between two variables. One way to understand the impact of a correlation of is that, if the correlation exists because respondents’ social desirability scores are affecting their other answers, those scores explain of the variation in the other question’s answers. For instance, if the correlation coefficient is .4, social desirability scores explain .16 or 16% of the variability in responses to the question. It can also be helpful to make a scatterplot with the responses to the question plotted against the social desirability scores. This can give a sense of the shape of the data, which puts the correlation coefficient in perspective and also allows nonlinear relationships to be seen, if they exist. Remember that if you are finding the correlation coefficient for a large number of comparisons, one or more is likely to be somewhat large by chance even if nothing is truly correlated; you will need to report this, but you do not need to be concerned about it.
4. Report the correlation with social desirability scores for any results you consider important. There is no amount of correlation that indicates whether a result is valid or invalid, but providing the actual correlation coefficients with your results gives important context for people interpreting your study.
Alternatively, if you are seeking to explain variation in a single characteristic based on an intervention and controlling for extraneous variables (gender, age, etc.) to account for outside influences, social desirability score would be an appropriate control variable to add to your analysis.
A social desirability scale can be useful for studies that do not control for social desirability bias through overall design, or in order to provide extra evidence that social desirability bias was not a large factor in the results obtained. The above scale is relatively short and provides a way to report on potential contamination of self-reported data by social desirability bias.
Barger, S. D. (2002). The Marlowe-Crowne affair: Short forms, psychometric structure, and social desirability. Journal of Personality Assessment, 79(2), 286-305.
Nederhof, A. J. (1985). Methods of coping with social desirability bias: A review. European Journal of Social Psychology, 15(3), 263-280.
Reynolds, W. M. (1982). Development of reliable and valid short forms of the Marlowe-Crowne social desirability scale. Journal of Clinical Psychology, 38(1), 119-125.