Children's Social Desirability and Dietary Reports

Abstract

We investigated telephone administration of the Children's Social Desirability (CSD) scale and our adaptation for children of the Social Desirability for Food scale (C-SDF). Each of 100 4th-graders completed 2 telephone interviews 28 days apart. CSD scores had adequate internal consistency and test—retest reliability, and a 14-item subset was identified that sufficiently measures the same construct. Our C-SDF scale performed less well in terms of internal consistency and test—retest reliability; factor analysis revealed 2 factors, 1 of which was moderately related to the CSD. The 14-item subset of the CSD scale may help researchers understand error in children's dietary reports.

INTRODUCTION

Research on personality has identified characteristics on which individuals vary systematically; among these is the tendency to respond in a socially desirable way. Socially desirable responding is presumed when an individual reports never performing a behavior that most everyone performs at least occasionally or always performing a behavior that most people usually perform but omit occasionally. Individuals who tend to respond in a socially desirable way may err systematically in responding to a variety of questions, including questions about dietary intake; thus, social desirability is an example of a response bias.¹

Among adults, social desirability is most commonly measured using the Marlowe-Crowne Social Desirability (MCSD) scale,² which consists of 33 items that describe either desirable but uncommon everyday behaviors (eg, admitting mistakes) or undesirable but common everyday behaviors (eg, gossiping).¹ Possible scores range from 0 to 33, with higher scores indicating a higher tendency to respond in a socially desirable way. High test—retest reliability³ is consistent with the presumption that an individual's level of social desirability is an enduring characteristic.

Crandall et al⁴ used the MCSD scale as a model to develop 2 versions of the Children's Social Desirability (CSD) scale: the version for grades 6 through 12 has 48 true-false items, and the version for grades 3 through 5 has 46 yes-no items. In the latter version, shown in Table 1, possible scores range from 0 to 46, with higher scores indicating a higher tendency toward socially desirable responding.⁵ Crandall et al evaluated the reliability of the CSD scale with a total sample of 956 children in grades 3 through 12, including about 110 children in each of grades 3 through 5.⁴ Spearman-Brown—corrected split-half reliabilities were high, ranging from .82 to .95 for subsamples of males and females at various grade levels. One-month test—retest reliabilities on subsamples of 63 younger children and 98 10th-graders were .90 and .85, respectively. Thus, children's social desirability, as measured by the CSD, was found to be both internally consistent and consistent over a 1-month interval. Socially desirable responses were more frequently given by younger than by older children, by less intelligent than by more intelligent children, by females than by males, and by black than by white children.⁴ Klein et al replicated the age and gender findings of Crandall et al⁴ with 1,008 males and females ages 7 to 14.⁶

Table 1.

Items on the Children's Social Desirability (CSD) Scale,⁵ with Factor Loadings and Endorsement Rates from Our Study with 100 4th-Grade Children^.

		Loadings		Endorse
Item		A1	A2	A1	A2
1.	Do you ever get angry if you have to stop in the middle of something you're doing to eat dinner or go to school? (N)	30	63	42	48
2.	Does it sometimes bother you to share your things with your friends? (N)	30	50	65	72
3.	Do you always enjoy yourself at a party? (Y)	31	40	91	83
4.	Are you always polite to older people? (Y)	38	45	82	80
5.	Do you sometimes tell a little lie? (N)	26	42	13	25
6.	Do you ever hit a boy or a girl who is smaller than you? (N)	16	42	85	83
7.	Sometimes, do you feel like doing other things instead of what your teacher wants you to do? (N)	41	69	55	54
8.	Do you ever act “fresh” or “talk back” to your mother or father? (N)	26	51	85	83
9.	When you make a mistake, do you always admit that you are wrong? (Y)	30	52	76	71
10.	Do you feel that your parents always show good judgment (ie, do they always make good choices)? (Y)	38	53	92	86
*11.	Have you ever felt like saying unkind things to a person? (N)	45	67	55	54
12.	Have you sometimes felt like throwing or breaking things? (N)	42	38	76	69
13.	Do you ever let someone else get blamed for what you do wrong? (N)	41	43	87	83
*14.	Are you always careful about keeping your clothing neat and your room picked up? (Y)	49	55	78	72
15.	Do you ever shout when you feel angry? (N)	27	46	64	57
*16.	Do you sometimes feel like staying home from school even if you are not sick? (N)	49	57	38	48
17.	Sometimes do you wish your parents didn't check up on you so closely? (N)	29	45	55	58
18.	Do you always help people who need help? (Y)	50	49	86	79
19.	Do you sometimes argue with your mother to let you do something she doesn't want you to do? (N)	34	44	77	78
*20.	Do you ever say anything that makes somebody else feel bad? (N)	63	65	74	68
21.	Do you think your teachers know more about everything than you do? (Y)	-14	-10	73	62
*22.	Are you always polite, even to people who are not very nice? (Y)	47	50	68	68
*23.	Sometimes do you do things you've been told not to do? (N)	53	69	37	50
24.	Do you ever get angry? (N)	21	30	10	16
25.	Do you sometimes want to own things just because your friends have them? (N)	42	37	63	54
*26.	Do you always listen to your parents? (Y)	61	56	78	79
27.	Do you ever forget to say “please” and “thank you”? (N)	31	42	47	44
*28.	Do you sometimes wish you could just play around instead of having to go to school? (N)	57	64	43	50
29.	Do you always wash your hands before every meal? (Y)	52	49	85	81
30.	Do you sometimes dislike helping your parents even though you know they need your help around the house? (N)	33	9	35	43
31.	Do you ever find it hard to make friends? (N)	19	12	38	58
*32.	Have you ever broken a rule? (N)	51	60	34	37
33.	Sometimes do you try to get even when someone does something to you that you don't like? (N)	45	53	42	48
*34.	Do you sometimes feel angry when you don't get your way? (N)	50	70	49	46
35.	Do you always help a hurt animal? (Y)	16	4	39	29
36.	Do you sometimes want to do things your parents think you are too young to do? (N)	21	51	55	53
*37.	Do you sometimes feel like making fun of other people? (N)	62	70	80	72
38.	Have you ever borrowed anything without asking permission first? (N)	35	61	73	72
39.	Do you sometimes get mad when someone disturbs something you've been working on? (N)	18	45	19	22
40.	Are you always glad to cooperate with others? (Y)	37	51	90	78
41.	Do you ever get angry when your best friend wants to do something you don't want to do? (N)	34	33	59	61
42.	Do you sometimes wish that the other kids would pay more attention to what you say? (N)	32	33	40	44
*43.	Do you always do the right things? (Y)	64	53	58	49
44.	Are there some times when you don't like to do what your parents tell you? (Mind your parents?) (N)	56	62	45	41
45.	Are there times that you don't like it if somebody asks you to do something for him? (N)	36	60	51	56
*46.	Do you sometimes get mad when people don't do what you want them to do? (N)	56	58	59	56

The Y or N in parentheses after each item indicates whether the yes or no answer is the socially desirable response. Loadings are on the first (only retained) factor, multiplied by 100. Endorse is the percentage of children who gave the socially desirable response. Al indicates first administration; A2, second administration. Items with an asterisk before the question were selected for the 14-item subset of the CSD.

Some evidence for the validity of the CSD scale is available: Crandall showed that CSD scores for 10th-graders were related in the predicted direction to certain subscales of the California Personality Inventory (social presence, self-acceptance, good impression, self-control); for younger children, CSD scores were related to certain social and achievement behaviors in free play.⁷ For children from 4th to 12th grade, CSD scores were negatively associated with achievement test performance. Klein et al found that social desirability was associated with children's inability to express dislike for peers.⁶

It is plausible that social desirability would be related to dietary reports in general and to dietary reporting error in particular. Because many foods are regarded generally as “good” or “bad”,^8-12 a respondent who tends to answer in a socially desirable way might underreport consumption of “bad” foods and overreport consumption of “good” foods.^10-12 Worsley et al developed a scale to measure Social Desirability for Foods (SDF) and tested it with 309 Australian Air Force recruits; a subsample of 96 recruits also completed the MCSD scale.¹⁰ SDF scores correlated positively to both MCSD scores and self-reported consumption of fresh fruit and vegetables on a food frequency questionnaire (FFQ) and correlated negatively to FFQ-reported consumption of snack foods. Worsley et al concluded that psychological characteristics such as social desirability might influence individuals' dietary reports. (Note that they did not assess response validity or the relationship of social desirability to reporting error.)

Over the past 10 years, evidence collected concerning the relationship of social desirability to adults' dietary reports has tended to show a negative association between social desirability and reporting accuracy, particularly for females,^13-17 although one study found no effect of social desirability on adults' dietary reports.¹⁸ Investigators of children's diets have also been concerned about the relationship of social desirability to children's dietary reports^19-21 but have neither assessed social desirability and children's self-reported dietary behaviors concurrently nor attempted to relate children's dietary reporting errors to social desirability. Anticipating such investigations, we carried out this study to (1) assess test—retest reliability of telephone administration of the CSD scale; (2) adapt Worsley et al's SDF scale¹⁰ for children and assess test—retest reliability of telephone administration of the resulting Children's Social Desirability for Foods (CSDF) scale (Table 2); (3) evaluate the relationship between the CSD and the C-SDF scales; and (4) identify a subset of CSD scale items to permit efficient measurement of social desirability.

Table 2.

Items on a Children's Social Desirability for Food (C-SDF) Scale Created by Adapting Worsley et al's¹⁰ Social Desirability for Food (SDF) Scale^.

1.	Do you usually drink all of your milk? (Y)
2.	Do you usually eat breakfast every day? (Y)
3.	Do you often eat candy? (N)
4.	Do you often watch TV or read while you eat a meal or snack? (N)
5.	Do you always wash your hands before every meal? (Y)
6.	Do you usually eat all of your vegetables? (Y)
7.	Do you usually drink regular soft drinks or sodas? (N)
8.	Do you usually eat your meals or snacks in a hurry? (N)
9.	Are your table manners at home as good as when you eat at school or in a restaurant? (Y)
10.	Do you usually eat fast foods? (N)
11.	Do you usually “pig out” or eat too much? (N)
12.	Do you always brush your teeth after each meal? (Y)
13.	Do you always eat everything on your plate even if you're not hungry? (Y)
14.	Do you usually eat chips? (N)

The Y or N in parentheses after each item indicates whether the yes or no answer is the socially desirable response.

DESCRIPTION OF THE EVALUATION

The study had Institutional Review Board approval. During the spring of 2002, all children from the 14 4th-grade classes at 3 public elementary schools in 1 district in a southeastern state were invited to participate. At the 3 schools, 96%, 91%, and 69%, respectively, of the children across all grades were eligible to receive free or reduced-price school meals. Of the 262 children invited to participate, 170 (65%) provided written child assent and parental consent. Of these 170 children, each of 100 randomly selected children was interviewed twice, with a minimum of 28 days between any child's 2 interviews. Each child's 2 interviews were conducted by a different 1 of 3 research dietitians. Interviews were conducted by telephone in the evening. Children were told that they would be asked “questions about things that happen to all children your age.” The 46-item CSD scale was administered first, followed by the 14-item C-SDF scale. Beginning and ending times were recorded, and interviews were audiorecorded. Children were paid $10 per interview.

Quality control for interviews was assessed daily throughout data collection. QCI assessed on 69 random interviews (35%) showed that interviewers adequately followed the protocol.

Answers that matched the socially desirable choice were scored 1 point. Possible scores ranged from 0 to 46 for the CSD scale and from 0 to 14 for our C-SDF scale, with higher scores indicating a greater tendency toward socially desirable responding. For each administration of each scale, internal consistency was estimated by Cronbach α. Test—retest reliability for each scale was estimated by Pearson's correlation coefficient. Mean scores and administration times for the repeat administrations were compared using paired t tests. Means by gender were compared using t tests. Factor analysis was used to investigate the latent structure of the scales and to guide the selection of items for a shortened social desirability scale.

LESSONS LEARNED

Across children, a mean of 29.34 days (minimum = 28, maximum = 40) separated the 2 interviews. The mean total length of questionnaire administration (for the 60 items of the 2 scales) was 10 minutes (minimum = 7, maximum = 16) and 9 minutes (minimum = 7, maximum = 15) for the first and second interviews, respectively.

CSD Scale

For the CSD scale, α's were .88 and .93 for the first and second administrations, respectively. Test—retest reliability was.79. Table 3 shows means and standard deviations of scores on the CSD scale by administration and gender. Mean scores of males and females did not differ significantly in either the first or second administration (P = .96 and .48, respectively), and overall means did not differ between the 2 administrations (P = .72).

Table 3.

Means (and Standard Deviations) by Administration and Gender for the Children's Social Desirability (CSD) Scale, the Children's Social Desirability for Food (C-SDF) Scale, and a 14-Item Subset of the CSD Scale^.

	Males	Females	All Childern
Administration	(n = 53)	(n = 47)	(n=100)
CSD Scale
(See Table 1 for 46 items on CSD Scale. Scores may range from 0 to 46.)
1	27.49 (7.43)	27.43 (8.79)	27.46 (8.06)
2	27.92 (9.89)	26.47 (10.76)	27.24 (10.28)
C-SDF Scale
(See Table 2 for 14 items on C-SDF Scale. Scores may range from 0 to 14.)
1	8.85 (2.27)	7.96 (2.57)	8.43 (2.45)
2	8.77 (2.60)	7.94 (2.69)	8.38 (2.66)
14-Item Subset of CSD Scale
(This subset consists of CSD scale items 11, 14, 16, 20, 22, 23, 26, 28, 32, 34, 37, 43, 44, and 46. Scores may range from 0 to 14.)
1	7.96 (3.46)	7.96 (4.25)	7.96 (3.83)
2	8.11 (4.08)	7.74 (4.54)	7.94 (4.28)

The sample size was 100 4th-graders with 10 white males, 5 white females, 39 black males, 37 black females, 4 other males, and 5 other females. For each scale, higher scores indicate a higher tendency toward socially desirable responding.

To examine the latent structure of the scale, factors were extracted from the data from each administration using the principal factors method with squared multiple correlations as prior communality estimates. In each set of data, the proportion of common variance accounted for by the first factor was vastly larger than accounted for by the second and subsequent factors. For the first administration, the first factor accounted for 31% of the common variance and the second accounted for only 9% (total common variance = 24.8); for the second administration, the first and second factors accounted for 40% and 10%, respectively, of the common variance (total common variance = 29.2). This pattern suggested that the scale measures a dominant construct and that we might select a subset of items that would measure social desirability in children efficiently yet adequately.

Items were selected for the shortened scale if they had high loadings on the first extracted factor on both administrations of the CSD and nonextreme endorsement rates (which is required for an item to contribute to differentiating among respondents); in addition, the set was selected to mirror the full scale in the proportion of items keyed “yes” for social desirability. Table 1 shows first-factor loadings and endorsement rates for each item in each administration. We identified 14 items from the 46-item CSD scale (items 11, 14, 16, 20, 22, 23, 26, 28, 32, 34, 37, 43, 44, and 46; see Table 1) according to these criteria. Ten of these items were among the 15 highest-loading items on the first extracted factor on both administrations of the scale; 3 were among the 15 highest-loading items on 1 administration and among the 17 highest loading on the other; and 1 item loaded adequately on the first factor on both administrations, although it was not among the highest loading. The median loadings of these items on the first factor extracted from the data of the first and second administrations were .55 (range .45 to .64) and .62 (range .51 to .71), respectively. In addition, the endorsement rate for each of these items on each administration, was between .2 and .8, and most rates were well within these limits. For these 14 items, the median endorsement rates of the more frequent response were .62 and .58 on the first and second administrations, respectively; for the other 32 items, the comparable statistics were .73 and .71. Four of the 14 items (29%) are keyed “yes” for social desirability; on the full 46-item scale, 28% of the items are keyed “yes”. Table 3 shows means and standard deviations of scores on the 14- item subset of the CSD scale by administration and gender. Means of males and females did not differ significantly in either the first or second administration (P = .99 and .67, respectively), and the difference between the means for the 2 administrations was not significant (P = .93). For the 14-item subset of the CSD scale, test—retest reliability was .83.

C-SDF Scale

For the C-SDF scale, for the first and second administrations, respectively, α's were .57 and .63, and correlations between the C-SDF scale and the CSD scale were .49 and .59. Test—retest reliability was .64.Table 3 shows means and standard deviations of scores on the C-SDF scale by administration and gender. Mean scores of males and females did not differ significantly in either the first or second administration (P = .07 and .12, respectively), and overall means did not differ between the 2 administrations (P = .29).

Factor analysis of the data from each administration using the principal factors method with squared multiple correlations as prior communality estimates suggested an underlying 2-factor structure. For the first and second administrations, the first and second extracted factors together accounted for 84% and 96%, respectively, of the common variance (total common variance for the first and second administrations was 3.28 and 3.81, respectively). After application of a varimax rotation to the 2-factor solution from each administration, we identified 7 items (3, 4, 7, 8, 10, 11, 14) as loading strongly on 1 factor and 5 items (1, 5, 6, 12, 13) as loading strongly on the other. (Two items, 2 and 9, loaded strongly on neither factor.) Test—retest reliability for each of these subsets was .71. One subset of items, which includes questions about drinking milk, eating vegetables, finishing all food, washing hands, and brushing teeth, is related to social desirability in the sense measured by the CSD scale; the correlations between this subset of items and the CSD scale were .60 and .64 for the 2 administrations, respectively. In contrast, the other subset of items, which includes questions about eating candy, chips, fast food, and drinking soda; eating in a hurry; eating too much; and watching television or reading while eating, was weakly related to the CSD scale, with correlations of .22 and .24 on the first and second administrations, respectively.

One item, concerning washing hands before every meal, appeared on both the CSD and C-SDF scales; during the interviews, these items were generally separated by 3 to 4 minutes. Thus, these items provide an opportunity to examine intra-administration reliability (for a single item) and provide a frame for expectations about scale reliability with 4th-grade children. On the first administration, 7 children changed their responses, giving a repeated-item reliability of .72; on the second administration, 4 children changed their response (1 of these was among the 7 who had done so on the first administration), giving a repeated-item reliability of .87.

DISCUSSION

For the sample of primarily black (76%) 4th-graders from whom data were collected, test—retest reliability was adequate for telephone administration of the CSD scale (.79). In addition, we identified a 14-item subset of the CSD scale that appears to measure the same construct and that also has adequate test—retest reliability (.83). These levels of reliability with retest after at least 28 days indicate that children's social desirability, measured by telephone administration of the CSD, is a reasonably stable characteristic. Our C-SDF scale, which is an adaptation for children of the SDF scale of Worsley et al,¹⁰ performed less well in terms of internal consistency and test—retest reliability. However, exploration of its latent structure revealed 2 factors, of which 1 was and 1 was not related to the CSD scale. Whether 1 or both of these subscales merits further study depends on their relevance to investigators' conceptualizations of sources of dietary reporting error. Further testing, including validation, is needed before the C-SDF scale is ready to use.

The children in the sample were from 3 schools in 1 school district and were primarily of 1 race; these study characteristics may limit generalizability of the results. Another limitation is that test—retest reliability of the 14-item subset of the CSD scale was not assessed by itself; responses to these items were collected only in the context of the other 32 items of the CSD scale. In addition, the C-SDF scale was always administered after the CSD scale.

The relatively poor performance of the C-SDF scale may have been due to the use of the word “usually” in many of the questions; this term may not be sufficiently absolute to identify socially desirable responding. In the CSD scale, all but 1 of the questions keyed “yes” include the term “always,” and all but 1 of the questions keyed “no” include either “sometimes” or “ever.” In contrast, 3 items on Worsley et al's¹⁰ SDF scale for adults include “always” or “never,” and the remaining 9 items include a variety of less absolute terms, such as “usually,” “rarely,” “often,” and “hardly ever.”

IMPLICATIONS FOR RESEARCH AND PRACTICE

There is a growing interest in measuring social desirability in studies of children's dietary reports and in exploring relatively stable personality characteristics as correlates of children's dietary reporting error. Many studies by us and others involve collecting dietary data from children by telephone^22-27; in this article, we estimated test—retest reliability for tele-phone administration of the CSD scale. (The test—retest reliability estimates of Crandall et al⁴ were from in-person administration of tape-recorded items.) In addition, we identified a subset of CSD items that should permit efficient, yet adequate, measurement of social desirability in children. A 46-item instrument is burdensome in research situations in which a variety of data must be collected under time constraints. Short forms of the MCSD scale have been developed^28-31; a short form of the CSD scale was needed as well.

Validation studies that compare self-reports of diet with a method (such as direct observation) that does not rely on memory indicate that the accuracy of children's dietary recalls is poor.^19,22,32-35 In adults, dietary reporting error has been shown to be related systematically to social desirability, a stable characteristic that differentiates individuals. It is plausible that social desirability is a systematic correlate of error in children's dietary reports as well. We recommend use of the 14-item subset of the CSD scale to investigate this issue.