`ALCOHOLISM: CLINICAL A N D EXPERIMENTAL RESEARCH
`
`Vol. 17, No. 4
`July/August 1993
`
`Influence of Socially Desirable Responding in a Study of
`Stress and Substance Abuse
`
`John W. Welte and Marcia Russell
`
`Socially desirable responding is the reluctance to admit unpopular
`beliefs or behavior in order to avoid making a negative impression.
`It poses a problem for researchers who rely on self-report of heavy
`drinking and drug use. The Marlowe-Crowne Social Desirability Scale
`(SDS), which measures socially desirable responding, was adminis-
`tered to 1933 respondents in a general population survey. The
`relationships between SDS and various self-report measures were
`examined. SDS was not correlated with gender or race; it increased
`with greater age and decreased with higher socioeconomic status.
`With age and socioeconomic status controlled, SDS had a moderate
`negative relationship with alcohol and drug use, and a strong nega-
`tive relationship with variables reflecting the expression of anger.
`Correlations between drinking/drug use and such variables as age,
`marital conflict, and stressful life events were not substantially
`changed by controlling for SDS. It was concluded that social desir-
`ability response bias probably results in underestimates of rates of
`heavy drinking and drug use, but does not compromise the study of
`predictors of heavy drinking or drug use.
`Key Words: Marlowe Crowne, Social Desirability, Alcohol, Drug
`Abuse.
`
`S for respondents to avoid admitting unpopular actions
`
`OCIALLY DESIRABLE responding is the tendency
`
`or beliefs in order to favorably impress an interviewer.
`The Marlowe-Crowne Social Desirability Scale (SDS)’
`was developed to measure socially desirable responding.
`It contains 33 categorically strong statements that are
`almost never (or almost always) true, but which tempt
`subjects to respond falsely in order to place themselves in
`a positive light. Examples are: “I have never been irked
`when people expressed ideas very different from my own,”
`and “There have been occasions when I took advantage
`of someone.” The developmental work of Marlowe and
`Crowne showed the SDS to have good internal consistency
`reliability (i.e., the items tended to be positively correlated
`with each other) and to be unrelated to any mental illness.
`The absolute nature of the statements means that few, if
`any, could give socially desirable answers with complete
`honesty. Therefore, the SDS was originally interpreted as
`a measure of bias or contamination in self-report data-
`bias caused by subjects’ lying to make a good impression.
`
`From the Research Institute on Addictions, 1021 Main Street, Buffalo,’
`New York.
`Receivedfor publication January 8, 1992; accepted January 28, 1993
`This study was supported by the National Institute on Alcohol Abuse
`and Alcoholism Grant ROI AA05702 awarded to M.R. This study was
`conducted at the Research Institute on Addictions.
`Reprint requests: John W. Welte, Ph.D.. Research Institute on Addic-
`tions, 1021 Main Street, Buffalo, NY 14203.
`Copyright 0 I993 by The Research Society on Alcoholism.
`750
`
`In previous studies of socially desirable responding it
`was reported that SDS scores increased with age and
`decreased with education, but were not correlated with
`Racial differences in SDS scores were observed by
`some researcher~,~ but not all.* Not surprisingly, socially
`desirable responding has been found to have a negative
`relationship with self-report of undesirable behavior.
`Henly and Winters4 found a strong negative correlation
`with a drug problems scale in a sample of adolescent drug
`abusers. Bradburn et a1.2 found a negative correlation of
`approximately the same magnitude between SDS and
`heavy drinking and SDS and marijuana use.
`The possibility that people lie about socially undesirable
`behavior is of concern to researchers who must evaluate
`the validity of self-reports of such socially undesirable
`behaviors as heavy drinking and drug use. However, the
`correlation between SDS and self-report of socially unde-
`sirable behavior is not necessarily due to contamination.
`High SDS scores could reflect the personality of individ-
`uals highly concerned about societal approval. Such indi-
`viduals might be expected to both represent themselves as
`having socially acceptable behavior and behave in socially
`approved ways.
`Evidence for the latter view has been put forth by
`Bradburn et a1.2 They interpreted the positive relationship
`between age and SDS as a cohort effect caused by older
`people having been brought up to think in categorical
`terms about behavior, arguing that older respondents were
`not portraying themselves as virtuous to impress the in-
`terviewer but that they truly thought of themselves as, for
`example, always being helpful. It was also hypothesized
`that if socially desirable lying were operating, the correla-
`tion between the SDS and marijuana use would be
`stronger than that between the SDS and alcohol, reflecting
`the fact that marijuana is illegal and more strongly pros-
`cribed than drinking.
`The researcher’s dilemma is to determine “whether the
`differences between persons with low and high Marlowe-
`Crowne scores are part of the real variance in (the) data
`or part of the error variance” (Bradburn et al., p. 88).2 In
`the latter case, it would be necessary to adjust socially
`sensi.tive self-report data from subjects with high SDS
`scores in order to discern the true nature of the relation-
`ships under study; in the former, adjusting for SDS scores
`might obscure relationships of interest.
`In this study we will investigate the effect of socially
`desirable responding on self-reports of alcohol and drug
`Alcohol Clin Exp Res. Vol 17, No 4, 1993: pp 758-761
`
`CFAD VI 1032-0001
`
`
`
`SoClAuY DESIRABLE RESPONDING IN STRESS AND SUBSTANCE ABUSE
`use in a general population survey. The contamination
`theory would predict that the SDS would be most highly
`correlated with those variables that involve the most so-
`cially damaging admissions. We will also investigate the
`effect of controlling statistically for SDS on relationships
`between measures of substance abuse and other variables
`of interest.
`
`METHODS
`
`h p l e
`Data for the present analyses came from a representative adult house-
`hold survey of stress, alcohol use, and hypertension conducted in Erie
`County, New York (n = 1933). Respondents were identified using a
`stratified, three-stage probability sampling procedure designed to yield
`approximately equal numbers of African Americans and all others at
`thne levels of education (less than high school, high school, at least some
`college). The overall sample completion rate was 78.3%, with the major-
`ity (84.5%) of noncompletions due to refusals. The present analyses were
`conducted without weighting.
`The sample was 42% White, 52% African American, and 6% others.
`It was 62% female, and ranged in age from 19 to 9 1 years.
`
`Racedures
`Data were collected by a corps of 27 interviewers in the summer and
`MI of 1986. Interviewers received 5 days of intensive training on general
`aad survey-specific interviewing techniques and 3 days of training on
`physical measurement, including blood pressure measurement. Inter-
`views were conducted in respondents’ homes using a highly structured
`interview schedule that included diet, smoking, and physical activity, as
`well as medical history. The entire procedure took about 90 min;
`nspondents were paid $25.
`
`Measures
`*Social desirabilify was measured using a 10-item version of the
`Mdowe-Crowne Social Desirability Scale developed by Strahan and
`Gerbasi? It was demonstrated to have correlations of 0.8 or better with
`tbG low form in various population^.^ To adjust for missing data, the
`~ u m k r of items answered in a socially desirable manner was divided by
`the IIUmbcr of items completed. Thus, SDS scores ranged from 0 to 1 .O.
`showed an Cronbach’s of 0.62, which might be
`Reliability anal*
`characterized as acceptable although not good. Alphas of 0.6 or better
`WCR observed among race- and sex-specific subgroups (data available
`from the authors).
`*Heavy drinking was defined as the number of days in the last 30 that
`five or more drinks were drunk.
`*Drug use was defined as the number of different types of illegal drugs
`used in a respondent’s life.
`*Job prestige rankings were derived from respondents’ reports of their
`job titles and duties using Hollingshead’s6 categories for occupational
`rank.
`*Depression was assessed using a 20-item scale developed by the
`Center for Epidemiological Studies.’
`*Drinking fo cope was estimated using a 5-item scale developed by
`Pobch and Onis* to measure the tendency to use alcohol to relieve
`dis$eJs.
`*Negative lfi? events was used to assess the extent to which the
`nspondent was distressed or upset by events in his/her life. Our measure
`was derived from the life events measure developed by Dohrenwend et
`d.9
`*Sel/.esteem, respondent’s perception of self-worth, was assessed using
`RoSenberg’~’~ measure of self-esteem.
`*Mastery, the extent to which the respondent feels in control of his
`or ber life, was based on Pearlin et al.’s” measure.
`
`7 59
`
`*Socia/ competence was assessed using a measure culled from the
`Social Avoidance and Distress Scale developed by Watson and Friend.”
`This scale reflects ease in social situations.
`.Anger-in/anger-out/anger-reflect measure styles of dealing with an-
`ger. They were assessed using anger-in (suppressing it and “sulking”) and
`anger-out (letting it out) scales developed by Spielberger et al.l3 and an
`anger-reflect (keep cool and think) scale developed by Harburg and
`Gleiberman. l 4
`‘Experience qf anger is a measure developed by Siegel15 to assess the
`frequency, duration, and strength of anger.
`*Marital conflict was estimated using a measure based on the work of
`Kessler. l 6
`
`RESULTS
`Bivariate analyses were done that examined the rela-
`tionship between SDS scores and several important de-
`mographic characteristics. SDS scores were not correlated
`with sex or race. SDS increased markedly with age, how-
`ever. Respondents aged 18-29 answered 5 l % of the items
`in the socially desirable direction, whereas the correspond-
`ing figure for those over 59 was 70%. SDS was negatively
`correlated with both education and income. Higher socio-
`economic status respondents answered fewer items in the
`socially desirable direction.
`Bivariate relationships between SDS and various meas-
`ures from our stress model (listed in “Measures”) were
`examined. Correlation coefficients were used, both Pear-
`son’s correlations ( r ) and partial correlations adjusted for
`age, income and education. These control variables were
`used to ensure against spurious correlations; for example,
`a negative correlation between SDS and substance abuse
`created solely because older respondents have both high
`SDS and low substance abuse. A comparison of the un-
`adjusted and adjusted correlations showed that holding
`age, education, and income constant generally lowered
`the absolute value of the correlation slightly, but did not
`change the basic pattern.
`The SDS was negatively associated with substance use,
`as expected. The magnitude of the correlation increased
`as level of social disapproval likely to be associated with
`the use increased. For cigarette smoking the association
`was not statistically significant, and for alcohol-related
`variables the magnitude of the correlations was in the
`modest -0.09 to -0.17 range. However, the correlation
`with number of illegal drugs used in the respondent’s
`lifetime was considerably higher at -0.27. Desirable per-
`sonality traits such as self-esteem, mastery, and social
`competence tended to be positively associated with socially
`desirable responding, whereas the experience of anger, a
`less desirable trait, was negatively associated. Sources of
`stress such as negative life events, depression, and marital
`conflict were negatively associated with the SDS. The SDS
`was positively associated with active coping styles, prob-
`lem-solving and anger-reflect, and negatively associated
`with avoidant styles of coping, anger-in, and anger-out.
`As mentioned earlier, adjusting for the effect of age, edu-
`cation, family income, and job prestige had no influence
`
`CFAD VI 1032-0002
`
`
`
`760
`
`WELTE AND RUSSELL
`
`on the direction of the relationships and little influence
`on the strength of the observed associations.
`The relationship of socially desirable responding to self-
`report of substance abuse is examined in greater detail in
`Table 1, in which the unadjusted and adjusted mean days
`of heavy drinking/month and lifetime use of illegal drug
`types are summarized for 10 levels of the SDS. Displayed
`in this manner, the relationships seem much stronger. For
`example, respondents in the lowest tenth of SDS reported
`an average of 0.96 different types of illegal drugs, whereas
`those in the top tenth report an average of only 0.02. As
`in the bivariate analyses mentioned earlier, adjusting for
`the effect of age, education, family income, and job pres-
`tige attenuated the relationship between substance abuse
`and SDS somewhat, but it was still strong. Therefore,
`confounding by age and socioeconomic status does not
`appear to account for the correlations between the SDS
`and alcohol/drug use.
`The possibility that the correlations observed between
`SDS and self-report of alcohol/drug use indicate a contam-
`ination effect cannot be ruled out. Therefore, we must be
`concerned about the impact of socially desirable respond-
`ing on the type of relational analyses often performed with
`such data. Table 2 shows the correlations of five variables
`with heavy drinking and illegal drug use. These five were
`selected a priori because their correlations with substance
`abuse have theoretical interest and are commonly studied.
`The correlations with SDS scores partialled out are also
`shown. None were changed significantly.
`Table 3 illustrates more directly the point made in Table
`2. For example, a marked decline in drug use with increas-
`
`Social
`desirability
`(deciles)
`
`Table 1. Heavy Drinking and Illegal Drug Use by Level of Social Desirability
`Frequency of heavy
`drinking (5+ drinks, mean
`daysponth)
`Unadjusted
`Adjusted'
`Unadjusted
`3.3
`3.0
`0.96
`1
`0.83
`2.6
`2.2
`2
`3
`1.6
`1.8
`0.78
`2.5
`4
`0.54
`2.7
`2.3
`5
`0.56
`2.7
`1.9
`6
`0.37
`1.9
`7
`1.3
`0.29
`1.5
`1 .I
`0.7
`8
`0.23
`0.9
`9
`0.23
`0.9
`1 .o
`10
`0.02
`0.6
`' Adjusted for age, education, family income, and job prestige.
`
`(mean no.
`''legal drug
`used in lifetime)
`
`Adjusted'
`0.78
`0.81
`0.64
`0.60
`0.60
`0.41
`0.44
`0.40
`0.39
`0.24
`
`Table 2. Influence Of Social Desirability on the Relationship of Heavy Drinking
`and Drug Use with Selected Variables (correlation coefficients)
`
`mean days/month)
`
`Unadjusted
`
`Adjusted
`for SDS
`
`Unadjusted
`
`Adjusted
`for SDS
`
`-0.05
`0.20'
`0.01
`0.30'
`0.05
`
`-0.02
`0.19'
`0.01
`0.29'
`0.03
`
`-0.44'
`0.16.
`0.07
`0.15'
`0.18'
`
`~~
`
`-0.39'
`0.15'
`0.06
`0.11'
`0.15'
`
`Age
`Cigarettes/day
`Negative life events
`Drinking to cope
`Marital conflict
`' p < 0.01.
`
`Table 3. Heavy Drinking and Illegal Drug Use Stratified by Social Desirability,
`Age, and Drinking to Cope
`Social desirability
`Medium
`
`~~
`
`High
`
`Low
`
`Frequency of heavy drinking.
`Age
`18-30
`31-50
`50+
`Drinking to cope
`Low
`Medium
`High
`Illegal drug uset
`Age
`18-30
`31 -50
`50+
`Drinking to cope
`Low
`Medium
`High
`Days drink 5+/month
`0
`1-5
`6+
`
`2.4
`1.9
`0.9
`
`0.54
`1.01
`4.13
`
`1.1
`0.8
`0.1 1
`
`0.51
`0.77
`1.18
`
`0.62
`1.29
`1.31
`
`2.6
`2.3
`1.7
`
`0.29
`2.26
`4.99
`
`0.8
`0.5
`0.08
`
`0.26
`0.64
`0.68
`
`0.34
`0.86
`0.98
`
`1.3
`1.3
`0.63
`
`0.24
`0.86
`3.46
`
`0.6
`0.4
`0.02
`
`0.20
`0.36
`0.34
`
`0.18
`0.62
`0.62
`
`* Cell entries are mean dayslmonth drank 5 or more drinks.
`t Cell entries are mean number of drug types used in lifetime.
`
`ing age was observed at each level of SDS, although the
`mean number of drugs reported was much lower for high
`SDS respondents. Put another way, the conclusion about
`the relationship between age and drug use would be the
`same whether the sample was composed solely of respond-
`ents with low SDS scores or solely of respondents with
`high SDS scores. This also holds true for the relationship
`between drinking to cope and substance abuse and that
`between heavy drinking and drug use.
`
`DISCUSSION
`These analyses indicated that socially desirable respond-
`ing was significantly correlated with self-reported sub-
`stance use and measures from our stress model, but that
`it did not appreciably influence correlational analyses of
`substance abuse with other factors of,interest. These results
`offer some hope that high SDS might not be just a contam-
`ination factor, but reflects a person who both answers and
`acts in a socially desirable manner. The strongest correla-
`tions with the SDS were variables that involve anger. The
`highest correlation was with anger-out, a high score on
`which means that the respondent admitted to doing things
`such as slamming doors and making sarcastic remarks.
`The correlation with depression was much lower, even
`though the depression measure includes items such as "I
`thought my life had been a failure." To the extent that
`most people would rather admit to sarcasm than to their
`life being a failure, the contamination view would predict
`that the SDS would be more highly correlated with depres-
`sion than with anger-out. The result observed is clearly
`not consistent with the contamination hypothesis. Anger-
`out is directly related to the presentation of oneself in a
`
`CFAD VI 1032-0003
`
`
`
`SOCIALLY DESIRABLE RESPONDING IN STRESS AND SUBSTANCE ABUSE
`
`76 I
`
`social situation. and the visible display of petulance may
`be disturbing to people who actually behave in a socially
`approved manner.
`Also encouraging is the observation that, even assuming
`the contamination view to be correct. many relational
`analyses do not seem to be affected. For example, we have
`demonstrated that the tendency for older persons to report
`less drug use is still strong after controlling for SDS,
`strengthening our belief that they really use fewer drugs
`than young adults. However, to the extent that socially
`desirable responding reflects contamination of the data.
`point estimates of behavior such as drug use will be greatly
`affected, even if the conclusions drawn from relational
`analyses don’t change. A glance at Table 1 shows that self-
`reported drug use and heavy drinking are much lower for
`those with high SDS scores. These results are in fact
`consistent with the conventional wisdom about drinking
`surveys. From comparisons with excise tax data, we know
`that drinking surveys underestimate average alcohol con-
`sumption,” although correlations with alcohol consump-
`tion are widely assumed to be valid.
`
`REFERENCES
`I. Crowne DP. Marlowe D: A new scale of social desirability
`independent of psychopathology. J Consult Psychol 24:349-354, 1960
`2. Bradburn NM. Sudman S. Blair E. Locander W. Miles C. Singer
`E, Stocking C: Reinterpreting the Marlowe-Crowne Scale. in Bradburn
`NM, Sudman S (eds): Improving Intenleu Method and Questionnaire
`Design. San Francisco, Jossey-Bass Publishers. 198 I. pp 85-106
`3. Massen D. Hornstra RK, Anderson PB: Influence of social
`desirability on symptom and mood reporting in a community sample. J
`Consult Clin Psychol 43:488-452. 1975
`
`4. Henly G A ? Winters KC: Development of problem severity scales
`for the assessment of adolescent alcohol and drug abuse. Int J Addict
`23:05-85. 1988
`5. Strahan R. Gerbasi KC: Short. homogeneous versions of the
`Marlowe-Crowne social desirabllit> scale. .I CIin Psbchol ?8:191-193.
`1972
`6. Hollinghhead AH: Four factor index ofsoc~al status. Department
`of Sociology. Yale Ilniversity, unpublished manuscript. 1975
`7. Radloff’ LS: The CES-D Scale: A self-report depression scale for
`rescarch in the general population. Appl Psychol Measurement 1385-
`401. 1977
`8. Polich JM. Orvis BR: Alcohol problems: Patterns and prevalence
`in the U.S. Air Force (Rept. no. R-2308-AF). Santa Monica. CA. Rand
`Corporation, 1979
`9. Dohrenwend BS. Krasnoff‘ L, Askenas) AR. Dohrenwend BP:
`Exemplification o f a method for scaling life events: The PER1 Life Events
`Scale. J Health Soc Behav 19:2Oj-229. 1978
`10. Roscnberg M: Society and the Adolescent Self-Image. Princeton.
`NJ. Princeton llnivcrsity Press. 1Y65
`11. Pearlin LI. Menaghan EG, Lieberman MA. Mullan JT: The
`stress process. J Health Soc Behav 12337-356, 1981
`11. Watson D. Friend R: Measurement of social-evaluative anxiety.
`J Consult Clin Psycho1 33:448-457, I969
`13. Spielbcrger CD. Johnson Efj. Russell SF, Crane RJ. Jacobs GA,
`Worden TJ: The experience and expression of anger: Construction and
`validation of an anger expression scale, in Chesney MA, Rosenman RH
`(eds): Anger and Hostilit? in Cardiovascular and Behavioral Disorders.
`Washington, DC, Hemisphere. 1985. pp 5-30
`14. Harburg E. Gleiberman I-: Unpubliahcd questionnaire. Ann
`Arbor, MI, Institute for Social Rcsearch. Liniversity of Michigan. 1986
`15. Siege1 JM: The measurement of anger as a multidimensional
`construct. in Chesney MA, Roscnman R11 (cds): Anger and Hostility in
`Cardiovascular and Behavioral Disorders. Washington. DC. Hemisphere.
`1985. pp 59-81
`16. Kcssler RC: Personal communication. 1986
`17. Midanik L: The validity of self-reported alcohol consumption
`and alcohol problems: A literature review. Br J Addict 77:357-382, 1982
`
`CFAD VI 1032-0004
`
`