Social desirability biases in self-reported alcohol consumption and harms

Addictive Behaviors 35 (2010) 302–311

Contents lists available at ScienceDirect

Addictive Behaviors

Social desirability biases in self-reported alcohol consumption and harms Christopher G. Davis a,b,⁎, Jennifer Thake a, Natalie Vilhena a,1 a b

Department of Psychology, Carleton University, Ottawa, Ontario, Canada Canadian Centre on Substance Abuse, Ottawa, Canada

a r t i c l e

i n f o

Keywords: Alcohol drinking patterns Screening Response bias

a b s t r a c t Aims: Self-reports remain the most common means of assessing alcohol consumption despite concern for their validity. The objective of this research is to assess the extent to which social desirability biases relate to self-reported consumption, hazardous use, and harms. Methods: In each of two studies presented, undergraduate students (N = 391 and N = 177) who reported that they had consumed alcohol in the past year completed online confidential surveys. Results: Both studies show consistent associations between impression management bias and self-reported consumption such that high impression managers report 20 to 33% less consumption and are about 50% less likely to report risky drinking. No significant correlations involving consumption were found for selfdeception bias. Study 2 also indicated that high impression managers report 30–50% fewer acute harms following a drinking episode, and that these effects are maintained after controlling statistically for trait impulsivity/constraint. Conclusions: Impression management bias represents a significant threat to the validity of self-reported alcohol use and harms. Such bias may lead to misspecification of models and under-estimates of harmful or hazardous use. © 2009 Elsevier Ltd. All rights reserved.

1. Introduction Surveys are frequently used to assess whether and the extent to which people use illicit drugs and alcohol, engage in risky sexual practices, or participate in other activities that may be illegal, highly personal or sensitive. Prevalence rates based on self-reports of such behavior, however, are regarded generally with some skepticism on the assumption that some respondents are unwilling to divulge such information, cannot recall accurately, or do not answer honestly owing to a desire to present themselves in a favorable light. It is important to estimate the extent of bias in these reports, and to reduce them, because such reports are often used for planning and policy purposes and for testing theoretical models (e.g. Cooper, 2002; Taylor, Rehm, Patra, Popova, & Baliunas, 2007). For example, when researchers attempt to assess the social costs of different consumption patterns, they rely on self-reports. In their analyses of the effect of alcohol consumption on health care costs, Taylor et al. (2007) applied a correction factor of 2.7 to self-reported quantity–frequency estimates for heavy drinkers, suggesting that heavy drinkers (but not light drinkers) are grossly underestimating their consumption. If

⁎ Corresponding author. Department of Psychology, Carleton University, 1125 Colonel By Drive, Ottawa, Ontario, Canada K1S 5B6. E-mail address: [email protected] (C.G. Davis). 1 Now at the University of Toronto, Toronto, Ontario, Canada. 0306-4603/$ – see front matter © 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.addbeh.2009.11.001

some heavy (or moderate) drinkers have understated their consumption, and are thus misclassified, the accuracy of the model is reduced. Several researchers have demonstrated that self-reports of consumption do not adequately capture how much alcohol is consumed (e.g., Knibbe & Bloomfield, 2001). Stockwell and colleagues (Stockwell et al., 2004; Stockwell, Zhao, Chikritzhs, & Greenfield, 2008) have compared self-reported consumption patterns from nationally representative samples to aggregate sales data, and report that the common quantity–frequency measure generally captures only about 50% of alcohol sales, suggesting significant under-reporting. Stockwell et al. (2008) report that other methods, such as the Yesterday method (where population estimates are based on each participant's recall of how much he or she drank yesterday) and graduated frequencies (where participants are asked to recall how often over a specified period of time they consumed various quantities of alcoholic beverages, starting with large quantities and working down to small quantities), tend to yield somewhat better estimates, but still significantly under-estimate actual sales data. Others have also shown that accuracy of reported consumption decreases the further back one is asked to recall. Gmel and Daeppen (2007) have shown that reports of daily consumption become progressively lower over the space of a week, arguing that this represents a reporting bias. These studies suggest that under-reporting occurs, but they do not indicate where, or in which segments of the population, misreporting is most prevalent. The issue of veracity of self-reported alcohol use has shifted from whether reports are accurate to analysis of the factors that affect the

C.G. Davis et al. / Addictive Behaviors 35 (2010) 302–311

degree of accuracy. Del Boca and Darkes (2003) have argued that contextual factors play a significant role in the accuracy of selfreported data on alcohol consumption. For example, prevalence rates vary as a function of the assessment method and whether anonymity or confidentiality is assured. In a recent review of this literature, Tourangeau and Yan (2007) found that self- and computer administered anonymous surveys yield higher endorsement rates on sensitive topics than do face-to-face, telephone, or non-anonymous interviews. It is important to note, however, that using online anonymous surveys to collect data on such topics may not entirely alleviate the problem (see e.g., Adams, Parkinson, Sanson-Fisher, & Walsh, 2008; BoothKewley, Larson, & Miyoshi, 2007). In the studies we report below, we take the approach that individuals differ in the extent to which they are willing to be forthright in their reports of alcohol use and related harms. That is, we examine the extent to which reports of alcohol consumption and harms are related to an independent assessment of one's tendency to respond in a socially desirable way. A socially desirable response is one that is influenced by the perception that others will be evaluating that response. Someone who is responding in a socially desirable way might tailor his or her reported attitude or behavior to conform to what he or she thinks is appropriate, acceptable, or desired by others. 1.1. Assessment of social desirability biases The assessment of the degree to which an individual is responding in a socially desirable way has been an issue for personality and clinical psychologists for more than 60 years. Concerned about lying and faking on the Minnesota Multiphasic Personality Inventory, Meehl and Hathaway (1946) empirically developed a scale using items that distinguished individuals with a “tendency to be defensive or to put oneself in too favorable a light” (p. 560). Crowne and Marlowe (1960) likewise developed a set of items, some from the MMPI, which reflected a motivation to respond in a socially desirable way. People who score high on the well-known Marlowe–Crowne Social Desirability Scale (M–C SDS) are susceptible to social influence and conformity—they want approval from others—and demonstrate this by tending to endorse self-descriptions that are too good to be true. From this perspective, the tendency to present oneself in a favorable light is regarded as an individual difference variable. People who score high on the Marlowe–Crowne Social Desirability Scale tend to adjust their responses to conform to the wishes or expectations of others whom they wish to please. To our knowledge, only one study has assessed the extent to which socially desirable response tendencies relate to reports of consumption of alcohol. Welte and Russell (1993) used a short form of the M–C SDS in a general population survey of alcohol and stress. They found that individuals scoring high on the abbreviated M–C SDS tended to report one half to one third the number of heavy drinking occasions (≥5 drinks) than those scoring low on the M–C SDS. They reported that alcohol-related variables correlated in the range of −.09 to −.17 with M–C SDS scores. The M–C SDS, however, has been criticized for its inability to distinguish distinct biases: the tendency to intentionally present oneself favorably, to appear good in the eyes of others (referred to as impression management), and an unconscious tendency to think of one's self in an overly positive way (i.e., to self-enhance or selfdeceive; Paulhus, 1984). Paulhus (1984, 1986; Paulhus & Reid, 1991) developed and validated a questionnaire that distinguishes these two motivations, known as the Balanced Inventory of Desirable Responding (BIDR). The BIDR represents an advance over the earlier M–C SDS insofar as it assesses separately impression management and selfdeception, is balanced in terms of positively-keyed items (i.e., endorsing good qualities) and negatively-keyed items (i.e., denying undesirable qualities), and does not contain items that directly reflect adjustment. The BIDR has now been used in hundreds of personality

303

and attitudinal research studies and is widely regarded as the best instrument for assessing these two sources of bias (e.g., Olson, Goffin, & Haynes, 2007). Below we describe these two components of social desirability. Impression management reflects a tendency to self-attribute saintly or virtuous characteristics and deny socially deviant impulses or behaviors (e.g., “I always pick up my litter on the street”). In contrast, self-deception reflects a tendency to exaggerate (unconsciously) desirable qualities (e.g., “my first impressions always turn out to be right”; Paulhus, 2002). Consistent with this, impression management scores are more responsive to instructions to “fake good” and role-playing instructions (e.g., imagine you are applying for a job) than are self-deception scores (Paulhus, Bruce, & Trapnell, 1995). Self-deception scores, on the other hand, tend to be more strongly related to measures of self-esteem and adjustment (Lanyon & Carle, 2007; Paulhus, 1991). To the extent that people care a great deal about the impression they are making on others, we expect that they will be motivated to downplay or “low-ball” the extent to which they consume alcohol and experience harms from use. Impression managers do not want to appear to have a “drinking problem” because to do so would be stigmatizing and unattractive. Significant associations are not anticipated for the relation of alcohol consumption to selfdeception. That is, there is no reason to believe that people understating their consumption are doing so because they view themselves as actually drinking less. However, people who do engage in selfdeception may be motivated to unconsciously deny the experience of harmful consequences as these drinking outcomes are especially difficult to integrate into their positive self view. Surprisingly, the BIDR has not been used as a means of assessing the extent to which reports of alcohol and/or drug use are underestimated. The BIDR has, however, been used to assess the extent to which self-deception and impression management tendencies correlate with reports of sexual attitudes and behavior among university students (Meston, Heiman, Trapnell, & Paulhus, 1998). Meston and colleagues found that self-deceptive enhancement was generally uncorrelated with reports of sexual behavior, but found modest but consistent correlations in the range of |.13| to |.26| between impression management scores and reports of sexual behavior, such that those with high impression management scores were more likely to report less sexual experience. Controlling for individual differences in personality traits, including conservatism, did not significantly attenuate these effects. It is important to note that impression management represents a conscious attempt to downplay or under-estimate what one perceives to be socially deviant behavior. To the extent that moderate consumption of alcohol is perceived to be socially acceptable, an impression manager is not likely to feel the need to under-estimate his or her consumption. It is only when consumption goes beyond what one considers socially acceptable will he or she be motivated to understate his or her consumption. Thus, we anticipate that impression managers will be inclined to under-estimate reports of quantity and frequency of consumption and the number of times they drink heavily. 1.2. Overview of studies In the research presented here we examine the extent to which self-deception and impression management scores correlate with selfreports of alcohol use in two independent university student samples. Our focus in Study 1 is on assessing the extent to which impression management and self-deception biases are associated with selfreported daily consumption, typical consumption, frequency of heavy drinking, and risky drinking. Our focus in Study 2 is on assessing the extent to which these biases are associated with reports of alcoholrelated harms and the extent to which these biases confound previously observed links between heavy consumption and harms.

304

C.G. Davis et al. / Addictive Behaviors 35 (2010) 302–311

2. Study 1 2.1. Introduction to Study 1 The purpose of the first study was to assess the extent to which self-reported drinking behavior correlates with the tendency to respond in a socially desirable way. Self-reported alcohol consumption was assessed in terms of typical drinking behavior as well as in terms of recall of daily drinking over the past week. We also assessed risky or harmful drinking with a standard measure of hazardous drinking (the Alcohol Use Disorders Identification Test; Saunders, Aasland, Babor, de la Fuente, & Grant, 1993). As noted above, we hypothesize that individuals who score high on impression management will report consuming less alcohol and report fewer alcoholrelated problems than those scoring low on impression management. We also anticipate that people who are impression managers will be particularly responsive to cues that they are drinking too much alcohol. One such cue comes in the form of daily reports of drinking, that is asking respondents how many alcoholic drinks they consumed on each day over the past week, starting with yesterday (e.g., Gmel & Daeppen, 2007). Impression managers are likely to be keenly aware that they appear to be drinking too much as they endorse multiple consecutive days of drinking. We hypothesize, therefore, that impression managers will be apt to report less alcohol consumption than others as they recall further back in the week. Moreover, we hypothesize that impression managers will be particularly likely to under-report drinking on days when they drink more heavily, as heavy drinking is more likely than moderate drinking to be perceived by impression managers as deviant. 2.2. Method 2.2.1. Participants Participants were 408 undergraduate students at a large Canadian university. Seventeen participants who reported consuming no alcohol in the past 12 months were excluded from further analyses, leaving a final sample consisting of 391 participants (65% female, M age = 19.6, SD = 2.3). Forty-one percent of participants (n = 157) were under the legal drinking age of 19 years. 2.2.2. Procedures Students in the Introductory Psychology subject pool volunteered to complete our Alcohol and Drug Use survey in exchange for a modest grade-raising credit. They were assured confidentiality during the informed consent process and were free to withdraw from the study or to skip any question they did not want to answer. Participants completed the survey online; there was no face-to-face contact between participants and experimenters. To facilitate accurate responding participants were assured that their names were only linked to their data in order to assign credit for completing the study, but would subsequently be deleted making responses anonymous. 2.2.3. Measures The online survey included a short demographics section, followed by modules on use of alcohol, cannabis, motives for use of each, harms, and social desirability, in that order. Our focus in this paper is on the relations of social desirability to alcohol use, and so we do not here describe other measures. More than 75% of participants completed the survey within 30 min. Alcohol use was assessed with questions used in prior population surveys (e.g., Adlaf, Begin & Sawka, 2005). In the alcohol module, respondents were first asked whether they had ever consumed alcoholic beverages. To assess frequency and quantity of alcohol consumed, participants were asked: (1) “How often did you drink alcoholic beverages in the past 12 months?” (range for response options: 1 = “less than once a month” to 7 = “every day”, as well as a

last category of “don't know”; M = 3.12 [corresponding to about 2–3 times/month], SD = 1.36)2; (2) “How often in the past 12 months have you had 5 or more drinks on one occasion?” if they were male or “… 4 or more drinks…” if they were female (range for response options: 1 = “never” to 4 = “more than once a week”, plus a “don't know” option; Mmales = 2.81, SD = 1.07; Mfemales = 2.78, SD = 0.99); and (3) “During the past 12 months, on those days when you drank, how many drinks did you usually have?” (M = 5.19 drinks, SD = 3.13).3 Participants were subsequently asked if they had consumed alcohol in the past week. If so (n = 207), they were then asked how many drinks they consumed yesterday, the day before, and so on for the past seven days. On average, past week drinkers reported consuming 9.52 alcoholic drinks over the week (SD = 9.27), and reported drinking on average 2.06 days per week (SD = 1.21). To measure risky or hazardous drinking, participants completed the Alcohol Use Disorders Identification Test (AUDIT; Saunders et al., 1993). The AUDIT was originally developed as a brief screening tool for use in primary care settings to detect individuals at risk for abuse. It has since been used, however, in a number of large general population and college student studies to assess risky or hazardous alcohol use (e.g., Fleming, Barry, & MacDonald, 1991; Kills Small, Simons, & Sticherz, 2007). The AUDIT includes three questions assessing consumption, three items assessing dependence, and four items assessing consequences. Most researchers report analyses based on the AUDIT by either summing the ten items or distinguishing those who exceed and do not exceed a cutoff score of 8 (where those scoring 8 or higher are considered at-risk drinkers; see Berner, Kriston, Bentele, & Harter, 2007, for a systematic review), although some factor analytic studies suggest that the instrument represents two dimensions (alcohol consumption and alcohol-related problems; see Doyle, Donovan, & Kivlahan, 2007; Shevlin & Smith, 2007). Both the full scale and the two component parts have good internal consistency (α N 0.80 for the unidimensional scale, α N 0.70 for the two subscales; Shields, Guttmannova, & Caruso, 2004; Thomas & McCambridge, 2008), and a great deal of research in clinical, general population, and college student samples supports its validity (e.g., Babor, Higgins-Biddle, Saunders, & Monteiro, 2001; Berner et al., 2007; Kokotailo et al., 2004). Since the first three questions of the AUDIT are nearly identical to the quantity– frequency questions asked above, these questions were not asked a second time in the AUDIT. Rather, responses to those questions were reformatted to correspond to the categories used in the AUDIT (e.g., the typical number of drinks per occasion were recoded into the AUDIT's categories of 1–2, 3–4, 5–6, 7–9, and 10+ drinks) for the purpose of calculating an overall AUDIT score. The remainder of the AUDIT was as specified in the original. Impression management and self-deception were assessed with the Balanced Inventory of Desirable Responding (BIDR-6; Paulhus, 1991). In this instrument, participants are asked to rate the extent to which they agree or disagree (on a 7-point scale) with each of 40 statements, none of which make any reference to alcohol. Twenty of these items reflect a tendency to self-enhance or deny undesirable qualities (half positively-keyed, half negatively-keyed; e.g., “I never regret my decisions”), and 20 items reflect a tendency to put across a good, socially acceptable impression of oneself (again, half positivelykeyed, half negatively-keyed; e.g., “I sometimes tell lies if I have to”). Scores on self-deception and impression management are determined by counting the number of items on which the individual responds in an extreme way (strongly disagree [1 or 2] on negatively-keyed items; strongly agree [6 or 7] on positively-keyed items). Scores obtained from this sample (M = 5.52, SD = 3.45 on self-deception and M = 2 “Drink” was defined for participants as: 1 bottle or can of beer or glass of draft; 1 glass of wine or wine cooler; or 1 drink or cocktail with 1.5 oz of liquor. 3 Extreme scores (≥20 drinks/day) were shifted to within 3 standard deviations of the mean.

C.G. Davis et al. / Addictive Behaviors 35 (2010) 302–311

5.26, SD = 3.39 on impression management) are comparable to those reported by Paulhus (1991) in other undergraduate samples (M = 6.8, SD = 3.1 on self-deception and M = 4.9, SD = 3.2 on impression management, n = 433). Cronbach's alpha for the subscales in the present sample was 0.72 for self-deception and 0.73 for impression management, which are comparable to those found in other reports (Paulhus, 1991). The subscales of the BIDR have been extensively validated (see Paulhus, 1991; Paulhus et al., 1995). For example, in an extensive validation study involving forensic and student samples, Lanyon and Carle (2007) reported that the impression management scale of the BIDR consistently correlates strongly (r N .50) with other measures of exaggerated virtue, whereas the self-deception scale consistently correlates strongly with indicators of exaggerated adjustment.

305

Table 1 Correlations between social desirability and self-reported alcohol consumption. Drinking item

Impression management r

Selfdeception r

Frequency of drinking in past 12 months (Males) Frequency of 5+ drinks/occasion (12 months) (Females) Frequency of 4+ drinks/occasion (12 months) Quantity of drinks/occasion (12 months) Number of drinks in past week

−.20⁎⁎⁎ −.16⁎

.01 −.06

−.28⁎⁎⁎

−.05

−.16⁎⁎ −.16⁎⁎

.05 .01

Notes. ns range from 358 to 372 for items 1, 4, 5; ns for males = 134 and females = 253. ⁎ p b .10. ⁎⁎ p b .01. ⁎⁎⁎ p b .001.

2.3. Results 2.3.1. Missing data analyses An analysis was conducted to assess the extent to which responses were missing or answered with a “don't know” response. For alcoholrelated questions, between 3% and 6% of responses were identified as missing or unknown per question. The extent to which one left a question blank, or indicated “don't know” was not significantly correlated with impression management score (r = −.05, ns) but was weakly correlated with self-deception scores (r = −.11, p b .05). 2.3.2. Effects of gender and legal drinking age Neither past 12 months drinking frequency nor the frequency with which one consumed 5+/4+ drinks in a sitting differed significantly as a function of gender or whether participant was of legal drinking age (ts b 1.2). However, in terms of typical consumption, males reported consuming significantly more drinks per drinking day than females (Mmales = 6.32, SD = 3.95 vs. Mfemales = 4.61, SD = 2.42; t = 4.44, p b .001). Typical consumption did not vary as a function of whether participant was of legal drinking age (t b 1). Males who reported drinking in the past week also reported consuming more drinks than females Mmales = 12.45, SD = 10.57 vs. Mfemales = 8.01, SD = 8.17; t = 3.07, p b .01). There were no differences as a function of whether participant was of legal drinking age (t b 1.2). With respect to risky alcohol use as measured by the AUDIT, participants scored an average of 8.95 (SD = 5.33, range 0–27, n = 382). Mean AUDIT scores did not differ by gender or drinking age (legal vs. underage) (ts ≤ 1.50, p N .10). Sex and age differences were found for social desirability scores. Females scored higher on impression management than males (M = 5.54, SD = 3.38 vs. M = 4.72, SD = 3.36; t = 2.27, p b .05), whereas males scored higher on self-deception (M = 6.04, SD = 3.79 vs. M = 5.24, SD = 3.23, t = 2.18, p b .05). Underage participants scored lower on impression management than drinking age participants (M = 4.64, SD = 3.00 vs. M = 5.67, SD = 3.57; t = 2.97, p b .01), and scored marginally lower on self-deception (M = 5.13, SD = 3.46 vs. M = 5.78, SD = 3.43; t = 1.85, p b .10). 2.3.3. Social desirability and alcohol consumption The main objective of the current study was to assess the extent to which impression management and self-deception tendencies correlate with self-reported alcohol consumption. Table 1 shows that scores on impression management, but not self-deception, were modestly but consistently correlated with measures of self-reported drinking. Those who scored higher on impression management reported drinking less. Next, we assessed whether the link between impression management and self-reported alcohol consumption was attenuated or moderated by gender or age. Statistically controlling for gender and age did not significantly reduce any of these effects. Regression analyses revealed no significant interactions of gender or age with impression management when predicting drinking frequency (past 12 months) or

number of drinks in the past week. However, a significant gender by impression management interaction was found for quantity of drinks/ occasion (β = .11, p b .05), such that the effect of impression management was stronger for males (β = −.21, p b .05) than it was for females (β = −.08, p = .21). To give a better sense of the difference in self-reported consumption of those who score high on impression management relative to others, impression management scores were divided into thirds. The results (summarized in Fig. 1) show a consistent pattern: those who score in the top tertile of the distribution on impression management report drinking less often and in smaller quantities (20–33% less) than those in the middle or bottom tertiles of the distribution. 2.3.4. Effect of impression management on daily reports To assess the effect of impression management on daily self-reports of alcohol consumption, a two-level hierarchical linear (growth) model was constructed where each individual's retrospective daily reported alcohol consumption (among those reporting past week alcohol consumption) was predicted from day of the week (i.e., Sunday, Monday, etc.) and the number of days since the day the participant completed the questionnaire (i.e., yesterday, the day before, etc.). According to participants, prime drinking days were Thursday through Saturday. Table 2 gives the means and standard deviations for number of self-reported drinks per day of the week. Given the evident quadratic pattern, our model includes a quadratic function (i.e., days of the week squared, where Sunday is Day −3 and Saturday is Day + 3). This quadratic function allowed us to model the steep increase in drinking that occurs at the end of the week. The first (within-subject) level of our two-level model involved modeling the effect of days since the assessment day (i.e., “days

Fig. 1. Error bars represent standard errors. For the dependent variable “how often drink”, 1 = “less than once/month,” 2 = “once/month,” 3 = “2–3 times/month,” and 4 = “once/week;” for other dependent variables, scores refer to number of alcoholic drinks consumed (truncated at 20).

306

C.G. Davis et al. / Addictive Behaviors 35 (2010) 302–311

Table 2 Mean number of alcoholic drinks consumed by day of the week. Day of the week

Mean number of drinks

Sunday Monday Tuesday Wednesday Thursday Friday Saturday

SD

0.89 0.46 0.67 0.71 1.61 2.51 2.80

2.29 1.52 1.84 1.88 3.24 3.80 3.88

n = 203–207.

since”) on self-reported consumption. This model also included the effect of day of the week and the effect of “day of the week—squared”. (Note that all predictor variables were centered about 0, as recommended by Aiken & West, 1991.) Thus, each individual has his/her own regression equation: Drink quantityij = b0j + b1j ðDays sinceij Þ + b2j ðDay of weekij Þ

ð1Þ

2

+ b3j ðDay of weekij Þ + eij where Drink quantityij represents the estimated number of drinks personj reported on dayi, b0j represents the intercept for personj, b1j represents the effect (unstandardized regression coefficient) of days since for personj, b2j represents the linear effect of day of the week (Sunday through to Saturday) for personj, b3j represents the effect of day of the week squared for personj, and eij represents the within person residual or error. To test whether impression management moderates the effects of days since, day of the week, and the day of the week squared on reports of daily consumption, impression management scores were used to predict the individual intercepts (b0) and slopes of day of the week and days since effects (b1–b3). That is, in these analyses we ask whether people who score high on the impression management scale report drinking less than those who score low on the impression management scale, and whether they report less drinking than those scoring low on impression management on days further back in time (the “days since” slope). In these analyses, day of the week is treated as a linear variable centered about the midday of the week (i.e., Sunday = −3, Monday = −2, …, and Saturday = 3). Values for all parameters in the 2-level model are estimated simultaneously with the first level model using HLM6.06 (Raudenbush, Bryk, & Congdon, 2008). The results given in Table 3 show that both the linear and quadratic effect of days of the week were signifi-

cant (b = .381, se = .035 and b = .107, se = .021, ps b .001), indicating that quantity of drinks increased greatly after midweek. The linear effect was moderated, however, by impression management (b = −.028, se = .011, p b .05) indicating that high impression managers tended to report less drinking later in the week (i.e., on heavier drinking days) than low impression managers. In addition to the day of the week effects, we also obtained a significant Days Since by impression management interaction (b = −.027, se = .012, p b .05) indicating that as high impression managers recall further back in the week, they tend to report less drinking than low impression managers. The combined effects of the two interactions yields an interesting pattern depicted in Fig. 2. High impression managers report fewer drinks consumed the further back in time they recall whereas low impression managers report more drinks consumed the further back in time they recall. The model predicts that someone scoring at +1 SD of the mean on impression management (a score of approximately 7) recalls drinking 2.3 drinks on Saturday night seven days ago, whereas someone scoring at − 1 SD of the mean (a score of approximately 2) recalls drinking 3.9 drinks on the same Saturday night seven days ago. If, in contrast, Saturday night was last night, the high impression manager estimated consuming 2.6 drinks to the low impression manager's estimate of 3.3 drinks. Interestingly, early in the week, high impression managers tend to report a little more drinking than low impression managers, but only if those days were more recent. A parallel model constructed with self-deception scores as the moderator variable yielded no significant effects involving this variable (ts b 1). 2.3.5. Social desirability and AUDIT scores In addition to self-reported quantity and frequency of alcohol consumption, impression management and self-deception scores were also correlated with AUDIT scores. Bivariate correlations revealed a significant correlation between total AUDIT scores and impression management, r(382) = −0.25, p b .001, such that those scoring high on impression management tended to obtain lower scores on the AUDIT. Self-deception scores were marginally correlated with AUDIT scores (r = −.10, p = .10). There was no significant interaction of gender or drinking age with impression management (p N .50). Because the correlation observed between AUDIT scores and impression management scores may have been due to the already established link between consumption and impression management, we correlated impression management scores to the sum of the seven harmful use and dependency items on the AUDIT. The sum of these AUDIT items negatively correlated with both impression management (r = −.26,

Table 3 Coefficients for hierarchical linear model on self-reported number of drinks consumed. Effect Intercept Intercept by Imp Mgmt Days since Days since by Imp Mgmt Day of the week Day of the week by Imp Mgmt Day of the week2 Day of the week2 by Imp Mgmt

b

se

t-ratio

df

p-value

0.947 − 0.037 0.025 − 0.027 0.381 − 0.028 0.107 − 0.007

0.124 0.040 0.036 0.012 0.035 0.011 0.021 0.007

7.63 − 0.94 0.69 − 2.33 10.86 − 2.53 5.16 − 1.12

203 203 1418 1418 1418 1418 1418 1418

b 0.001 ns ns 0.020 b 0.001 0.012 b 0.001 ns

Note: n = 205. Imp Mgmt refers to impression management score. All independent variables are grand mean centered. Dependent variable is self-reported drinks consumed on dayj.

Fig. 2. High impression management is defined as + 1 SD of the mean; low impression management is defined as − 1 SD of the mean. Arrows indicate that as day of recall increases from yesterday to the day before, and so on, the estimated number of drinks shifts in the direction of the arrow: high impression managers recall less consumption as one goes back in time whereas low impression managers recall greater consumption.

C.G. Davis et al. / Addictive Behaviors 35 (2010) 302–311

p b .001) and self-deception scores (r = −.17, p = .001). No significant interactions involving gender or drinking age were observed. The mean difference in AUDIT scores of those in the top third of the distribution on impression management scores from those in the middle or bottom third of the distribution was 2.35 and 2.87 points, respectively (F(2, 379) = 11.16, p b .001). Most high impression managers do not score at or above the criterion for hazardous drinking (43% score 8+), whereas most of those scoring low or moderate in the impression management scale do score above the criterion (65% and 59%, respectively, Χ2 (df = 2) = 12.1, p b .01). 2.4. Discussion Study 1 suggests that even in the context of an online, confidential survey, individuals with an impression management bias consistently under-estimate their reports of alcohol consumption. Impression management bias appears to be greatest on recall of more distant drinking events and on traditional higher-consumption days (Thursday through Saturday), suggesting that impression managers don't mind reporting (modest) alcohol consumption in the recent past, but appear to become more modest in their recall as they go back in time. Although the correlation estimates are modest (rs in the range of −.15 to −.25), one should not assume that they are inconsequential. Those who are in the top third of the distribution on impression management report that they drank on average 33% less in the past week; they reported drinking at least one less drink (about 20% less) per drinking episode; they reported that they typically drink less frequently; and they are about 50% as likely to meet criteria for hazardous drinking relative to those in the middle or bottom thirds of the distribution. It might be argued, however, that individuals who score high on our measure of impression management really do drink less. This implies that the impression management subscale of the BIDR is confounded with some personality trait, such as conservatism or constraint. That is, people who are socially conservative or unimpulsive may both be less likely to drink excessively and more likely to want to put across a good impression. There are at least two arguments against this possibility. First, prior research by Meston et al. (1998) showed that covarying scores on major traits as well as a measure of conservatism did not attenuate correlations of impression management with self-reported sexual activity. Second, the personality explanation cannot account for the tendency of people who score high on impression management to progressively under-estimate their consumption the further back in time they are recalling. On the other hand, it makes sense that impression managers who admit to one day of drinking would be less likely to reporting a second drinking day that week. To do so might suggest that one drinks too much. Nevertheless, to squarely address this concern, in our second study, we include a measure of impulsivity– constraint to demonstrate that the link between impression management and self-reported consumption of alcohol is not due to this trait. Although drinking variables were consistently negatively correlated with impression management scores, we found no evidence of such links with self-deception scores. However, both higher impression management and higher self-deception scores were significantly associated with fewer self-reported indicators of harmful use. That self-deception bias emerges as significant on the AUDIT problems subscale but not on consumption is in step with the notion that selfdeceivers tend to be somewhat oblivious to their problems. They will admit to drinking, but tend not to acknowledge its negative effects. 3. Study 2 3.1. Introduction to Study 2 The primary goals of Study 2 are to (1) assess the extent to which impression management and self-deception biases relate to reports of

307

alcohol consumption on a single drinking occasion and subsequent acute alcohol-related harms, and (2) estimate the effect that this potential confound has on the link between heavy drinking and selfreported alcohol-related harms. Ganster, Hennessey, and Luthans (1983) noted that when social desirability is associated with both the independent and dependent variables (or correlated variables of interest), the observed correlation is likely inaccurate. A secondary goal of Study 2 is to address the possibility that the consistent effects involving impression management that we observed in Study 1 are attributable to some personality construct other than impression management. That is, it might be the case that individuals who are (truly) highly principled and self-disciplined tend to score high on impression management and low on our measures of alcohol consumption. A number of studies have indicated that people who are impulsive tend to consume more alcohol, use illicit drugs, and experience more harms from use (e.g., Magid, MacLean, & Colder, 2007). Impulsivity, as a personality trait, is scored along a continuum where the low end of the distribution is referred to as impulse control, constraint, or selfcontrol (for a review, see Carver, 2005). That is, those who are not impulsive tend to be inhibited, reflective and deliberate in their behavior, and demonstrate self-control. Block and Block (2006) refer to these individuals as ego-controlled, suggesting that they are able to curb impulses to act inappropriately and recklessly. Indeed, Hampson, Severson, Burns, Slovic, and Fisher (2001) demonstrated that adolescents who are ego-controlled report less consumption of alcohol and report fewer incidences of risky behavior. From the perspective of the current research, people who score low on trait measures of impulsivity (which we refer henceforth as those with the trait of constraint) may report drinking fewer drinks and report fewer harms because they are more likely to act carefully, limit their consumption of alcohol, and consider future consequences when drinking, rather than taking risks and acting on impulse. In Study 2, we assess the extent to which trait impulsivity–constraint is associated with consumption of alcohol on a specified occasion, harms attributable to that drinking event, and impression management scores. We then assess the extent to which the link between impression management and reports of drinking (and alcohol-related harms) is attenuated when impulsivity–constraint scores are controlled statistically. Study 1 revealed that impression management appears to have an effect on estimates of drinking as reports become more retrospective. This is likely due to participants becoming aware that their consecutive reports of daily drinking are beginning to “look bad”, and thus attempt to under-estimate subsequent reports of drinking. In order to reduce this bias, in Study 2 we asked our participants to recall a single, recent drinking occasion. That is, our participants were asked to recall how many drinks they consumed, and what harms they experienced, on a single drinking occasion that occurred this past weekend. Furthermore, to minimize any bias that might be introduced by assessing impression management and drinking behavior in the same survey, in Study 2 we assessed tendency towards impression management and reports of drinking behavior on different days, about a week apart. 3.2. Methods 3.2.1. Participants Participants were 177 undergraduate students at a large Canadian university (76% female; M age = 19.88; SD = 2.28) who reported that they normally drink alcohol at least once per week. Thirty-three percent were under the legal drinking age of 19 years. Eighty-eight percent of participants (n = 155) completed both the pre- and post-drinking surveys, although 12 of these 155 reported not drinking over the weekend, and so did not report any harms. Participants who dropped out of the study before completing the second survey did not differ significantly from remaining participants on demographic variables, the AUDIT or any personality variables assessed in the pre-drinking survey,

308

C.G. Davis et al. / Addictive Behaviors 35 (2010) 302–311

except scoring lower on measures of impression management and selfdeception (M = 3.27, SD = 2.69 vs. M = 5.39, SD = 3.09; t = 3.06, p b .01 for impression management and, M = 3.82, SD= 2.99 vs. M = 5.21, SD = 3.08; t = 2.00, p b .05 for self-deception). 3.2.2. Procedures Students in the Introductory Psychology subject pool were recruited to participate in an online study about alcohol use by university students. Participants were assured confidentiality during the informed consent process and were free to withdraw from the study or to skip any question they did not want to answer. Participants were informed that they would complete one survey now and a second follow-up survey the following week. For the first survey, participants completed a series of online questionnaires assessing personality and the AUDIT. On the following Monday, participants were sent an email requesting them to fill out the follow-up survey within the next couple of days. Because we wanted to avoid problems associated with a long recall, we closed the survey on Wednesday. The median recall interval was 2 days (M = 2.5, SD = 1.2). At the beginning of the second survey, participants were asked if they had consumed alcohol between Thursday and Sunday night. If they had not, they were directed to the final debriefing page and were not asked any further questions. Those who did report drinking alcoholic beverages were asked a series of questions about the one drinking occasion since Thursday when they consumed the most alcohol. 3.2.3. Measures The pre-drinking survey included a short demographics section, followed by the AUDIT, the BIDR, and various measures of personality, of which impulsivity–constraint is most relevant to this report. The postdrinking survey asked a series of questions about the weekend's drinking events, including questions about number of drinks consumed, perceived intoxication, motives for drinking, strategies used to decrease alcohol-related consequences, and a questionnaire assessing alcoholrelated consequences. Our focus in this paper is on the relations of social desirability and trait impulsivity–constraint to alcohol consumption and self-reported harmful consequences. Scores on the AUDIT and on the impression management and selfdeception scales of the BIDR were very similar to those obtained in Study 1. Mean AUDIT score in this sample was 8.94 (SD = 5.60), mean impression management score was 5.13 (SD = 3.11), and mean selfdeception score was 5.04 (SD = 3.09). Constraint/Impulsivity was measured in the pretest using the Barratt Impulsiveness Scale (BIS-11; Patton, Stanford, & Barratt, 1995). The BIS11 was chosen as a measure of impulsivity–constraint because items assess both ends of the continuum (e.g., some items ask about effortful self-control and careful and deliberate planning before acting, such as “I am a careful thinker” and “I plan tasks carefully”). No items make reference to alcohol. Participants are asked to indicate how true each item is of them on a 4-point scale ranging from rarely/never to almost always/always. Low scores on this measure indicated impulse control or constraint. This measure has been commonly used in other research assessing the relationship between impulsivity and risk-taking behavior in young adults (e.g., Stanford, Greve, Boudreaux, & Mathias, 1996). Patton et al. (1995) report high internal consistency coefficients for separate populations of under-graduates, substance-abuse patients, general psychiatric patients, and prison inmates. Coefficient alpha for the current study is .85. The mean score for the current sample (M = 65.50, SD = 10.25) is similar to that found in another college sample (Stanford et al., 1996). 3.2.4. Specifics of the acute drinking experience On the post-drinking survey, participants were asked a series of questions pertaining to alcohol consumption over the previous weekend. Participants checked off whether they had consumed alcohol on Thursday (until 4 am Friday), Friday (until 4 am Saturday),

Saturday (until 4 am Sunday) and/or Sunday (until 4 am Monday). They were then instructed to indicate the day on which they consumed the most alcohol (i.e., Thursday, Friday, Saturday or Sunday) and to keep that day in mind when answering the remaining questions: (1) “How many standard drinks do you believe you consumed on that day?” and (2) “At the point in the day in which you had consumed the most alcohol, how intoxicated did you feel on a scale from 1 to 10?” (1 = completely sober and 10 = completely intoxicated). The number of drinks consumed was imputed for two participants who did not know how many drinks they consumed (imputation based on a linear regression of hours spent drinking, how intoxicated they felt they were, and their gender; Multiple R = .78). Excluding these two participants from all analyses had no effect on findings. We developed an index of harms attributable to drinking on a single occasion based on items used in other indices and from prior research on harms experienced by college students (e.g., Gmel, Rehm, Room, & Greenfield, 2000; White & Labouvie, 1989). Our list of 50 potential harms included direct consequences of excessive consumption (e.g., blacking out, vomiting), consequences attributable to alcohol-impaired judgment (e.g., driving under the influence, had unprotected sex, did something I regret), as well as next-day alcohol-related consequences (e.g., missed a day of work or school). For each item, participants were asked whether the event happened, and if so, to rate its seriousness (1= “not serious” to 3 = “a big deal”). The mean number of items endorsed was 6.39 (SD = 6.95), and the mean sum of the items weighted by seriousness was 8.41 (SD = 10.83). Since the number of items endorsed and the sum of the weighted items correlate r = .93, we report only the number of items endorsed. Table 4 indicates the proportion of past weekend drinkers reporting each harm. 3.3. Results 3.3.1. Acute drinking episode Mean consumption (after truncating extreme scores at 20) was 5.90 standard drinks (SD = 3.79, n = 141); however males drank significantly more than females (M = 7.78, SD = 5.23 vs. M = 5.23, SD = 3.09; t = 2.99, p b .01). At the point in which they had consumed the most alcohol, males reported feeling slightly more intoxicated than women, although this effect was marginal (M = 6.14, SD = 2.31 vs. M = 5.33, SD = 2.36; t = 1.80, p b .10). Participants below the legal drinking age did not differ from legal drinking age participants on any of the drinking variables. 3.3.2. Impression management, self-deception, and reports of alcohol use and harms Impression management scores correlated significantly and negatively with the number of drinks consumed (r = −.17, p b .05), how intoxicated one felt one was (r = −.18, p b .05), the number of harms one experienced (r = −.38, p b .001), as well as AUDIT scores (assessed at pretest, r = −.30, p b .001). Self-deception scores were not significantly correlated with any of these measures, although the correlation approached conventional significance levels for harms (r = −.14, p b .10). Consistent with what was observed in Study 1, relative to those in the middle and bottom of the distribution of impression management scores, those scoring in the top third reported 30% less alcohol consumption (4.7 drinks vs. 6.6; F(2,138) = 4.64, p = .01), indicated that they were 25% less intoxicated (4.6 on 10 point scale vs. 6.1; F(2,138) = 6.80, p b .01), reported about half the number of harms (3.9 vs 8.0; F(2,138) = 8.99, p b .001), and scored about 3 points lower on the AUDIT (6.8 vs 9.8; F(2,174) = 10.93, p b .001). 3.3.3. Impulsivity–constraint Scores on impulsivity–constraint were significantly correlated with impression management scores (r = −.37, p b .001), the number of harms (r = .35, p b .001), and AUDIT scores (r = .33, p b .001), but were only marginally associated with the number of drinks consumed

C.G. Davis et al. / Addictive Behaviors 35 (2010) 302–311

309

Table 4 Acute harms from use of alcohol. Harms

N

%

Noticed a change in your personality Experienced a headache(s) related to your use Had trouble sleeping or slept too much Spent too much money on alcohol Felt that you needed more alcohol than usual to get the same effect Neglected your responsibilities Felt inappropriately angry, upset or frustrated with someone Did something that you regret Did not remember passing-out/falling asleep Did not use a seatbelt while in a vehicle Got into a verbal fight/argument with a friend(s) or relative(s) Felt queasy or threw up the next day Caused shame or embarrassment to someone Felt queasy or vomited while drinking Had a bad time Intentionally told someone something that you knew was untrue Was told by someone to stop or cut down on drinking Missed a day (or part of a day) of school or work Kept drinking when you told yourself that you should not Received a noise complaint Got into a verbal fight/argument with a stranger(s) Had unprotected or other risky sexual behavior Found yourself in a place that you could not remember getting to Had open booze in the car Could not function in a social situation due to alcohol Were physically hurt or injured (e.g., bruises, broken bones) Blacked out Disappointed your partner because you chose to drink instead of spending time with them Lost personal belongings while drinking (e.g., wallet, keys) Poor work performance Had to borrow money to pay for alcohol Wanted to stop drinking but couldn't Felt physically or psychologically dependent on alcohol Do not fully remember a sexual interaction you had with someone Felt that you have a problem with alcohol Engaged in sexual activities with a stranger(s) Was asked to leave a social situation due to intoxication Rode in a car in which the driver was likely over the legal limit Chose to use substances rather than go to work Got into a physical fight with a friend(s) or relative(s) Experienced a decreased sex drive or had difficulty becoming sexually aroused/reaching orgasm Do not remember drinking somewhere Went to work under the influence of alcohol Drank alcohol in the morning to stop the negative effects of substances from the night before Drove a car after you might have had too much to drink Fined, arrested or charged with a drinking offence (e.g., public intoxication, assault, disturbing the peace) Property in or outside of the home was destroyed or damaged Was pulled over for dangerous or reckless driving Got into a physical fight with a stranger Was in an accident or close to being involved in an accident because the driver was intoxicated

141 141 140 141 139 140 140 141 140 141 141 140 140 141 141 140 140 140 141 141 141 141 141 141 140 140 141 141

55 38 37 36 24 22 21 20 19 19 19 19 16 16 16 14 14 14 14 13 13 13 12 11 11 11 10 10

140 141 141 140 140 141 141 141 141 141 141 141 141

9 9 8 7 7 7 7 6 6 6 6 6 6

141 141 141

5 5 5

140 140

4 4

141 141 141 141

3 2 2 1

(r = .14, p b .10) and not significantly correlated with perceived level of intoxication (r = .04). Controlling for impulsivity–constraint scores in the ANOVAs reported above did not significantly alter the results (see Fig. 3 for means adjusted for impulsivity–constraint). 3.3.4. Controlling for impression management Since impression management scores were significantly associated with both self-reported consumption and harms, we examined the effect that controlling for impression management scores would have on the correlation among AUDIT scores, number of drinks consumed, and harms. Table 5 shows that partialling impression management scores modestly attenuated these relations. Fig. 4 shows that for participants who score low to moderate on impression management, greater consumption is associated with more harms (r(88) = .29, p b .01). However, for those scoring in the top third of the distribution

Fig. 3. Error bars represent standard errors. Means are adjusted for impulsivity– constraint scores.

on impression management, increasing consumption is not associated with more harms (r(49) = .14, ns). 3.4. Discussion Study 2 provided further evidence that impression managers underestimate reports of drinking. When asked to recall how much they drank (and how intoxicated they felt they were) on a just-past weekend drinking occasion, high impression managers reported drinking 30% fewer drinks and reported being 25% less intoxicated. Impression managers also tend to report significantly fewer negative consequences as a result of their drinking and scored 3 points lower on the AUDIT. These effects held after controlling for trait impulsivity–constraint, suggesting that impression managers do not report drinking less or experiencing fewer harms because they live a constrained, selfcontrolled lifestyle. Rather they appear to be motivated to underestimate reports of drinking in order to present themselves in a favorable light. These findings suggest that a degree of skepticism is warranted when interpreting self-reported estimates of alcohol consumption and harms. Impression management bias was evident even in the context of factors that should reduce this bias: use of an online survey with assurances of confidentiality and anonymity; recall of a single, recent drinking occasion vs. recall of average or typical drinking activity; and assessment of social desirability tendencies and drinking behavior on different days. In contrast, self-deception, or the tendency to view oneself in an overly positive way does not appear to influence responses—except perhaps drinking related consequences. Since this result was only marginally significant, it should be considered tentative. Given that impression managers tend to under-report drinking and harms, we examined the effect that such bias might have on the link between excessive drinking and harmful consequences. Although statistically controlling for impression management scores had little effect on the magnitude of correlations, the data do suggest that the Table 5 The effect of controlling for impression management scores on the correlations among AUDIT scores, number of drinks consumed, and number of harms experienced.

# drinks Harms AUDIT

# drinks

Harms

AUDIT

– .30⁎⁎⁎ .54⁎⁎⁎

.26⁎⁎ – .39⁎⁎

.52⁎⁎⁎ .30⁎⁎⁎ –

Notes: Partial correlations are presented above the diagonal, zero-order correlations are below the diagonal. n = 141. ⁎⁎ p b .01. ⁎⁎⁎ p b .001.

310

C.G. Davis et al. / Addictive Behaviors 35 (2010) 302–311

research needs to assess the extent to which this approach limits impression management bias. 4.1. Limitations and future directions

Fig. 4. Error bars represent standard errors. Number of harms is truncated at 20.

link between excessive consumption and harmful consequences is much stronger among those scoring in the normal range of impression management. Those who are impression managers tend not to report excessive drinking, and when they do, they are unlikely to acknowledge harmful consequences. 4. General discussion The motivation to put across a good impression has a significant impact on self-reports of alcohol consumption and harmful consequences. Even in the context of a confidential, online survey, people with a stronger motivation to impress under-report their consumption by 20 to 33% and under-estimate the consequences by about 50%. As a result, they are significantly less likely to be recognized on screening instruments as problematic or risky drinkers. To the extent that impression managers really do experience both the acute and long-term consequences of their drinking behavior (and there is no reason to believe that they do not), their under-reporting of consumption introduces error into models used to project the social costs of drinking, treatment needs, and any other model that relies on selfreported consumption. Although it might be assumed that some college students would be motivated to exaggerate reports of their consumption of alcohol in order to garner social acceptance from their peer group, our findings suggest that, in general, the opposite is the case: they appear to under-state their consumption. We suspect that impression managers are sensitive to the audience. That is, impression managers might exaggerate their reported consumption to their peers but tell a different story to an apparent authority (e.g., a representative of the university, their employer, or their parents). This implies that the norms perceived by impression managers are not fixed but vary by context. Given the apparent impression management bias in self-reported alcohol use and harms, attention should turn to means of limiting or adjusting scores to compensate. Although some researchers have suggested that increased accuracy might come from assessing recent (yesterday or past week) consumption rather than assessing “typical” consumption, our findings suggest that this approach will not necessarily reduce the extent of bias due to impression management. Of the various ways of assessing alcohol consumption in surveys, the graduated frequency approach (Greenfield & Kerr, 2008; Stockwell et al., 2004) may be less prone to impression management bias to the extent that it implicitly normalizes excessive and frequent drinking. Participants tend to view the middle response options as normal or neutral and use this as a frame of reference when estimating their own use (Schwarz, 1999). By providing response options reflecting extreme quantities or frequencies, and thus, shifting the normal or neutral response upwards, impression managers may be more willing to admit to greater consumption (Embree & Whitehead, 1993). Future

Although most research on impression management bias involves university students, such bias is likely not limited to this population. However, given that both our samples were made up of university students, it is important that future research establishes the extent of bias (and its effects) in other populations. To the extent that this bias is evident in other populations (including for example, those seeking or referred to treatment, those presenting at emergency departments or in primary care, and those participating in national surveys), greater research attention needs to be focused on means of reducing this bias. It also needs to be acknowledged that impression management is not the only factor contributing to inaccurate self-reports. Impression management bias does not fully account for the mismatch of sales data with self-reported consumption reported by Stockwell et al. (2004) and Knibbe and Bloomfield (2001). Nonetheless, impression management does represent one significant source of inaccuracy that leads to under-estimation of rates of consumption, risky use, and harmful consequences. Thus, we are not suggesting that people who score low on impression management are accurate in their reports, only that the accuracy of their responses does not appear to be influenced by this particular bias. Finally, it is important that future research assesses whether impression managers are, in fact, misrepresenting their consumption and harms. We have attempted to rule out alternative explanations for our findings but acknowledge the possibility that impression managers really do drink less and experience fewer harmful consequences. 5. Conclusions Skepticism with respect to self-reported consumption of alcohol and alcohol-related harms is, to some extent, warranted. Impression managers appear to significantly under-estimate their typical consumption, frequency of drinking, and daily reported consumption. They also appear to under-estimate the harmful consequences of their drinking. Although such bias may affect point estimates, prevalence rates, and the like, it does not appear to have a dramatic effect on the link between heavy drinking and acute harmful consequences. Nevertheless, future research that relies on self-reported consumption should take into account social desirability issues and attempt to reduce their influence. Role of Funding Sources This research was supported by an institutional research grant from Carleton University. Contributor The authors gratefully acknowledge the assistance of Stef Geiger who conducted Study 2 as part of his Honors thesis. Conflict of Interest The authors confirm no conflicts of interest.

References Adams, J., Parkinson, L., Sanson-Fisher, R., & Walsh, R. A. (2008). Enhancing self-report of adolescent smoking: The effects of bogus pipeline and anonymity. Addictive Behaviors, 33, 1291−1296. Adlaf, E. M., Begin, P., & Sawka, E. (Eds.). (2005). Canadian Addiction Survey (CAS): A national survey of Canadians’ use of alcohol and other drugs: Prevalence of use and related harms: Detailed report. Ottawa, ON: Canadian Centre on Substance Abuse. Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage. Babor, T. F., Higgins-Biddle, J. C., Saunders, J. B., & Monteiro, M. G. (Eds.). (2001). The Alcohol Use Disorders Identification Test: Guidelines for use in primary care, 2nd Ed. WHO document WHO/MSD/MSB/01.6a. Geneva, Switzerland: World Health Organization.

C.G. Davis et al. / Addictive Behaviors 35 (2010) 302–311 Berner, M. M., Kriston, L., Bentele, M., & Harter, M. (2007). The Alcohol Use Disorders Identification Test for detecting at-risk drinking: A systematic review and metaanalysis. Journal of Studies on Alcohol and Drugs, 68, 461−473. Block, J., & Block, J. H. (2006). Venturing a 30-year longitudinal study. American Psychologist, 61, 315−327. Booth-Kewley, S., Larson, G. E., & Miyoshi, D. K. (2007). Social desirability effects on computerized and paper-and-pencil questionnaires. Computers in Human Behavior, 23, 463−477. Carver, C. S. (2005). Impulse and constraint: Perspectives from personality psychology, convergence with theory in other areas, and potential for integration. Personality and Social Psychology Review, 9, 312−333. Cooper, L. (2002). Alcohol use and risky sexual behavior among college students and youth: Evaluating the evidence. Journal of Studies on Alcohol, 101−117 Supplement No. 14. Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology, 24, 349−354. Del Boca, F. K., & Darkes, J. (2003). The validity of self-reports of alcohol consumption: state of the science and challenges for research. Addiction, 98 (Supplment no. 2), 1−12. Doyle, S. R., Donovan, D. M., & Kivlahan, D. R. (2007). The factor structure of the Alcohol Use Disorders Identification Test (AUDIT). Journal of Studies on Alcohol and Drugs, 68, 474−479. Embree, B. G., & Whitehead, P. C. (1993). The validity and reliability of self-reported drinking behavior: Dealing with the problem of response bias. Journal of Studies on Alcohol, 54, 334−344. Fleming, M. F., Barry, K. L., & MacDonald, R. (1991). The Alcohol Use Disorders Identification Test (AUDIT) in a college sample. International Journal of the Addictions, 26, 1173−1185. Ganster, D. C., Hennessey, H. W., & Luthans, F. (1983). Social desirability response effects: Three alternative models. Academy of Management Journal, 26, 321−331. Gmel, G., & Daeppen, J. (2007). Recall bias for seven-day recall measurement of alcohol consumption among emergency department patients: Implications for casecrossover designs. Journal of Studies on Alcohol and Drugs, 68, 303−310. Gmel, G., Rehm, J., Room, R., & Greenfield, T. K. (2000). Dimensions of alcohol-related social and health consequences in survey research. Journal of Substance Abuse, 12, 113−138. Greenfield, T. K., & Kerr, W. C. (2008). Alcohol measurement methodology in epidemiology: Recent advances and opportunities. Addiction, 103, 1082−1099. Hampson, S. E., Severson, H. H., Burns, W. J., Slovic, P., & Fisher, J. K. (2001). Risk perception, personality factors and alcohol use among adolescents. Personality and Individual Differences, 30, 167−181. Kills Small, N. J., Simons, J. S., & Stricherz, M. (2007). Assessing criterion validity of the Simple Screening Instrument for Alcohol and Other Drug Abuse (SSI-AOD) in a college population. Addictive Behaviors, 32, 2425−2431. Knibbe, R. A., & Bloomfield, K. (2001). Alcohol consumption estimates in surveys in Europe: Comparability and sensitivity for gender differences. Substance Abuse, 22, 33−38. Kokotailo, P. K., Egan, J., Gangnon, R., Brown, D., Mundt, M., & Fleming, M. (2004). Validity of the Alcohol Use Disorders Identification Test in college students. Alcoholism, Clinical and Experimental Research, 28, 914−920. Lanyon, R. I., & Carle, A. C. (2007). The interna land external validity of scores on the Balanced Inventory of Desirable Responding and the Paulhus Deception Scales. Educational and Psychological Measurement, 67, 859−876. Magid, V., MacLean, M. G., & Colder, C. R. (2007). Differentiating between sensation seeking and impulsivity through their mediated relations with alcohol use and problems. Addictive Behaviors, 32, 2046−2061. Meehl, P. E., & Hathaway, S. R. (1946). The K factor as a suppressor variable in the Minnesota Multiphasic Personality Inventory. Journal of Applied Psychology, 30, 525−564.

311

Meston, C. M., Heiman, J. R., Trapnell, P. D., & Paulhus, D. L. (1998). Socially desirable responding and sexuality self-reports. The Journal of Sex Research, 35, 148−157. Olson, J. M., Goffin, R. D., & Haynes, G. A. (2007). Relative versus absolute measures of explicit attitudes: Implications for predicting diverse attitude-relevant criteria. Journal of Personality and Social Psychology, 93, 907−926. Patton, J. H., Stanford, M. S., & Barratt, E. S. (1995). Factor structure of the Barratt Impulsiveness Scale. Journal of Clinical Psychology, 51, 768−774. Paulhus, D. L. (1984). Two-component models of socially desirable responding. Journal of Personality and Social Psychology, 46, 598−609. Paulhus, D. L. (1986). Self-deception and impression management in test responses. In A. Angleitner & J. S. Wiggins (Eds.), Personality assessment via questionnaire (pp. 143−165). New York: Springer-Verlag. Paulhus, D. L. (1991). Measurement and control of response bias. In J. P. Robinson, P. R. Shaver, & L. S. Wrightsman (Eds.), Measures of personality and social psychological attitudes (pp. 17−59). New York: Academic Press. Paulhus, D. L. (2002). Socially desirable responding: The evolution of a construct. In H. I. Braun, D. N. Jackson, & D. E. Wiley (Eds.), The role of constructs in psychological and educational measurement (pp. 46−69). Mahwah, NJ: Lawrence Erlbaum Associates. Paulhus, D. L., & Reid, D. B. (1991). Enhancement and denial in socially desirable responding. Journal of Personality and Social Psychology, 60, 307−317. Paulhus, D. L., Bruce, N. M., & Trapnell, P. D. (1995). Effects of self-presentation strategies on personality profiles and their structure. Personality and Social Psychology Bulletin, 21, 100−108. Raudenbush, S., Bryk, A., & Congdon, R. (2008). HLM6.06 (Computer software) Lincolnwood, IL: Scientific Software International. Saunders, J. B., Aasland, O. G., Babor, T. F., de la Fuente, J. R., & Grant, M. (1993). Development of the Alcohol Use Disorders Identification Test (AUDIT): WHO collaborative project on early detection of persons with harmful alcohol consumption. II. Addiction, 88(791–804), 1993. Schwarz, N. (1999). Self-reports: How the questions shape the answers. American Psychologist, 54, 93−105. Shevlin, M., & Smith, G. W. (2007). The factor structure and concurrent validity of the Alcohol Use Disorder Indentification Test based on a nationally representative UK sample. Alcohol and Alcoholism, 42, 582−587. Shields, A. L., Guttmannova, K., & Caruso, J. C. (2004). An examination of the factor structure of the Alcohol Use Disorders Identification Test in two high-risk samples. Substance Use & Misuse, 39, 1161−1182. Stanford, M. S., Greve, K. W., Boudreaux, J. K., & Mathias, C. W. (1996). Impulsiveness and risk-taking behavior: Comparison of high-school and college students using the Barratt Impulsiveness Scale. Personality and Individual Differences, 21, 1073−1075. Stockwell, T., Donath, S., Cooper-Stanbury, M., Chikritzhs, T., Catalano, P., & Mateo, C. (2004). Under-reporting of alcohol consumption in household surveys: A comparison of quantity–frequency, graduated-frequency, and recent recall. Addiction, 99, 1024−1033. Stockwell, T., Zhao, J., Chikritzhs, T., & Greenfield, T. K. (2008). What did you drink yesterday? Public health relevance of a recent recall method used in the 2004 Australian National Drug Strategy Household Survey. Addiction, 103, 919−928. Taylor, B., Rehm, J., Patra, J., Popova, S., & Baliunas, D. (2007). Alcohol-attributable morbidity and resulting health care costs in Canada in 2002: Recommendations for policy and prevention. Journal of Studies on Alcohol and Drugs, 68, 36−47. Thomas, B. A., & McCambridge, J. (2008). Comparative psychometric study of a range of hazardous drinking measures online in a youth population. Drug and Alcohol Dependence, 96, 121−127. Tourangeau, R., & Yan, T. (2007). Sensitive questions in surveys. Psychological Bulletin, 133, 859−883. Welte, J. W., & Russell, M. (1993). Influence of socially desirable responding in a study of stress and substance abuse. Alcoholism, Clinical and Experimental Research, 17, 758−761. White, E. W., & Labouvie, E. W. (1989). Towards the assessment of adolescent problem drinking. Journal of Studies on Alcohol, 50, 30−37.