|Home | About | Journals | Submit | Contact Us | Français|
It is important to examine social desirability when interpreting self-report data from substance abusers. Social desirability is the tendency to respond on surveys that make people appear more favorable to others; thus, a strong desire for social approval is related to minimized reports of substance use. In the present study, the relationship between social desirability and different types of social support was examined within 582 residents of communal-living recovery homes (i.e., Oxford Houses). Although effect sizes were small, results may suggest that participants reported social network variables in a socially desirable manner; this tendency towards self-deception even predicted misrepresentations of these constructs eight months later. In addition, self-reports of the substance use habits of friends and family were more prone to social desirability than the reporting of other social network characteristics. Overall, it is suggested that social desirability might be taken into account when examining substance abusers’ self-reports of social support variables.
For individuals in substance abuse recovery, contextual characteristics of the social environment such as social network characteristics may affect treatment acceptance and provide resources that influence post-treatment functioning (Finney, Moos, & Mewborn, 1980). Consequently, it is important to explore the social support networks available to people in recovery. Although social support is defined as the resources that other people provide, it is a concept that may be broken down into different dimensions (Cohen, Underwood, & Gottlieb, 2000; Cohen & Wills, 1985; Haber, Cohen, Lucas, & Baltes, 2007).
It may be important to distinguish between general social support and alcohol or drug-specific support (Longabaugh & Beattie, 1986). General social support is defined as support for the recipient’s overall well-being (Cohen et al., 2000; Cohen & Wills, 1985). Measures of general support often combine structural aspects (e.g., the number of people in a network) with functional aspects (e.g., the meaningfulness of that support) to obtain a global assessment of network social support. Specific social support, in contrast, may be directly tied to certain functions such as alcohol or illicit drug use (Longabaugh & Beattie). This specific support has either a positive or negative impact on recovery depending on whether the relationships provide encouragement for abstinence/reduced use (i.e., specific support for abstinence) or encouragement for drinking (i.e., specific support for alcohol or drug use; Falkin & Strauss, 2003).
Researchers (e.g., Rychtarik, Tarnowski, & St. Lawrence, 1989) argued that social desirability is important to examine when interpreting self-report data from substance abusers (e.g., social network information). Socially desirable responding includes the tendency to supply answers on self-report surveys that make a respondent appear more favorable to others than in reality (Bradburn, 1983). The prevailing model of social desirability contains two dimensions: impression management, the tendency to make purposeful modifications of information to impress others or to avoid negative appraisal, and self-deceptive enhancement, the tendency to believe augmented positive statements about ones’ self, even if they are not completely accurate (Paulhus, 1984; 1991; Paulhus & Reid, 1991).
A negative relationship has been found between socially desirable responding and reports of substance abuse, and this type of response bias likely results in underestimated rates of use, with heavy users actually reporting less use than light users (Cox, Swinson, Direnfeld, & Bourdeau, 1994; Richards & Pai, 2003). A strong desire for social approval, therefore, may be associated with minimizing reports of use. However, no published study has examined how social desirability affects self-reports of social network variables among substance abusers. If it is found, for example, that socially desirable responding affects reports of others’ substance use or support for use from others, then it may be important for future researchers and clinicians to control for these tendencies when having individuals report on the socially undesirable behaviors of others.
In the present study, the relationship between the two aspects of social desirability and self-reports of social support variables on Clifford and colleagues’ (Clifford & Longabaugh, 1991; Clifford, Longabaugh, & Beattie, 1992) Important People Inventory were examined within a group recovery home sample. Given that substance abusers tend to falsely report substance abuse, it was expected that recovering substance abusers living in Oxford House group recovery homes would unrealistically inflate reports of social support (i.e., higher general support, fewer drinking behaviors of friends and family, and less support for drinking from social contacts). It also was hypothesized that lower self-deception scores would predict more positive social support variables over time. If persons in recovery become more “honest” with themselves, then their support systems may improve.
Participants were 582 adults (67.7% men, 32.3% women; M age = 39.6 years old, SD = 9.3) who completed the Balanced Inventory of Desirable Responding at two data collection waves (i.e., Waves 2 and 4). This sample was drawn from a larger two-year U.S. nationwide longitudinal study (see Jason, Davis, Ferrari, & Anderson, 2007) containing 897 Oxford House group home residents at the start of the study. Founded in 1975, Oxford House (OH) provides a national network of supportive, democratic, self-run, communal-living setting for recovering substance abusers. Because residents maintain financial responsibility by paying for their own rent, food, utilities, and by sharing in house chores, OH is no more expensive than any other place of residence. Residents may stay indefinitely, provided that they pay rent, abstain from alcohol and drug use, and avoid disruptive behavior. Additionally, OH is completely devoid of professional therapists or treatment providers (Ferrari, Jason, Olson, Davis, & Alvarez, 2002; Jason, Ferrari, Dvorchak, Groessl, & Malloy, 1997).
The sample was ethnically diverse, with 57.8% European American, 34.6% African American, 3.4% Hispanic/Latino, and 4.3% others. At the start of the study, the average education level was 12.7 years (SD = 2.0). With respect to employment, 71.3% reported being employed full-time, 13.3% worked part-time, 10.9% were unemployed, and 4.5% were retired/disabled. The average monthly income of participants was $1014.7 (SD = 850.6). In the 90 days previous to entry into the study, 9.6% of participants reported using alcohol, and 7.3% reported using illicit drugs. On average, participants had undergone alcohol treatment of 2.6 times (SD = 3.6) and drug treatment 2.9 times (SD = 3.3) during their lives. The average length of abstinence from alcohol was 2.1 years (SD = 3.0) and 2.2 years (SD = 3.3) for drugs.
The majority of participants were recruited through an announcement published in a monthly newsletter circulated throughout these recovery home settings. Members of the research team then contacted the homes via letters to house presidents, conducted follow-up phone calls to the houses, and where possible, arranged to visit the houses. The remainder of participants were recruited at an annual Oxford House convention. Analyses of data collected at the convention versus data collected using the first method did not reveal significant differences in outcome or socio-demographic variables. This study was approved by an institutional review board; thus, all attempts were made to adhere to ethical standards. The nature, purpose, and goals of the study were explained to potential participants. Informed consent was given, and research team members explained that participation was entirely voluntary and withdrawal from the study was possible at any time. Furthermore, participants were assured that their responses would remain confidential. Payments of $15 were made to participants following each survey. We focused on two measurement waves (i.e., Waves 2 and 4) from the Jason et al. (2007) data set separated by an 8-month period. Correlations were run between the 11 IP indices and the two social desirability constructs (i.e., self-deception and impression management). Additionally, linear regression analyses were run to determine if Wave 2 self-deception scores predicted social support variables at Wave 4.
At both data collection waves, participants completed the Important People Inventory (IP, Clifford & Longabaugh, 1991; Clifford et al., 1992). This structured interview required participants to identify important members in their networks with whom they frequently contacted within the past 6 months. In the first section of the IP, labeled the Important People section, a participant was asked to identify up to 12 social contacts over the age of 12 years. For each person the participant listed in his/her network, the inventory examined the type of relationship, the duration of relationship in years, and the frequency of contact. In addition, the participant assessed how often the network member drank, how much the network member drank on a maximum drinking day, and the network member’s overall drinking status. In the next section, called the Most Important People (MIP) section, the participant chose up to four network members who were the most important over the past six months. The participant then rated each network member’s importance, how much he/she liked the person, and how the person reacted to the participant’s drinking. The current study utilized the 9-index, three-factor model of the IP proposed by Groh, Olson, Jason, Ferrari, and Davis (2007), namely: a) General Social Support (indices 1-3; initial wave Cronbach’s alpha = 0.33); b) Drinking Behaviors of Network Members (indices 4,7,8; initial wave Cronbach’s alpha = 0.78); and, c) Support for Drinking from Network Members (indices 9-11; initial wave Cronbach’s alpha = 0.81). This model has good internal reliability and is structurally supported by factor analyses. Individual mean scores are not interpretable due to standardization of the 11 indices before computing the factor scores.
Several researchers customized the IP to assess support specific to participants’ illicit drug use in addition to alcohol use (see Jason et al., 2006; Jason et al., 2007; Majer et al., 2002; Schmitt, 2003). Because it is possible that social support for drug use has an even greater association with self-deception because of the stigmatized nature of drug as opposed to alcohol use, we focused on support for drug use in addition to alcohol use in the present study. To assess drug use, a few additional questions were added. The questions that contributed to the alcohol-specific indices (i.e., indices 4, 7-11) were repeated, but with the words “drinking” altered to “drug use.” Likewise, the alcohol-related composite scores were accordingly changed.
In addition, participants completed Paulhus’ (1998) Balanced Inventory of Desirable Responding (BIDR). This scale contained 40 items rated along a 7-point Likert-type response scale (1= not true; 7= very true). Unlike other similar measures (e.g. the Marlowe-Crowne Social Desirability Scale [Crowne & Marlow, 1960]), the BIDR separated social desirability into two separate but related concepts each measured by 20-items: self-enhancement and impression management. The self-deceptive enhancement subscale (M = 81.10, SD = 12.91), which examined a person’s tendency to engage in statements that enhance or over exaggerates one’s abilities and skills, was administered at Wave 2 of the larger study. In contrast, the impression management subscale (M = 81.93, SD = 18.78), an assessment of one’s self-presentation style to favorably impress others, was administered at Wave 4, 8 months later. Paulhus (1988) reported Cronbach’s alphas between 0.74 and 0.86 for the two subscales, respectively, and the alphas in the present study were 0.65 and 0.78, respectively. Factor analyses of the scale conducted by the scale’s author demonstrated strong discriminant validity for both subscales across numerous other response distortion scales (see Paulhus, 1984; 1988; Paulhus & Reid, 1991). Paulhus (1994) authorized two scoring methods of the BIDR: continuous (i.e., all answers are utilized) and dichotomous (i.e., only extreme scores are utilized). Because the continuous scoring method has demonstrated stronger convergent and internal validity (see Stober, Dette, & Musch, 2002), we decided in the present study to use the continuous scoring method in all data analyses.
Regarding social support networks, of a possible 12 people, participants listed an average of 6.5 (SD = 3.5) members in their social networks. Regarding gender, 52.2% of network members were male, and 47.8% were female. The mean length of relationship was 14.3 (SD = 9.7) years, and participants on average had contact with members about once or twice a week. Almost all network members were considered to be generally supportive (98.0%). About three quarters (74.8%) of network members abstained from alcohol or were in alcohol recovery; 91.1% of network contacts abstained from illegal drugs or were in drug abuse recovery.
For all outcome analyses, the probability level was set at .01 in order to control for Type 1 error. To test the relationship between self-deception and self-reported general and alcohol-specific social support variables, correlations were run between the 11 IP indices and BIDR self-deception scores at Wave 2 (see Table 1). Results indicated that self-deception scores were significantly related to number of people in the network, r(535) = −.12, p = .007. Individuals with high self-deceptive tendencies were more likely to report having social support networks that were smaller. These self-deception analyses were run again focusing on support for drug use instead of alcohol use for the relevant indices (i.e., 4 through 11), but significant relationships were not found.
Next we correlated the 11 IP indices with scores for the second BIDR subscale, impression management, at Wave 4 (see Table 1). Results indicated that impression management scores were significantly related to percentage of heavy drinkers in the network, r(527) = −.12, p = .005, and percentage of abstainers and recovering alcoholics in the network, r(527) = .17, p = .000. In other words, individuals with high impression management tendencies were more likely to report having social support networks containing a lower percentage of heavy drinkers, and a lower percentage of abstainers and recovering drug addicts. In addition, the Drinking Behaviors of Network Members composite score had a significant relationship with impression management, r(527) = −.15, p = .001. Individuals with high impression management were more likely to report that their network contacts displayed fewer drinking behaviors. Concerning drug use, self-deception had a significant positive relationship with percentage of heavy drug users in the network, r(581) = .13, p = .003, and a significant negative relationship with percentage of abstainers and recovering drug addicts in the network, r(526) = .18, p = .000. In addition, the Drug Use Behaviors of Network Members composite score was found to have a significant negative relationship with self-deception, r(582) = −.13, p = .002.
Because the impression management subscale was only given at Wave 4, we were unable to longitudinally examine the influence of this construct. However, since the IP was administered at all waves, and the self-deception subscale of the BIDR was administered at Wave 2, we were able to test longitudinally the influence of self-deception on social support over a eight-month time period. Specifically, linear regression analyses were run to determine if Wave 2 self-deception scores predicted social support variables at Wave 4 (see Table 2). Five separate models were run: self-deception was the predictor variable in each case, and the outcome variables included the different IP composite scores. The only statistically significant regression model included self-deception predicting lower Drinking Behaviors of Network Members scores, β = −.12, t(525) = −2.68, p = .007. Thus, individuals with greater tendencies towards self-deception at Wave 2 reported having fewer friends and family members who consumed alcohol at Wave 4. As suggested earlier in this paper, it is believed that participants with these socially desirability biases misrepresented their social support networks to appear more favorable to others. However, it is important to note that the effect sizes for these significant relations were low.
Data from correlational and regression analyses suggest that participants residing in communal-living recovery homes (i.e., Oxford Houses) might have reported certain social network variables in a socially desirable manner (although effect sizes were low). Notably, most of the significant relations with social desirability variables focused on the drinking and drug use habits of important network contacts. Perhaps the reporting of friends and family members’ substance use within a communal-living recovery sample was a more sensitive issue than the reporting of general support or support for drinking. This pattern is consistent with other research emphasizing friendships with abstainers and decreasing friendships with users common to various treatments ranging from cognitive-behavioral therapy (e.g., Bandura, 1986) to 12-step and mutual-help models (Humphreys & Noke, 1997). It may be difficult for communal-living residents (many of whom also attend 12-step programs) to report that their important social contacts are using drugs and alcohol when they receive so much pressure to avoid spending time with these individuals. Because of this finding, future researchers may choose to control for these tendencies when having individuals report on the socially undesirable behaviors of their friends and family.
A surprising finding was that more significant relations with social desirability were seen for alcohol-specific social support than drug-specific social support variables. This may suggest that alcohol-related social support in this sample was more prone to social desirability than drug-related support. We originally expected that social support for drug use would have a greater association with self-deception due to the greater stigma associated with illicit drug use. However, this reversed effect might have been found because more participants used alcohol than drugs in the past three months (suggesting more recent alcohol use in this sample); thus, participants would have more reason to misrepresent alcohol than drug use.
There were several limitations in the present study. To start out with, the results were mostly correlational in nature; thus, causal relations cannot be assumed. Overall, effect sizes were low, so alternate explanations are certainly possible besides the effects of social desirability; however, the probability level was set at .01 in order to control for Type 1 error. In addition, the General Social Support composite score of the IP had low internal reliability, which may help explain the general lack of significant findings for this factor. Also, it was not possible to compare the results for impression management and self-deception subscales because these measures were administered one year apart. The inability to show that time 1 self-deception predicted reports of substance use by others at time 2 while controlling for time 2 self-deception scores provides another weakness of this study.
Furthermore, some selection bias might have occurred, and the low rates of current alcohol and drug use by participants may indicate that only the more successful or motivated Oxford House residents participated and completed the necessary measures. Perhaps, future research assessing self-deception should consider a sample with more variability with regards to substance use and stages of recovery. It has also been argued that Oxford House residents represent a self-selected sample within the substance abuse recovery population. Research by Jason, Davis, Ferrari, and Bishop (2001) refuted this claim, however, demonstrating that the demographic profile of Oxford House residents matches the typical profile of recovering alcoholics in traditional treatment settings. Even so, findings might not generalize to others in recovery for several reasons. First of all, because Oxford House residents as a rule reside with friends instead of family, perhaps social support is very different in these settings. For instance, support provided by friends is likely more essential for this sample than for others in recovery, for whom support from family is more essential. Additionally, OH residents must be sober upon entry into the program and may be further along in their recovery than people in other treatments. Consequently, it can not be said that social support and social desirability play the exact same role in recovery for those who are not yet at that stage. In addition, OH provides a highly structured abstinent environment, and perhaps social support and social desirability play a different role in the real world, which contains less structure and more opportunities for substance use and criminal behavior.
In any case, social desirability is a major concern when working with substance abuse populations, and it is important to attempt to understand these types of biases. The present study indicated that similar to self-reports of substance use, self-reports of social network variables might also susceptible to social desirable responding (particularly in structured residential recovery settings such as Oxford House). BIDR social desirability scores demonstrated relationships with several social network characteristics, in particular those focusing on the drinking behaviors of one’s friends and family. Therefore, it may be important for clinicians and researchers to assess self-deception and impression management when utilizing self-reports of social networks or other potentially socially undesirable variables within substance abuse populations.
Funding was made possible in part through National Institute on Drug Abuse (NIDA) grants #5F31DA16037 and # R01DA13231.