Beyond Projection: Specifying the Types of Peer Delinquency Misperception at the Item and Scale Levels

Abstract

The perceptual inaccuracy in peer delinquency perceptions has been attributed to projection, a hypothesis stating that individuals impute their own delinquency into perceptions of peer delinquency. While logical, projection can truthfully only explain peer delinquency misperception in one specific situation. Moving beyond projection, we propose there are actually four types of misperception. Using dyadic data, misperceptions are assessed at both the item and scale levels. At the item level, projections of delinquency are not the most frequent type of misperception. Scale-level results reveal that most people are inaccurate and tend to mix various types of misperceptions. Although projection is often linked to overestimations of peer delinquency, those who exclusively misperceive through projection mechanisms tend to underperceive peer delinquency in scale-level perceptions. A key finding shared between item- and scale-level results is perceptions of peer non-delinquency are far more accurate than perceptions of peer delinquency, thus posing serious validity concerns with perceptual measures.

ACKNOWLEDGMENTS

The authors cordially thank Marvin D. Krohn, D. Wayne Osgood, and our anonymous reviewers for helpful comments on earlier versions of this article.

Notes

1. ¹Gottfredson and Hirschi’s argument was rooted in a body of developmental psychology research predominantly focused on middle school students that had, by the early 1990s, demonstrated a high degree of similarity between one’s perceptions of peers’ tobacco, alcohol, and cocaine use and his/her own self-reported use of these same behaviors (Bauman and Fisher 1986; Bauman and Koch 1983; Fisher and Bauman 1988; Graham et al. 1991; Iannotti and Bush 1992; Jussim and Osgood 1989; Marks and Miller 1987; Urberg et al. 1990).

2. ²In most research, “actual” peer behavior is frequently measured using peer self-reports. While not without some flaws, the self-reported methodology has been found to yield reliable and valid data on delinquency (see Krohn et al. 2010).

3. ³Beyond the logical line of reasoning that rejection processes are possible, can it be expected that these processes will account for some misperception? Gottfredson and Hirschi (1990:157) make no mention of such processes, instead suggesting that indirect peer delinquency measures are either tantamount to respondent self-reported delinquency or, alternatively, hearsay with questionable validity. However, psychological literature suggests potential reasons why a respondent might report that his/her peer does not act in the same way as he/she does, even though the peer does in actuality. Individuals with lower self-esteem may seek enhancement via their associations with others (Brown 1993; Suls et al. 2002). Those with low self-esteem tend to describe themselves as having more flaws than their friends (Suls et al. 2002). And, evidence suggests that low self-esteem contributes to participation in delinquency and anti-social behavior (Donnellan et al. 2005; Rosenberg et al. 1989). Therefore, some delinquent individuals might be motivated to reject the delinquency of his or her peers. Those with high self-esteem, however, are found to rate themselves higher than their close friends on positive attributes (Suls et al. 2002). To the extent that conformity might be considered a positive attribute, this may be a reason for why individuals reject non-delinquency. It should be noted that we do not seek to test these ideas explicitly; rather, we discuss them to provide an additional rationale for why we might expect rejection processes to account for some misperception in peer delinquency.

4. ⁴The number of “purists” in any given dataset will almost certainly fluctuate depending on how many delinquency items are investigated. Intuitively, the more peer delinquency items any person is asked to perceive, the more likely that they will become inaccurate, meaning less purists will probably be found in datasets that inquire about many different peer behaviors.

5. ⁵The use of dyadic data raises several issues that require unique approaches that are not found in single-person data; one is the structure of the datafile itself. The data for the current project are structured in a “double-entry” type datafile (see Kenny et al. 2006). Unique to dyadic data analysis, this file structure places each respondent (frequently referred to as the “actor”) onto his/her own line of data with his/her friend’s (the “partner”) data attached to the end of the line of data. Thus, each dyad has two individual lines of data—one for each of the two people in the dyad—and each dyad member is treated as both the actor and the partner. Stated differently, each dyad member is treated as a target respondent (actor) who has a friend (partner) and as a partner for their friend, the actor.

6. ⁶If frequency were the focus of the investigation, there would be a much greater chance for misperception to occur. If an actor perceives the peer committed a different number of offenses than the peer self-reports (e.g., perceiving nine acts when the peer actually committed ten acts), it technically is a misperception. We comment on this further in the concluding remarks of the study.

7. ⁷Missing data was an extremely minor issue, as no perceptual or self-report delinquency item showed more than 1% missingness. The average percent missing on the perceptual items and self-reported delinquency items was 0.37% and 0.28%, respectively. Although iterative imputations are often useful, they can be detrimental with dyadic data because the iterative process can create minor differences between nested self-report delinquency measures that should be identical. Therefore, perceptual and self-reported measures were imputed with a regression based method prior to item dichotomization. However, two dyads (four individuals) were eliminated from analyses because of chronic missing data, resulting in an analytical sample size of 2,150 individuals nested in 1,075 dyads.

8. ⁸The three items that show an opposite trend all inquire about alcohol use; those who have used alcohol have been more successful in perceiving their peer’s alcohol use than non-alcohol users. It should be emphasized that this is a collegiate sample gathered from a university where the vast majority of students report using alcohol (84% report they have drank in the past year). The drinkers’ higher accuracy in perceptions of peer alcohol use could reflect the possibility that the friends in the dyads are drinking alcohol together, and co-using alcohol allows them to directly observe the other’s alcohol use, thus increasing perceptual accuracy.

9. ⁹Pointed actor perceptions are almost completely unrelated to the peer’s self-reports (r = .026, NS) because the relationship shared between perceptions and peer self-reports for the underperceiving and overperceiving pointed actors is offsetting when the entire group is considered. To help interpret this offsetting effect, this correlation was investigated separately depending on whether the pointed actor overestimated (n = 138) or underestimated (n = 283). The perceptions of both pointed overestimators and underestimators share strong positive relationships with the peer’s self-reported delinquency (r = .630, p ≤ .001 and r = .618, p ≤ .001, respectively).

10. ¹⁰Subsequent analyses were performed to determine how many purists were friends with peers who reported zero delinquent acts in the past year. Of the 139 friends who self-reported zero delinquency, 73 of them were friends of purists. Thus, over half of the purists have been completely accurate because they are consistently indicating that their friends are non-delinquent when this is in fact true. Of the remaining 66 totally non-delinquent friends, 11 are friends with rejectors, 38 are friends with projectors, and 17 are friends with pointed perceivers. All of the 677 erratic perceivers are friends with individuals who have self-reported at least one delinquent act in the past year.

11. ¹¹Using information from Table 4, the formula for calculating the number of misperceptions per item is: (1 – Median [% Correct]) * (Median [N]). Note that this formula accounts for the sample sizes of the different misperception types. We employ median values here because the N is substantially different for several items in the different scenarios (e.g., the N for items 11 and 12 in the rejection of delinquency type of misperception is very large). For comparison purposes, the misperceptions per item calculated with mean values are: Rejection of delinquency = 90.4; projection of delinquency = 74.2; projection of non-delinquency = 185.7; rejection of non-delinquency = 59.8).

Additional information

Notes on contributors

John H. Boman

JOHN H. BOMAN, IV is an Assistant Professor at the University of Wyoming in the Department of Criminal Justice. His research interests are primarily focused on the behavioral influence that peers exert on behavior, how this influence changes over the life-course, gender, theory, substance use, measurement, and construct validation. Recently, his work has appeared in Crime and Delinquency, the Journal of Criminal Justice, Criminal Justice and Behavior, the Journal of Crime and Justice, and the Journal of Youth and Adolescence.

Jeffrey T. Ward

JEFFREY T. WARD is an Assistant Professor in the Department of Criminal Justice at the University of Texas at San Antonio. His research interests include developmental and life-course criminology, sanction effects, gangs, and quantitative methodology. His work has recently appeared in Crime & Delinquency, Criminal Justice and Behavior, Criminology, and Journal of Criminal Justice.