module

AGE DIFFERENCES IN DELINQUENCY: A TEST OF THEORY*

RANDY L. LaGRANGE University of North Carolina, Wilmington

HELENE RASKIN WHITE Rutgers, The State University of New Jersey

This study tests an integrated theoretical model of delinquency on a representative sample of 341 male New Jersey adolescents. The model is a modified version of Hirschi's contro! theory that integrates, in part, cul- tural deviance theory. This study addresses two questions: (1) how well does the theory explain delinquency within different adolescent age groups? and (2) does the salience of individual components in the model differ from one age group to another? Path analysis indicates that param- eter estimates vary substantially across age groups. The infiuence of par- ents and the school peak considerably in midadolescence while the infiuence of delinquent companions is greatest among the oldest male group. The findings indicate that the processes related to delinquency change considerably as youths age through adolescence, thus implying that the issue of "age generalizability" warrants greater attention than pres- ently given in delinquency theory and research.

The past few decades have witnessed a considerable advance in delin- quency theory and research. Initiated by Hirscbi's (1969) seminal work on control theory, a number of investigations have tested this theoretical frame- work on youth samples from many parts of the country (Hindelang, 1973; Kelly and Pink, 1973; Linden and Hackler, 1973; Megargee, 1973; Schoen- berg, 1975; Aultman, 1976, 1979; Conger, 1976; Hepburn, 1977; Rankin, 1977; Cemkovich, 1978; Eve, 1978; Regoli and Poole, 1978; Thomas and Hyman, 1978; Johnson, 1979; Little, 1979; Mannle and Lewis, 1979; Mook- het^ee and Hogan, 1979; Poole and Regoli, 1979; Wiatrowski, Griswold, and Roberts, 1981). In general, control theory has amassed substantial empirical support. Moreover, several researchers have taken Hirschi's framework an important step further by testing control theory separately on male and female samples (Jensen and Eve, 1976; Hagan, Simpson, and Gillis, 1979;

* An earlier version of this paper was presented at the American Society of Criminology Meeting, November, 1983, Denver, Colorado. The writing of this paper was supported in part by grants from the National Institute of Alcohol Abuse and Alcoholism and the National Institute of Justice (#AA-35O 9-05 and #AA-O5823-O1} and a grant from the National Institute on Drug Abuse (#DA-O3395-O1). We extend our gratitude to Allan V. Horwitz and Erich W. Labouvie for valuable comments.

CRIMINOLOGY VOLUME 23 NUMBER 1 1985 19

20 L A G R A N G E AND WHITE

Shover, Norland, James, and Thorton, 1979; Smith, 1979). These studies suggest that the etiological processes predicted by control theory cross gender lines and that separate explanations of male and female delinquency are unnecessary.

The comparison of structural parameters across clearly demarcated youth groups provides a depth of understanding not achieved when large heteroge- neous youth groups—or small homogeneous groups—are tested. However, a serious neglect in delinquency research exists regarding the applicability of theory to different adolescent age groups. That is, if we recognize first that the rate of delinquency varies considerably by age (Greenberg, 1977) and sec- ond that adolescence represents a stage of almost unrivaled physical, psycho- logical, and social change, we also need to recognize the possible limited generalizability of our theories to the entire age range of adolescence. This study takes an initial step in filling the current void by path analyzing a modi- fied version of Hirschi's control theory on three different male groups and then comparing results across samples. •

THE THEORETICAL MODEL

Control theory posits that deviance results when tbe mechanisms that should prevent it are not operating effectively. Youths with weak bonds to society (in the form of poor relationships with others and little commitment to conventional lines of activity) are to a certain extent "free" to engage in delinquent behavior. Contrary to Hirschi's original formulation, however, the state of "psychological freedom" speaks to only part of the motivational structure: only the costs of committing delinquency enter tbe decisional calculus while the potential benefits gained are left unstated. The assumption of "natural motivation" (Hirschi, 1969: 31-34, 230) provides a comfortable crutcb for control theorists to lean upon, yet it also is the source of much intellectual slack (see Cohen and Short, 1966). Therefore, the present theory attempts to avoid the conceptual shortcoming of a "pure" control theory by explicitly incorporating the influence of delinquent companions as a major source of delinquency motivation. In so doing we have wed two different— though not opposing—theories of delinquency: namely, control theory and adolescent culture theory. This approach closely resembles the structural equations model advanced by Johnson (1979). Figure 1 displays the compo- nents of tbe model and indicates the hypothesized relations among variables. The following paragraphs provide a brief discussion of each component in the model. A more complete specification and rationalization of tbe model is available elsewhere (LaGrange, 1983).

I. To prevent potential age-by-sex Interactions from influencing the results, and to increase the clarity of interpretation, females are excluded from this analysis. A more com- prehensive analysis with both male and female adolescents is presently in the works.

AGE DIFFERENCES IN DELINQUENCY 21

o Si S tiUi

X S S g , tfi lO

22 L A G R A N G E AND WHITE

SOCIOECONOMIC STATUS

Social stratification theory offers two plausible justifications for the inclu- sion of SES in a control theory of delinquency (Toby, 1974: 91). First, strati- fication theory proposes that adolescents at the lower end of the stratification hierarchy are less subject to the formal and informal networks of social con- trol that typically scrutinize the behavior of higher status youths (cf Lewis, 1965). Second, status attainment research (Blau and Duncan, 1967; Haller and Portes, 1973) suggests that socioeconomic status has an infiuence on important socialization processes. Although social status is not likely to have a strong direct effect upon delinquency, its indirect effects—primarily through parental and school influences—are presumed to be important. Socioeconomic status is the only background variable in the theory and it is treated as an exogenous variable.

PARENTAL INFLUENCES

A persistent finding in delinquency research is the infiuence of parent-child relationships on delinquency involvement (Healy and Bronner, 1936; Zucker, 1943; Glueck and Glueck, 1950; Nye, 1958; Gold, 1963; Hirschi. 1969; Jen- sen, 1972). Consistent with previous control theory formulations, the present theory presumes that family influences are a first-line defense in preventing delinquency. Figure 1 reveals that two components of family infiuences are included: (a) the degree of "parental love" toward the youth, as perceived by the youth, and (b) the strength of attachment to parents.

This separation of family infiuences into two distinct yet related controls against delinquency is guided by theory and has received empirical support (Johnson, 1979). Drawing primarily from a social learning approach (e.g., Conger, 1976), strong parental attachments presumably are most likely to develop in children whose parents provide them with ample love and positive feedback. Children who receive reinforcements in the home are likely to develop stronger ties of affection to their parents than youths who do not receive the same degree of emotional support. In other words, attachment to parents is not automatic; it is quite variable and depends to a large extent upon the degree of love and emotional support parents provide their children. Therefore, rather than conceiving of ties to parents as the first stage in the theory as Hirschi (1969) does, the concept of "parental love for child" is brought into the theory prior to "parental attachment." The double-lined arrow in Figure 1 shows that the expected influence of "parental love" on "parental attachment" is predicted to be quite strong. This claim has received considerable suppori in the literature (Bandura and Walters, 1959; Gold, 1963; Gibbons, 1976; Johnson, 1979).

AGE DIFFERENCES IN DELINQUENCY 23

SCHOOL INFLUENCES

The importance of the school in American society as a primary socializing force is well recognized (Parsons, 1959; Smelser and Halpem, 1978). More- over, the evidence linking negative school experiences with delinquency is overwhelming (Slocum and Stone, 1963; Hargreaves, 1967; Hirschi, 1969; Polk, 1971; Frease, 1973; Hindelang, 1973; Kelly and Pink, 1973; Elliott and Voss, 1974; Gibbons, 1976; Thomas, Kreps, and Cage, 1977; Johnson, 1979). In the present delinquency model the influences of the school have been bro- ken down into three separate components: school performance, school attach- ment, and commitment to education. According to the model and prior theory (Toby, 1957), youths' perceptions of the school and chances for aca- demic success are shaped in large part by their parents. Youths who do poorly in school (school performance) are not likely to develop strong feelings of attachment to school (school attachment), nor are they likely to commit themselves to educational pursuits (educational commitment). This negative process increases the probability that these youths will enter into delinquent associations which, in tum, increase the probability of delinquent behavior (refer to Figure 1).^

DELINQUENT ASSOCIATES

The structural equations model developed to this point is predominantly a control theory. Youths who lack adequate attachments to parents and schools and who do not have strong commitments to the conventional system are, in a sense, "free" to engage in delinquency. Provided that opportunities avail themselves (i.e., weak social control), the only defense against delin- quency at this point is the perceived risk of apprehension and punishment. Yet an essential link in the theory is missing: given Xht freedom to deviate, what invokes the desire to deviate? This conceptual link in the model is partly filled by the infiuence of delinquent associates. The model predicts a strong relationship between delinquent associates and delinquent behavior (accounting for the double-lined arrow between these two variables in Figure 1). Paths leading to the dependent variable that do not pass through the

2. To this point we have remained silent on the subject of causal order. We recog- nize that the recursive model posited here requires several rather bold theoretical assump- tions, all of which cannot be completely justified. However, following the logic of Hirschi's analysis and (we further argue) consistent with the processes involved in much primary deviation, delinquency Is believed to be the final element in the causal flow. Thus, as Hir- schi (1969: 132) was able to demonstrate:

The causal chain runs from academic incompetence to poor school pwifonnance to disliking of school to rejection of the school's authority to the commission of delinquent acts. All statistical relations relevant to this causal chain have been presented, and all are in fact consistent with it.

24 L A G R A N G E AND WHITE

delinquent associates variable refiect the assumption that control theory vari- ables have an efi"ect upon delinquent behavior independent of delinquent com- panions.

METHODS AND MEASURES This investigation analyzes data from a much larger study, the Rutgers

Health and Human Development Project. The present data are comprised of 341 adolescent males selected from the general population of New Jersey youth. The sample is stratified on the basis of age with the following break- down: 122 12-year-olds, 138 15-year-olds, and 81 18-year-olds.3 These age groups roughly represent the developmental benchmarks of beginning, mid- dle, and late adolescence. We should note that this sample has the desirable quality of spanning the entire age range commonly called "adolescence." The three-year interval that separates each successive male cohort maximizes the age variation in the data.

The sampling procedures involved several steps. Potential participants were initially identified through a series of randomly generated telephone calls. If a household contained a youth falling within an eligible age group and agreed to participate a more extensive telephone interview was com- pleted, followed by a home visit. At this time several questionnaires were completed by the subjects and their parents. Following this contact, subjects came on-site for a full day of testing including physical examinations, physio- logical and perceptual-behavior tests, psychological inventories, and the com- pletion of several questionnaires. Initial analysis of the data indicates that refusers are similar to study participants in terms of race, religion, and alco- hol consumption. However, refusers tend to be slightly overrepresented by the lower social classes.**

DELINQUENT BEHAVIOR

A general delinquency index is created from the responses to 15 self-report items (see Appendix). These items are designed to tap a wide range of behav- iors that are either inappropriate or illegal for youth to commit (e.g., breaking and entering, assault, theft, vandalism, truancy, avoiding payment). The

3. A large number of items used to measure school-related influences only applied to youth who were currently enrolled in school (e.g., grade school, high school, vocational school, or college). This requirement excludes from some measures a sizable proportion of the 18-year-olds (37%) who have already left the formal school system. A series of t-tests indicated that these out-of-school youth did not differ significantly from in-school youth on a wide range of measures. There were two exceptions: out-of-school youth reported signifi- cantly/etver simple assaults and significantly less parental supervision than in-school youth. Important race or social status differences were not identified.

4. A more complete discussion of the methodological strategy can be found in Les- ter, Pandina, White, and Labouvie (in press).

AGE DIFFERENCES IN DELINQUENCY 25

delinquency index has a possible range from 15 to 75 with an alpha reliability (Cronbach) of .76. The data reveal that the proportion of males reporting at least one act of delinquency is considerable. Seventy-two percent of the 12- year-olds have nonzero scores on the delinquency index while 77% of the 15- year-olds and 94% of the 18-year-olds also have nonzero scores.

PARENTAL INFLUENCES

The parental love measure is an additive index created from 16 items reflecting the amount and quality of love parents bestow upon their children (alpha reliability of .89). Parental attachment is an additive index of five items asking youth how strongly they currently love and respect their parents (alpha reliability of .78). Each index measures the maximum level of parental love or parental attachment. This is similar to the measurement procedure employed by Johnson (1979: 80-81).5 Moreover, the "parental love" index (as well as all components in the model) is measured from the viewpoint of the child, not the parent.*

SCHOOL INFLUENCES

A number of items are utilized to measure the multiple and complex influ- ences of the school on adolescent behavior. First, three items measuring school grades are summed into a school performance index. Second, school attachment is measured through five questions concerning the amount of sat- isfaction and enjoyment a youth derives from school. Finally, educational commitment is measured from six items reflecting the amount of work and emotional energy a youth presently expends on school."^ Reliability estimates

5. Johnson provides a convincing justification against the simple addition of mother and father scores, especially in the case of a single parent household. Avoiding an additive model of parental influence not only makes theoretical sense but allows the inclusion of single-parent families in the analysis.

6. A preliminary test of the theory was performed to determine whether mother- child relationships play a substantially different role than father-child relationships. Meas- ures of "mother's love toward child" and "attachment to mother" were employed in the first regression along with all other variables in the model. Comparable measures of "father's love toward child" and "attachment to father" were employed in the second regression. The two path analytic solutions generated largely replicated one another (not shown). Despite several minor differences, the magnitude of regression coefficients remained relatively stable across models. Moreover, the proportion of variance explained by each model was virtually identical (R^ = .35). Similar analyses performed within each age group also suggested that little additional information is gained when maternal and paternal infiuences are individually assessed. Therefore, the separate effects of mother's and father's love toward the child have been combined into one "parental tove" index. Similarly, the two measures of parental attachment have been combined into a single "parental attachment" index.

7. In the delinquency literature the concept of "commitment" is defined in broad terms and includes one's "stake" in a number of legitimate sectors of society. For our

26 L A G R A N G E AND WHITE

for these indices are relatively high, ranging from .75 to .82.^

DELINQUENT ASSOCIATES

The delinquent associates measure is a composite index of nine items that attempt to identify the proportion of friends that engage in various forms of delinquency. This index ranges from 9 to 54 with a reliability estimate of .88.

SOCIOECONOMIC STATUS

The measure of social status is a composite of two indicators: (1) the high- est level of parental education attained and (2) family income.

ANALYTIC DESIGN

The theoretical model is tested through path analysis. The path analytic coefficients are interpreted as effect eoeflicients. That is, the coefficients measure the expected change in the dependent variable when the independent variable is actually changed by one unit (Nie, Hull, Jenkins, Steinbrenner, and Bent, 1975: 384).^ When comparing parameter estimates across age sub- samples unstandardized regression coefficients are reported. Standardized coefficients are reported when testing the theory on the entire sample or within a subsample.'<• Missing data are deleted through a "listwise" procedure. • '

purposes, this concept is operationalized in terms of present educational commitment only. While this obviously narrows the conceptual domain of the commitment variable—possibly robbing it of some explanatory jwwer—it does increase the clarity of the concept and inter- pretation of results.

8. Since there are two "conceptually" related measures of parental influence and three "conceptually" related measures of school infiuence, the question of multicoUinearity must be addressed (see Lewis-Beck. 1980, for a discussion of multicoUinearity problems). In the present sample, the correlations among these several indices are of moderate strength. For example, the correlation between parental hve and parental attachment is .51. Similarly, the correlations among the three school-related measures range from .35 to .49. In other words, multicoUinearity is not serious and is unlikely to cause analytic problems.

9. Without the assumptions of weak causal order and causal closure, the regression coefiicients can only be interpreted as the difference in Y between two groups that are different on X by one unit. The idea of actually "manipulating" the independent variable requires the two assumptions above.

10. Standardized coefficients are most appropriate when one is interested in estimat- ing the relative effects of different independent variables on Y in the same population, or when the independent variables are measured in different units and one is interested in assessing the explanatory importance of one variable compared to another. When the interest is in comparing the parameters from one sample to another (or from one subsample to another), unstandardized coefficients are preferable (see Blalock, 1967).

H. This means that missing values on any variable entered into the equation cause that case to be eliminated from the entire run, thereby ensuring that all coefficients are

AGE DIFFERENCES IN DELINQUENCY 27

RESULTS MODEL TESTED ON ENTIRE SAMPLE

Table 1 presents the standardized and unstandardized path coefficients resulting from the test of the delinquency model on the entire adolescent sam- ple. These findings are reported as a point of reference for subsample com- parisons and to detennine whether the theory is performing as hypothesized. Briefly summarizing the information contained in the table, we note that the pattem of coefficients is largely consistent with expectations. This holds true with respect to the direction of individual path coefficients and the magnitude of their effects. Furthermore, the amount of variance in delinquency explained by the model compares favorably with other self-report research (R2 = .36).

The decomposition of effects shown in Table 2 provides a clearer under- standing of the path coefficients. ^ 2 The total effects column (which reflects all direct plus indirect effects) indicates that delinquent associates have the larg- est relative influence on delinquency (.410, p < .001), followed in order by school performance ( — .281, p < .001), educational commitment ( — .277, p < .001), parental attachment ( - . 2 1 2 , p < .001), parental love ( - . 1 8 5 , p < .001), and SES (—.131, p < .05). Attachment to school is the oniy compo- nent of the model that fails to have a significan't total effect.

Table 2 also indicates how the variables are contributing to the model. As noted above, parental love has a moderately strong overall eflFect on delin- quency but does not influence delinquency directly. That is, the efFect of parental love is almost entirely mediated through other components of the model, most notably through parental attachment and school performance

based upon the same sample. "Pairwise" deletion only drops a case from analysis when that variable is involved, resulting in different coefficients being estimated from different cases. Though pariwise deletion "saves" a number of cases which listwise deletion elimi- nates, little confidence can be placed in the statistics it produces (Nie, et al., 1975: 353).

12. This table is "decomposed" in the manner described by Alwin and Hauser (1975), and provides useful infonnation for intetpreting pattems of coefficients. The total associa- tion column represents the zero-order correlation between the delinqtiency tneasure and each of the independent variables. The total effect refers to the amount of change in dehn- quent behavior induced by a change in a given independent variable, ignoring the interven- ing mechanisms through which the effects flow. Stated another way, the total efFect of one variable on another is that part of the total association that is "causal." The direct effect represents that part of the total effect that is not transmitted through any other variable in the model. (The indirect effect of any variable can be easily obtained by subtracting the "direct" effect from the "total" effect.)

LAGRANGE AND WHITE

3

o V (U

rt

rt H

D O

I 2i g-.S

I 8 Q <

3 B

35

c St-

=

O- 8 0 00 (N - ^ cN

o o O O(N

0 0 - * — r*i OO OO

O r ^ a ^ O - ^ — (N

I I I I I

O O ^ fO 00 O « CN

(N Q in cn

O CN O r^ fo oo <N ( N CN

Q * £ !

•a N

•3

TJ C

M

D

rt

lO O

l

"rt + j

c V l - l rt

3 c u

CU

ne e

1 n <u B

ta e

<

> I-J

s

0 0 0

ro (N — O —

r- rO

o

fN

r- O

UJ cn

ar en

i

Qa

.2 0

t«

ne e

rt

a! Q

cn

c CJ

d 0

lu c

e s5

_̂,c 3 O *

.E Q

V

u O

.a o- •t3 ;

p V

AGE DIFFERENCES IN DELINQUENCY 29

(from Table 1).'^ At the other extreme, all of the eflfeet of delinquent assoei- ates on delinquent behavior is direet sinee no other variable intervenes in the model.

Table 2. Decomposition of Effects of Theory Tested on All Males

Independent Variable

SBS

Parental

Love

Parental

Attachment

Sehool

Performanee

Sehool

Attachment

Educational

Commitment

Delinquent

Assoeiates

Total Association

- . 1 3 1 *

- . 1 9 5 * * *

- . 2 6 2 * * *

- . 3 3 8 * * *

- . 1 9 3 * * *

- . 4 0 4 * * *

.507***

Total EflFeet

- . 1 3 1

(-.206)*

-.185

(-.060)***

-.212 (-.445)***

-.281

(-.288)***

-.059

(-.058)

-.277 (-.280)***

.410

(.268)***

Direet Effeet

-.004

(-.006)

.006

(.002)

-.128 (-.269)**

-.138 (-.141)**

.048 (.048)

-.198 (-.200)***

.410

(.268)***

Unstandardized effects are in parentheses. N = 341

R2 = .36 •p<.05 ••p<.01 •••p<.001

Figure 2 displays the path diagram when unimportant paths from the ini- tial run (i.e., standardized path coeffieients less than .10) are deleted and the

13. Note that the direct effect of the parental love variable is essentially zero (i.e., .006 "standardized" efFect). Therefore, the fact that the coefficient is in a positive direction is of no great interest. However, a similar finding of slightly greater magnitude exists with the school attachment measure indicating that a suppressor effect is operating. Though we do not pursue this probletn further, it does testify to the complexity of relationships among variables iti the model.

30 LAGRANGE AND WHITE

AGE DIFFERENCES IN DELINQUENCY 31

model retested.i'* Error terms are omitted for graphic clarity. The path dia- gram is used principally for heuristic purposes since the information it con- veys is largely redundant (see Tables 1 and 2). However, of particular interest is the R̂ when this more parsimonious model is compared to the saturated model reported in Table 1. With the exception of some rounding differences, the proportion of variance explained remains virtually unchanged.

The findings summarized to this point are, in general, consistent with the- ory. While the model has slightly underestimated the role of school perform- ance and overestimated the effects of school attachment, these are minor departures and easily incorporated into the theoretical framework. These results are comparable to Johnson's (1979) findings and provide a partial rep- lication of his work.

Table 3. Decomposition of Effects of Theory Tested on Twelve-Year-Old Males

Independent Variable

SES

Parental Love

Parental Attachment

School Performance

School Attachment

Educational Commitment

Delinquent Associates

TotaJ Association

-.015

-.081

-.186*

-.119

- . 1 9 1 *

-.215*

.455***

Total Effect

-.015 (-.016) -.079

(-.017) -.182

(-.339) -.129

(-.102) -.150

(-.099) -.123

(-.091) .457

(.253)***

Direct Effect

-.124 (•133)

-.009 (-.002) -.086

(-.160) -.091

(-•072) -.006

(-.004) -.172

(-•127) .457

(.253)*** Unstandardized effects are in parentheses. N R2 = .27 •p<.05 •*p<.01 •••p<.001

= 122

14. While this cut-off point for deleting path coefficients is an arbitrary figure (and should not be confused with the .10 probability level), it is sufficiently restrictive to ensure that the theoretical model does not capitalize on chance variation among variables. In fact, Land (1969) maintains that even in the absence of an adequate interpretative framework, sundardized path coefficients of only .05 or greater may be retained. As will be evident later, however, a number of paths attain the predetermined level of substantive importance (i.e., pxy < .10) yet, because of the reduced number of cases available in the several sub- samples, statistical significance (p < .05) may be just barely achieved or just barely missed.

32 . L A G R A N G E A N D W H I T E

MODEL TESTED ACROSS AGE GROUPS

The theory is tested separately on each of the three age groups; Table 3 reports the decomposition of path coefficients for the 12-year-olds, Table 4 for the 15-year-olds, and Table 5 for the 18-year-olds. R-squares and significance

Table 4. Decomposition of Effects of Theory Tested on Fifteen-Year-Old Males

Independent Variable

SES

Parental Love

Parental Attachment

School Performance

School Attachment

Educational Commitment

Delinquent Associates

Total Association

-.118

-.251**

-.310***

-.403***

-.376***

-.504***

.418***

Total Effect

-.118 (-.178) -.244

(-.082)** -.257

(-.492)** -.334

(-.324)*** -.228

(-.240)** -.332

(-.348)*** .238

(.147)**

Direct Effect

-.057 (.086)

-.045 (-.015) -.185

(-.354)* -.149

(-.144) -.027

(-.028) -.278

(-.291)** .238

(.147)**

Unstandardized eflFects are in parentheses. N = 138

R2=.36 •p<.05 •*p<.01 •••p<.001

levels are also reported. Judging from the proportion of variance explained it is apparent that the theory is not performing the same within each adolescent group. The R̂ is lowest in the youngest male group (R^ = .27), reaches its highest level among the 15-year-olds (R^ = .36), and drops slightly in the oldest group (R^ = .32). Overall, however, the size of each of these R- squares is of moderate strength, thus suggesting that the theory—or part of the theory—is predictive of delinquency.

When we examine more closely the influence of specific variables within each age group, considerable variation is found. Among the 12-year-olds, for example (Table 3), delinquent associates is the only component of the model exhibiting a significant "total effect"; the conventional controls from parents and school have little predictive value. The pattern of coefficients is similar

AGE DIFFERENCES IN DELINQUENCY 33

Table 5. Decomposition of Effects of Theory Tested on Eighteen-Year-Old Males

Independent Variable

SES

Parental Love

Parental Attachment

School Performance

School Attachment

Educational Commitment

Delinquent Associates

Total Association

- . 2 5 2 *

- . 0 8 2

-.178

-.222*

-.145

-.210*

.528***

Total Effect

-.252 (-.483)* -.010

(-.039) -.105

(-.240) -.137

(-.170) -.023

(-.026) -.108

(-.130) .485

(.368)***

Direct Effect

-.083 (-.160) -.003

(-.001) -.099

(-.226) -.095

(-.118) -.014

(-.016) -.049

(-•059) .485

(.368)***

Unstandardized effects in parentheses. N = 81

R2 = .32 •p<.05 ••p<.01 ***p<.OOI

among the oldest male group (Table 5) where delinquent associates and SES are the only variables with a significant "total effect" upon delinquency. These findings seem to support an adolescent culture explanation of delin- quency while failing to support control theory.

Interestingly, a much different pattern of coefficients emerges in the 15- year-old-group. As shown in Table 4, each of the five control theory vari- ables, as well as the delinquent associates indicator, has significant and sub- stantial "total effects" upon delinquency; SES is the only component of the model that does not achieve significance. The relative total influence of delin- quent associates (.238, p < .01) is surpassed by four control variables: school performance ( — .334, p < .001), educational commitment (—.332, p < .001), parental attachment (—.257, p < .01), and parental love ( - . 2 4 4 , p < .01). The relative impact of the fifth control variable (school attachment) falls just below the total impact of the delinquent associates measure ( — .228, p < .01).

In short, the path analytic findings generated among the 15-year-old males

34 L A G R A N G E AND WHITE

appear to support both a control and adolescent culture explanation of delin- quency. This is considerably different from the findings that emerge in the youngest and oldest adolescent groups. Table 6 provides additional support for this argument. Among the 15-year-olds, the combined effects of parental and school influences explain over 30% of the variance in delinquency (10% by parental influences, 20% by school influences). Adding delinquent associ- ates to the model increases the explained variance by only another 4 % . However, the combined effects of parental and school influences explain only around 8% of the variance among the 12-year-olds and approximately 4 % among the 18-year-olds. Adding delinquent associates to the model increases the R2 by .19 and .22, respectively, in the youngest and oldest samples.

Table 6. Summary Table of R̂ Change By Age

Independent Variable

SES

Parental Influences Parental Love Parental Attachment

School Influences School Performance School Attachment Educational Commitment

Delinquent Associates

Age 12

•001

•034 .006 •028

.044

.014

.019

.011

• 193

.27

R2 Change

Age 15

.014

JQQ

xm

.103 X)36

.36

Age 18

.063

.020

.010

.010

mi .015 .001 .007

.32

Variables entered in theoretical order.

Finally, an alternative method of testing the importance of specific vari- ables across the three age groups is to compare the unstandardized total effects. These coefficients are reported in Table 7 with their standard errors. To identify the most important differences and to minimize the influence of random fluctuations in the data, we adopt an arbitrary criterion requiring that coefficients lie at least 1.0 standard errors away from each other.'5 For each between-group comparison we use the larger of the two standard errors which, in each case, turns out to be from the "older" group. Consistent with

15. See Portes and Wilson (1976) for a similar test of the difference among regression coefficients.

AGE DIFFERENCES IN DELINQUENCY 35

Table 7. Summary of Total Unstandardized Effects of Variables By Age

Independent Variable

SES

Parental Love

Parental Attachment

School Performance

School Attachment

Educational Commitment

Delinquent Associates

R2

Age 12

- . 0 1 6 (.098)

- . 0 1 7 (.020)

- . 3 3 9 (.183)

- . 1 0 2 (.076)

- . 0 9 9 (.064)

- . 0 9 1 (.077)

.253*** (.046)

.27

Age 15

- . 1 7 8 (.129)

- . 0 8 2 * * (.028)

- . 4 9 2 * * (.194)

- . 3 2 4 * * * (.078)

- . 2 4 0 * * (.095)

- . 3 4 8 * * * (.098)

.147** (.049)

.36

Age 18

- . 4 8 3 * (.209)

- . 0 3 9 (.043)

- . 2 4 0 (.272)

- . 1 7 0 (.145)

- . 0 2 6 (.144)

- . 1 3 0 (.174)

.368*** (.076)

.32

Standard errors are in parentheses.

•p<.05 *'p<.OI ***p<.001

the findings reported above, parental and school influences tend to peak con- siderably in the middle adolescent group. For example, among the 15-year- olds the coefficients for parental love, school performance, school attachment, and educational commitment all exceed the comparable coefficients from the 12- and 18-year-old groups by the arbitrary criterion. By contrast, the pre- dictive capacity of delinquent associates is substantially greater in the young- est and oldest groups than in the middle adolescent group. Note further that the influence of socioeconomic status increases steadily with increasing age, thus failing to support the common view (e.g., Reiss and Rhodes, 1961; Clark and Wenninnger, 1962; Johnstone, 1978; Glynn, 1980) that family social sta- tus is outgrown during late adolescence.'*

16. We must caution, however, that sample sizes are not large for this type of analysis (especially in the 18-year-old group), meaning that coefficients are prone to capitalize on chance fluctuations in the data. For this reason, our analysis has focused upon the more "general" trends found in the data rather than concentrating upon "specific" findings or anomalies. Clearly a need exists for additional research.

36 ^ L A G R A N G E AND WHITE

DISCUSSION

The results of this study suggest that the causal processes informed by Hir- schi's control theory are most applicable among youths in the middle of ado- lescence. Perhaps one explanation of this finding is that the middle teenage years represent a critical period. This may be especially evident in Western societies hke the U.S. where midadolescent youths occupy a rather dubious status; they are generally less than fully oriented to their family of origin and at the same time are rigorously denied access to full participation in society. To state this status dilemma alternatively, it is too late for many of these youths to turn back and too early for them to proceed. From this view, then, attachments to proper role models, such as parents and teachers, and com- mitments to conventional lines of activity are likely to be most important for youths in middle adolescence because they function as "psychological anchors" to conformity that otherwise may be missing. Moreover, this devel- opmental pattern closely follows the behavioral trends observed in self-report studies (e.g., Wilson, 1972; Elliott, Knowles, and Canter, 1981) and official data alike (Greenberg, 1977) where most forms of delinquency (excluding serious violent offenses) increase until age 15 or 16 and then gradually decline as youths age out of adolescence.

If our findings are generalizable, their implications for dehnquency theory and research are significant. Given the static conception of adolescent mis- conduct that dominates the delinquency literature, these results imply that a dynamic age framework may prove more valuable. That is, if adolescence is a stage in hfe of tremendous change, growth, and development, it is hkely that the most salient etiological processes of delinquency may also change as youths age through adolescence. Yet most theories have a tendency to view the adolescent years as a "unitary" period in life; explanations of why some youths engage in delinquency while others do not presuppose that the age of youths is unimportant.'^ For example, in two of the most recent and impor- tant applications of control theory (Johnson, 1979; Wiatrowski et al., 1981), the youth samples are restricted to adolescents in their middle teen years (i.e., approximately 15 years old). Wiatrowski and his associates apparently assume their findings are equally applicable to adolescents of all ages since no cautionary remarks concerning the generalizability of their results are stated. Johnson (1979: 72) recognizes that his findings may be, in part, "age depen- dent," but casts the problem of age generalizability to future research. Also

17. Although the concept of an "adolescent career" has escaped largely unnoticed by delinquency theorists and researchers alike, the occupational analogy of juvenile "criminal careers" has not (e.g., Robbins and Wise, 1977). This latter term offers a useful conception of career-line movement among the very delinquent, but it loses much of its relevance when applied to a general population of youth. For this reason the term "criminal career" has been avoided throughout the discussion.

AGE DIFFERENCES IN DELINQUENCY 37

worth mentioning, the Richmond Youth Project contained data on youths from grades seven through twelve, yet Hirschi (1969) chose not to address the problem of age effects in his main analysis. This decision is all the more curious in light of his decree that "there is little doubt that middle adoles- cence is the period of maximum dehnquent activity, a fact with theoretical significance that is out of proportion to the magnitude of the empirical rela- tion" (1969: 236; emphasis added).

At the very least, then, the present investigation suggests that the age-delin- quency relationship has been a neglected area of research far too long and remains a serious weakness in our delinquency theorizing. Greenberg's (1977) insightful essay offers one of the few systematic developments in this direction. In this article, Greenberg attempts to tie age patterns of delin- quency into the larger historical and structural patterns of society. This work is all the more valuable in relation to the present study because the social/ psychological framework utilized here has few structural moorings.

The need for appropriate data is apparent. Long-term investigations that trace the development of youths through their adolescent and young adult years are most useful though difficult and expensive to perform. Multi-cohort cross-sectional data (such as the data utilized in the present study) also yield relevant information concerning the effects of age. However, other poten- tially confounding influences, e.g. "generational" effects, are left free to vary. While single cohort cross-sectional designs are frequently used in delinquency research and do provide valuable information, one should realize that the relationship between age and dehnquency is actually "removed" from the data. Users of this type of data therefore should more carefully clarify the potential age range limitations of their findings.

There are several limiting aspects of the present study that temper the strength of our conclusions. First, the problem of restricted sample size— especially in the 18-year-old group—has increased the possibility that chance fluctuations in the data have influenced the results. Second, Hirschi's concept of "belief' was not included in the model because appropriate indicators were not available. This is no small oversight. The predictive capacity of this vari- able is repeatedly borne out in the literature (Hirschi, 1969; Hindelang, 1973; Jensen and Eve, 1976; Johnson, 1979; Wiatrowski et al., 1981), and is found to play an important mediating role between the influence of delinquent friends and delinquent behavior (Matsueda, 1982). Further research is advised to give greater attention to this concept.

A third potential limitation with this analysis concerns the problem of causal order. Given the cross-sectional nature of our data we are unable to firmly establish which variables precede others in the causal scheme. Our justification for causal order relies primarily upon prior research, presumed time sequences of events, and common sense. Similarly, we are unable to

38 L A G R A N G E AND WHITE

specify how the infiuence variables change as youths move through adoles- cence. We can only determine the salience of various mechanisms as they impinge upon different aged youths at a single point in time (even though our interpretive framework assumes that "aging," not "cohort change," is the driving force). In other words, since the theoretical orientation applied here is fundamentally dynamic and assumes significant adolescent change over time, moving from a cross-sectional to a longitudinal design would provide a more powerful test of the theory. Of course in the absence of an appropriate longitudinal design, specifying potential developmental patterns is a specula- tive and risky endeavor. These findings should prove valuable if only for the purpose of theoretical clarification, or as a springboard for future delinquency research.

REFERENCES

Alwin, Duane F. and Robert M. Hauser 1975 The decomposition of effects in path analysis. American Sociological Review

40: 37-47. • .

Auhman, Madeline G. 1976 An empirical examination of some major delinquency correlates. Unpublished

doctoral dissertation. Tallahassee: Florida State University.

1979 Delinquency causation: A typological comparison of path models. Journal of Criminal Law and Criminology 70: 152-163.

Bandura, Albert and Richard H. Walters 1959 Adolescent Aggression. New York: Ronald Press.

Blalock, Hubert M., Jr. 1967 Path coefficients vs. regression coefficients. American Journal of Sociology 72:

675-676.

Blau, Peter M. and Otis D, Duncan 1967 The American Occupational Structure. New York: Wiley.

Cemkovich, Stepheti A. 1978 Evaluating two models of delinquency causation. Criminology 16: 335-352.

Clark, John P. and Eugene P. Wenninger 1962 Socio-economic class and area as correlates of illegal behavio'r among juveniles.

American Sociological Review 27: 826-834.

Cohen, Albert K. and James F. Short, Jr. 1966 Juvenile delinquency. In Robert K. Merton and Robert A. Nisbet (eds.). Con-

temporary Social Problems. 2nd ed. New York: Harcourt, Brace, and World.

Conger, Rand D, 1976 Social control and social learning models of delinquent behavior. Criminology

14: 17-39.

Elliott, Deibert S. and Harwin L. Voss . 1974 Delinquency and Dropout. Lexington, MA: Lexington.

AGE DIFFERENCES IN DELINQUENCY 39

Elliott, Delbert S., Brian A. Knowles, and Rachelle J. Canter 1981 The Epidemiology of Delinquent Behavior and Drug Use Among American

Adolescents. Vol. 1. The National Youth Survey Project Report No. 14. Boul- der: Behavioral Research Institute.

Eve, Raymond A. 1978 A study of the efficacy and interactions of several theories for explaining rebel-

liousness among high school students. Journal of Criminal Law and Criminol- ogy 69: 115-125.

Frease, Dean E. 1973 Delinquency, social class, and the schools. Sociology and Social Research 57:

443-459.

Gibbons, Don C. 1976 Delinquent Behavior. 2nd ed. Englewood Cliffs, NJ: Prentice-Hall.

Glueck, Sheldon and Eleanor Glueck 1950 Unraveling Juvenile Delinquency. Cambridge: Harvard University Press.

Glynn, Thomas J. 1980 From family to peer: Transitions of influence among drug using youth. Paper

presented at the Research Analysis and Utilization Meeting, National Institute on Drug Abuse, Maryland, September.

Gold, Martin 1963 Status Forces in Delinquent Boys. Institute for Social Research. Ann Arbor:

University of Michigan.

Greenberg, David F. 1977 Delinquency and the age structure of society. Contemporary Crises 1: 189-223.

Hagan, John, John H. Simpson, and A.R. Gillis 1979 The sexual stratification of social control: A gender-based perspective on crime

and delinquency. British Journal of Sociology 30: 25-38.

Hargreaves, David 1967 Social Relations in a Secondary School. New York: Humanity Press.

Healy, William and Augusta Bronner 1936 New Light on Delinquency and Its Treatment. New Haven: Yale University

Press.

Hepburn, John R. 1977 Testing alternative models of delinquency causation. Journal of Criminal Law

and Criminology 67: 450-460.

Hindelang, Michael J. 1973 Causes of delinquency: A partial replication and extension. Social Problems 20:

471-487.

Hirschi, Travis 1969 Causes of Delinquency. Berkeley: University of California Press.

Jensen, Gary F. 1972 Parents, peers and delinquent action: A test of differential association theory.

American Journal of Sociology 78: 562-575.

Jensen, Gary F. and Raymond Eve 1976 Sex differences in delinquency. Criminology 13: 427-448.

40 L A G R A N G E AND WHITE

Johnson, Richard E. 1979 Juvenile Delinquency and Its Origins: An Integrated Theoretical Approach.

Cambridge: Cambridge University Press.

Johnstone, John W.C. 1978 Social class, social areas and delinquency. Sociology and Social Research 63:

49-72,

KeUy, Delos H. and William T. Pink 1973 School commitment, youth rebellion, and delinquency. Criminology U: 473-

485.

LaGrange, Randy L. 1983 Juvenile Offenders: An integrated explanation of age and gender differentials in

delinquency. Unpublished doctoral dissertation. New Brunswick, N.J.: Rutgers University.

Land, Kenneth C. 1969 Principles of path analysis. In Edgar F. Borgatta (ed.), Sociological Methodol-

ogy. San Francisco: Jossey-Bass.

Lewis, Hylan 1965 Child rearing among low-income families. In Louis A. Ferman, Joyce L.

Kombluh, and Alan Haher (eds.), Poverty in America: A Book of Readings. Ann Arbor: University of Michigan Press.

Lewis-Beck, Michael S. 1980 Applied Regression: An Introduction. Beverly Hills: Sage.

Lester, David, Robert J. Pandina, Helene R. White, and Erich W. Labouvie In Press The Rutgers health and human development project: A longitudinal study of

alcohol and drug use. In Sarnoff A. Mednick and Michel Harway (eds.), U.S. Longitudinal Studies. New York: Praeger.

Linden, Eric and James C. Hackler 1973 Affective ties and delinquency. Pacific Sociological Review 16: 27-46.

Little, C.B. 1979 Delinquent behavior and trouble with the police: An assessment of some com-

peting explanations. Presented at the 28th Annual Meeting of the Society for the Study of Social Problems.

Mannle, Henry W. and Peter W. Lewis 1979 Control theory reexamined: Race and the use of neutralizations among institu-

tionalized dehnquents. Criminology 17: 58-74.

Matsueda, Ross L. 1982 Testing control theory and differential association: A causal modeling

approach. American Sociological Review 47: 489-504.

Megargee, Edwin I. 1973 Recent research on overcontrolled and undercontrolled personality patterns

among violent offenders. Sociological Symposium 9; 37-50.

Mookherjee, Harsha N. and H. Wayne Hogan 1979 Delinquency as a function of personal and social controls. Paper presented at

the Rural Sociological Society Meetings.

Nie, Norman H., C. Hadlai Hull, Jean G. Jenkins, Karin Steinbrenner, and Dale H. Bent 1975 Statistical Package for the Social Sciences. 2nd ed. New York: McGraw-Hill.

AGE DIFFERENCES IN DELINQUENCY 41

Nye, F. Ivan 1958 Family Relationships and Delinquent Behavior. New York: Wiley.

Parsons, Talcott 1959 The school class as a social system: Some of its functions in American society.

Harvard Educational Review 29: 297-318.

Polk, Kenneth 1971 A reassessment of middle-class delinquency. Youth and Society 2: 333-353.

Poole, Eric D. and Robert M, Regoli 1979 Parental support, dehnquent friends, and delinquency: A test of interaction

effects. Journal of Criminal Law and Criminology 70: 187-193.

Portes, Alejandro and Kenneth L. Wilson 1976 Black-white differences in educational attainment. American Sociological

Review 41; 414-431.

Rankin, Joseph H. 1977 Investigating the interrelations among social control variables and conformity.

Journal of Criminal Law and Criminology 69: 470-480.

Regoli, Robert M. and Eric D. Poole 1978 The commitment of delinquents to their misdeeds: A reexamination. Presented

at the World Congress of the International Sociological Association.

Reiss, Albert J., Jr, and A. Lewis Rhodes 1961 The distribution of juvenile delinquency in the social class structure. American

Sociological Review 26; 720-732.

Robhins, Lee N. and Eric Wise 1977 Childhood deviance as a developmental process: A study of 223 urban hlack

men from birth to 18. Social Forces 56: 448-473.

Schoenberg, Ronald J. 1975 A structural model of delinquency. Unpublished doctoral dissertation. Seattle:

University of Washington.

Shover, Neal, Stephen Norland, Jennifer James, and William E. Thorton 1979 Gender roles and delinquency. Social Forces 58: 162-175.

Slocum, Walter L. and Carol L. Stone 1963 Family culture patterns and delinquent-type behavior. Journal of Marriage and

Family Living 25: 202-205.

Smelser, Neil J. and Sydney Halpwrn 1978 The historical triangulation of family, economy and education. American Jour-

nal of Sociology 84: 288-315.

Smith, Douglas A. 1979 Sex and deviance: An assessment of major sociological variahles. Sociological

Quarterly 20: 183-196.

Thomas, Charles W. and Jeffrey M. Hyman 1978 Compliance theory, control theory and juvenile delinquency. In Marvin Krohn

and Ronald L. Akers (eds.). Crime, Law and Sanctions. London; Sage.

Thomas, Charles W., Gary A. Kreps, and Robin J. Cage 1977 An application of compliance theory to the study of juvenile delinquency. Soci-

ology and Social Research 61: 156-175.

42 L A G R A N G E AND WHITE

Toby, Jackson 1957 Social disorganization and stake in conformity: Complementary factors in the

predatory behavior of hoodlums. Journal of Criminal Law, Criminology and Police Science 48: 12-17.

1974 The socialization and control of deviant motivation. In Daniel Glaser (ed.). Handbook of Criminology. Chicago: Rand McNally,

Wiatrowski, Michael D., David B. Griswold, and Mary K. Roberts 1981 Social control theory and delinquency. American Sociological Review 46: 525-

541.

Wilson, Nanci K. 1972 Risk ratios in juvenile delinquency. Ann Arbor: University Microfilms.

Zucker, Herbert J. 1943 Affectional identification and delinquency. Archives of Psychology 286: 1-60.

Randy L. LaGrange is an Assistant Professor of Sociology at the University of North Carolina at Wilmington. His research interests include juvenile delinquency, social control, and the police.

Helene Raskin White is an Assistant Professor of Sociology at the Center of Alcohol Studies, Rutgers, The State University of New Jersey. Her current interests center on ado- lescent substance use and delinquency, and she has authored several articles in the field of alcohol and drug studies.

AGE DIFFERENCES IN DELINQUENCY 43

APPENDIX

Items Used For Index Construction

A. General Delinquency Index HOW MANY TIMES IN THR LAST THREE YEARS HAVE

YOU. . . . 1. Avoided paying for such things as movies, bus or subway rides, and

food? 2. Broken into a building to look for something to steal or to steal

something? 3. Used a weapon, like a club, knife, or gun in a fight? 4. Stolen (or tried to steal) a motor vehicle, such as a car or

motorcycle? 5. Run away from home? 6. Hurt someone badly enough so that the person needed bandages or

a doctor? 7. Damaged property on purpose (such as slashing tires, breaking

windows, setting fire to someone else's property)? 8. Stolen things worth LESS THAN $50.00? 9. Stolen things worth OVER $50.00?

10. Obtained money or gifts by engaging in sex or by finding customers for prostitutes?

11. Knowingly bought, sold or held stolen property? 12. Run numbers or had a job connected with illegal gambling? 13. Been involved in forgery (signed somebody else's name to checks,

etc.)? 14. During the last year, how many times did you cut school (play

hooky) because you had other things you wanted to do? 15. During the last year, how many times did you get into trouble at

school for fighting?

B. Parental Love HOW OFTEN DOES YOUR MOTHHR/FATHER. . . .

1. Praise or compliment you? 2. Comfort you when you are afraid? 3. Notice when you are good at home or in school? 4. Enjoy talking things over with you? 5. Seem to know what you need or want? 6. Talk with you? 7. Make you feel better after talking over your worries with them? 8. Give you a lot of care and attention? 9. Act very patient with you?

44 L A G R A N G E AND WHITE

10. Speak to you with a warm and friendly voice? 11. Cheer you up when you're sad? 12. Smile at you? 13. Speak of the good things you do? 14. Enjoy doing things you do? 15. Enjoy going on drives, trips, or visits with you? 16. Make you feel wanted?

C. Attachment to parents 1. How much do you respect your mother/father? 2. How much do you trust your mother/father? 3. How much do you love your mother/father? 4. How much do you admire your mother/father? 5. How much would you like to be the kind of person your mother/

father is?

D. School Performance 1. Over the last three years, what is the average grade you received in

ENGLISH? 2. Over the last three years, what is the average grade you received in

MATH? 3. What is the average grade you received on your last report card?

E. Attachment to School 1. How often do you find school satisfying because it gives you a sense

of accomplishment? 2. How often do you think school is a real chance for you: that it can

make a real difference in your hfe? 3. How often do you feel that school is boring and you are not learn-

ing what you feel is important? 4. How often do you enjoy school because it gives you a chance to be

with people your own age and to do a lot of things that are fun? 5. How many of your teachers do you like and get along with?

F. Educational Commitment 1. How often do you put a lot of energy into what you do in school? 2. How often do you try to get the best grade? 3. How often do you stick to classwork in your classes? 4. How often do you try to be the first one in class to answer

questions? 5. How often do you take part in class discussions or activities? 6. How often do you do your best work in school?

A G E D I F F E R E N C E S I N D E L I N Q U E N C Y 45

G. Delinquent Associates HOW MANY OF YOUR CLOSE FRIENDS. . . .

1. Do things that might get them into trouble with the law? 2. Have been picked up or arrested by the pohce? 3. Have used a weapon, like a club, knife or gun, in a fight? 4. Have stolen things worth LESS THAN $50.00? 5. Have stolen things worth MORE THAN $50.00? 6. Have broken into buildings to look for something to steal or to steal

something? 7. Have damaged other people's property on purpose? 8. Have been on probation or parole? 9. Have been suspended from school?