Descriptive Statistics
In this section, we will compare adolescents living in non-urbanized, sparsely urbanized areas and the rest of the country, based on the results shown in Table
1, which provides average scores and percentages for all indicators, by the area in which adolescents live.
Concerning delinquent behavior, adolescents living in sparsely urbanized areas engage in slightly less frequent delinquent behavior as compared to adolescents living in the rest of the country, whereas adolescents in non-urbanized areas do not differ from them. Note that standard deviations exceed mean scores, indicating that variation within the rural categories and within the rest of the country category is large.
We now turn to the indicators of communal social control. Both categories of rural adolescents participate in activities of associations and clubs more often as compared to all others. Both categories of rural adolescents also more often indicate that they belong to a religion. Adolescents living in non-urbanized areas go to religious meetings more often, as compared to their counterparts living elsewhere in the country, while the difference between adolescents living in sparsely urbanization areas and those living in the rest of the country is not significant in this respect.
With regard to peers and deviant behavior, there are no significant differences concerning the frequency of meeting friends. The number of alcoholic consumptions in the weekend (note that standard deviations exceed the mean scores) and the frequency of visiting pubs are clearly higher among the two categories of rural adolescents. Pub visiting is also higher among non-urbanized adolescents than among adolescents living in sparsely urbanized areas. Drug consumption and truancy do not differ significantly between the three categories of adolescents.
As for the socio-demographic control variables, the sample shows an equal gender distribution across the categories. Furthermore, adolescents living in non-urbanized areas are slightly older than their counterparts living in sparsely urbanized areas. This result is probably due to sampling bias. Also, the sample probably underestimates the proportion of ethnic minorities overall. However, for our purpose, the distribution of ethnic minorities across areas is more relevant. At this point, the sample does resemble the population: only very few of the rural adolescents are of ethnic minority descent, whereas the proportion of youngsters of non-Dutch ethnic descent is clearly higher in the rest of the country. Concerning educational level, no significant differences appeared between the three categories. Finally, the rates of parental divorce are significantly lower among the two categories of rural adolescents as compared to adolescents living in the rest of the country.
We conclude that the delinquency rates of adolescents living in sparsely urbanized areas are significantly lower than those of non-rural adolescents, while adolescents living in non-urbanized areas do not differ from their non-rural counterparts in this respect. Behind these delinquency rates different factors may play a role in delinquent behavior. Some indicators thought to reduce the chance of delinquent behavior score higher among both categories of rural adolescents (participating in activities of clubs or associations, religious affiliation, lower rate of parental divorce). On the other hand, some other factors that are known to increase the likelihood of delinquent behavior are higher among both categories of rural youngsters (drinking alcohol). In the following section, we will assess the impact of these factors in multivariate analyses.
Multivariate Statistics
We now set out to test the hypotheses. Table
2 shows the results of linear regression analyses of adolescents’ delinquent behavior (in these analyses, we transformed the delinquent behavior scale by taking its square root). The first model compares adolescents living in non-urbanized and sparsely urbanized areas with their counterparts living in the rest of the country (the reference category), only controlling for the year of data collection. It turns out that both rural categories engage in slightly less frequent delinquent behavior than adolescents living in the rest of the country.
In model 2, socio-demographic controls are added. In line with standard criminological research on juvenile delinquency, we see that males clearly are more likely to commit delinquent acts and that age also increases the likelihood of delinquent behavior. Ethnic minorities report a lower likelihood of engaging in delinquent behavior. The latter finding is inconsistent with prior Dutch criminological research based on self-report data (Kruissink and Essers
2004), which shows a slight overrepresentation of ethnic minorities in various forms of delinquent behavior. At this point, the sample might be biased, as our measure of ethnicity is not fully accurate. Finally, we see a negative effect of educational level. This means that adolescents following higher educational streams are less likely to engage in delinquent behavior. Including these controls only marginally alters the effects of both rural categories. Thus, the main conclusion we draw from model 2 is that rural adolescents have just a little bit lower delinquency rates after controlling for socio-demographic variables. We will use model 2 and the effects of both categories of rural adolescents specifically, as the reference for the models to follow.
We now turn to model 3, which includes indicators of communal social control. The first finding is that religious affiliation, church attendance and weekly participation in activities of clubs or associations slightly decrease the likelihood of engaging in delinquent behavior. The second finding is that controlling for these indicators of communal social control only marginally reduces the effects of both categories of rural adolescents, as compared to model 2. The latter finding rejects hypothesis 1: rural adolescents’ greater exposure to communal social control does not reduce their likelihood of committing delinquent acts substantially.
Model 4 includes the frequency of meeting friends, alcohol and drugs consumption, the frequency of visiting pubs and playing truant. Except for the frequency of meeting friends, these indicators significantly influence the likelihood of engaging in delinquent behavior and also considerably increase the overall explanatory power of the model. We observe that controlling for these variables only slightly affects the effects of both categories of rural adolescents as compared to model 2. This is due to controlling for their larger alcohol intake and, to a lesser extent, their greater frequency of visiting pubs. If we include only alcohol intake and visits to the pub in this model, the beta’s increase and are significant: −.073, and −.086 for adolescents living in non-urbanized and sparsely urbanized areas respectively. This means that drinking alcohol and visits to the pubs are associated with rural adolescents’ delinquent behavior in particular.
Model 5 includes variables of communal social control to test hypothesis 2 stating that the association between rural adolescents’ risk behavior and their delinquent behavior is substantially mitigated by greater exposure to communal social control. However, the effects of both categories of rural adolescents are reduced only marginally, revealing that rural adolescents’ exposure to communal social control does not decrease substantially the association between their risk behavior and their engagement in delinquent activities.
Model 6 includes two interaction terms that differentiate males according to their place of living in order to test hypothesis 3 that claims that the difference in delinquency rates between rural and non-rural girls is larger than that between rural and non-rural boys. The interaction terms do not yield significant results and the effects of the main terms of both rural categories increased slightly. This means that, in accordance with hypothesis 3, rural boys are on a par with boys living in the rest of the country while rural girls, indicated by the increased main effects of both rural categories, are less likely to engage in delinquent behavior than girls living in the rest of the country. Note also that the gender gap among rural adolescents is slightly larger among adolescents living in non-urbanized areas, as the change of this effect is of greater magnitude than that of adolescents living in sparsely urbanized areas.
In model 7, we include the consumption of alcohol during weekends and the frequency of visiting pubs in order to test hypothesis 4, which claims that drinking alcohol and visits to pubs have a stronger effect on young rural males’ delinquent behavior as compared to young males living elsewhere. If this hypothesis is valid, we expect that including drinking alcohol and the frequency of visits to the pub decreases the effect of the interaction terms Non-urbanized × Males and Sparsely urbanized × Males. Indeed, we observe that the effects of both interaction terms have decreased, although it should be noted that the reduction is of small magnitude. We thus find support, albeit modest, for hypothesis 4. Moreover, the negative effect of the two main terms (non-urbanized and sparsely urbanized areas) also increased, meaning that alcohol intake and visits to the pub also affect female rural adolescents more strongly than female adolescents living in the rest of the country. And this association is slightly stronger for rural female adolescents living in non-urbanized areas than for their counterparts living in sparsely urbanized areas. As we already observed, drinking alcohol and visiting pubs are correlates of delinquent behavior for both rural girls and rural boys in particular.
Finally, we included parental divorce in model 8 to test hypothesis 5. This hypothesis claims that the lower rate of parental divorce substantially reduces rural adolescents’ likelihood of committing delinquent acts. Here we observe that parental divorce yields a significant effect on the likelihood of engaging in delinquent behavior. Unlike hypothesis 5, however, this results only in very small changes in the effects of both categories of rural adolescents.
We can infer the following from these analyses. First, we found somewhat lower delinquency rates among both categories of rural adolescents, after controlling for socio-demographic variables. Second, this difference cannot be accounted for by the greater exposure to communal social control that rural adolescents experience, as communal social control only very marginally inhibits delinquent behavior. Third, the greater numbers of alcoholic drinks in the weekend and the greater frequency of visiting pubs are a correlate of delinquent behavior for both rural girls and rural boys in particular, and this is even more the case for adolescents living in non-urbanized areas. Fourth, communal social control only marginally reduces the association between delinquent and risk behavior among rural adolescents. Fifth, the likelihood of engaging in delinquent behavior does not differ between rural and non-rural boys, whereas rural girls commit delinquent acts less often than non-rural girls. This also means that the somewhat lower delinquency rates of rural adolescents are largely the result of rural girls’ lower likelihood of engaging in such behavior. The gender gap in delinquent behavior among rural adolescents is slightly larger in the non-urbanized areas than in the sparsely urbanized areas. Sixth, even though divorce rates are lower among the parents of rural adolescents, this only very slightly reduces delinquent behavior among rural youth.
Before we move to the discussion, we first show the results of additional analyses in which we used different indicators to measure degrees of urbanization. We ran the same models as presented in Table
2 twice. In the first round we added a “big city” dummy variable to the two rural categories. The additional dummy variable concerns the very strongly urbanized areas containing the larger part of what the Dutch consider “big cities” of over 400,000 inhabitants (Amsterdam, Rotterdam, The Hague and, to a lesser extent, Utrecht and some parts of smaller cities). The reference category now consists of intermediate categories of urbanization: moderately and strongly urbanized areas (see the section on method above). In the second round, we replaced the dummy variables with the five-item ordinal scale of urbanization. Table
3 gives the beta coefficients of the “big city” dummy variable and the five-item ordinal urbanization measure. The effects of all other indicators were very similar to the ones displayed in Table
2.
Adding the “big city” dummy variable to the other two rural category dummy variables neither improves model fit, nor yields any significant effects. This means that adolescents living in these most strongly urbanized areas do not differ from adolescents living in the rest of the country. Replacing the two rural dummy variables with the ordinal urbanization measure yields a small significant positive effect in all models. As we just saw that adolescents living in the most strongly urbanized areas do not differ from those living in the rest of the country, the positive effect of the ordinal urbanization measure is not so much displaying a linear association between delinquent behavior and population density, but shows the existence of a threshold between rural and non-rural areas. To put it differently: Dutch rural adolescents are somewhat less likely to engage in delinquent behavior, not just compared to their counterparts living in big cities but also vis-à-vis adolescents living in all non-rural areas.