Informal Social Control and High Drug Use in Louisville and Lexington - Statistics Project Example

Informal Social Control and High Drug Use in Louisville and Lexington, Kentucky Introduction This research paper investigates the roles that informal social control plays in the rising high drug usage areas of Louisville and Lexington, Kentucky. Using data from the Interuniversity Consortium for Political and Social Research (ICPSR), the relationship between informal social control and high drug usage is sought. Background Social control denotes a form of regulation exercised by the authorities on groups of people or individuals to make them conform to a specific set of rules or social-legal parameters. In essence, social control is the key aspect behind the transformation of most immigrant families into taking up norms and practices that are practiced in their new settlement areas. Regulation takes two broad forms: formal (constitutes regulation by authorities to prevent society from falling into some ungovernable state; for instance, employment of sanctions by law enforcement authorities) and informal (which involves an adaptation process triggered by socialization – the process of transforming persons with a potentially wide range of behavioral traits to conform to a narrower set of values accepted within the society they are adapting to (Warner, Leukefeld & Kraman, 2003). Yet transforming to an extent of fully conforming to the set social and legal standards that govern a group of individuals is not easy. This not only affects individuals who are new to a locality but it also forms a basis for social and legal strife among members of all ranks. Notably, the transformation process has been associated with challenges, including engagement in drug abuse and alcoholism, burglary among other antisocial behaviors (Fulkerson, Pasch, Perry & Komro, 2008). Research into the effects of informal social control is largely confined within the principles of the social control theory. The theory proposes four types of control: internal, indirect, direct, and control through satisfaction (Pratt, Gau & Franklin, 2011). Informal social control is of specific importance since it involves aspects of the individual’s ability to interact and socialize effectively in order to conform to the desired socio-legal model. Warner, Leukefeld and Kraman (2003) noted that the influence of social ties in mediating conformity to set social standards is unclear. For instance, ghetto-related behaviors tend to mutilate community values, and the resulting community is characterized by ambiguity in value adoption; with a section of the population drifting towards the larger community’s ideals and the remaining section adapting to ideals specific to the locality. This way, social ties have no overall influence on social control, but can still be attributed to partial influence. On the other hand, social ties are perceived as a pathway to effecting informal social control. For instance, where a community is largely radicalized, social ties are more likely to result in reduced levels of informal social control. The above two contrasting notions tend to draw our attention to a realm that is boundless due to the contrasting nature of results from which Warner, Leukefeld and Kraman (2003) borrowed their literature. It is, therefore, very important to seek a justifiable, results-based opinion on the issue, as a way to supplement the theoretical understanding of the issue. What these opinions point to in general is that social classification (for instance, on the basis of income, and other demographic factors such as gender, number of family members in neighborhood) plays an important role in deciding the effectiveness of social control across the various social classes (Green, Doherty, Reisinger, Chilcoat & Ensminger, 2010). This is because persons with higher income will tend to present a more formal image of themselves, and will be less likely to engage in petty crime and other relatively antisocial behaviors than those with low pay or no employment (Schroeder, Giordano & Cernkovich, 2007). These arguments lead us to a number of relevant research questions that this research seeks to address. Research Questions The present research seeks answers to three vital questions that appear important in deciding how members of any community conform to informal social control. The main research question draws from the relationship between the strength of informal social control based on the number of relatives living in the neighborhood and prevalence of drug-related arrests. At this point, the researcher poses the question: does having more relatives from the neighborhood (who could actively engage in influencing the process of social control) lead to higher lower number of arrests? The general perception is that when there are more family members to trigger successful social control, the process is more likely to be fast and successful, which rightly explains an expected inverse correlational relationship between the number of relatives and the number of arrests (Schroeder, Giordano & Cernkovich, 2007). The number of arrests is interpreted to mean poor uptake of social control. Accordingly, the commission of more crimes are posited to be directly proportional to a lower number of relatives in the neighborhood. The corresponding hypothesis is listed first at the bottom of this sub-section. Equating the number of drug-related arrests to poor uptake of social control, the relationship between the variable and income is likely to shed light on whether the number of arrests is a function of the income status. The research question: Are drug-related arrests a function of the income? Hypothetically, the people who tend to be influenced more into drug dealing are the financially vulnerable – people without jobs or those hanging on to poorly-paying ones. As such, the number of arrests are expected to have an inverse relationship with the income – the higher the income the lower the number of arrests on drug-related charges and the vice versa. Several other important variables are estimated to affect the number of arrests made on an individual based on drug-related charges. For instance, age is relevant in the sense that more young people are more likely to yield to peer pressure and engage in crime while employment status marks the ability to harness some income and consequently refrain from engaging in criminal activities (Jones-Johnson, DeLisi & Hochstetler, 2012; Nilsson, Estrada & Backman, 2014). The two variables will be added to the initial regression model developed above to determine whether they have significant impact on the suitability of the model. Hypotheses sets 2 and 3 are used to test the significance of the univariate and multivariate models. Hypothesis 1: H0 (Null): Having more relatives (informal social control mechanism) is significantly, inversely correlated with the number of drug-related arrests made (unsuccessful social control). H1 (Alternative): Having more relatives (informal social control mechanism) is not significantly, inversely correlated with the number of drug-related arrests made (unsuccessful social control). Hypothesis 2: H0 (Null): The regression coefficient for income is not equal to zero (the model is significant). H1 (Alternative): The regression coefficient for income is equal to zero (the model is not significant). Hypothesis 3: H0 (Null): The regression coefficients for income, age and employment status are not equal to zero. H1 (Alternative): The regression coefficients for income, age and employment status are equal to zero. Methods For the analysis, data from the neighborhood-level study “Informal Social Control of Crime in High Drug Use Neighborhoods in Louisville and Lexington, Kentucky, 2000” by Warner (2003) was used. The data was collected between February and June 2000, representing 66 neighborhoods within the cities of Louisville and Lexington. The initial aim of collecting the data was to conduct a vast number of analysis to help understand the community context within which behaviors occur and to help predict the behavioral patterns associated with particular sociological aspects. The data is clearly capable of providing insights into the research questions posed above. Both survey and administrative records data were available. For the purposes of this analysis, the administrative records’ data was relied upon for its choice of variables. The variables are described below. Dependent Variable (Response) The same dependent variable was used across the study. The variable, ‘number of drug-related arrests’ highlights the inter-neighborhood arrests emanating from drug-related offences, such as sale and open use. The variable is denoted as a continuous, scale variable indicating the number of such arrests made in each neighborhood. The variable marked the failure or success (where fewer arrests were made) of informal social control in reigning on drug use. Independent Variables Four independent variables were used at two levels of the study. The first independent variable (IV) to be used was the ‘number of relatives living in the neighborhood’. The variable provided an alternative method of social control that could help prevent an individual from engaging in drug-related crimes. The second IV is the ‘median income of residents within a neighborhood’. Having high incomes is considered a deterrent for drug-related crimes, therefore a form of informal social control. It is a continuous, scale variable. Similarly, the third IV, ‘proportion of unemployed individuals within a neighborhood’ represents the fraction of unemployed to total neighborhood population. It is a scale variable ranging between 0% and 100%. The fourth IV was the average age of the population living within the neighborhoods. The population sampled comprised all mature individuals, capable of making legally binding decisions and whose exposure to working is considered practicable. Age is a continuous, scale variable. Analytical Plan The hypotheses were investigated using two basic techniques: correlation and regression analysis. However, the basic characteristics of the data were initially established using both descriptive and frequency analysis. Frequency analysis was used to indicate the percentage of respondents drawn from each of the two cities – Louisville and Lexington. Descriptive analysis was used to provide insights into the population distribution, proportions of males across the neighborhoods listed against the two cities, the proportions of the three leading ethnic groups (White, African-American and Hispanic) within the cities, the proportions of full-time employees in every neighborhood, the average number of drug-related arrests made per neighborhood, average age of persons in every neighborhood, the median income, and the proportion of the population that is unemployed. The next step involved ascertaining whether the data was fit for the intended tests. In particular, regression analysis requires that data meets some prerequisites, all related to distribution, multicollinearity, linearity, and homoscedasticity. Once ascertained, the two final tests (regression and correlation analysis) were run in the SPSS software to investigate relationships among various variables. Regression analysis established the nature and existence of both bivariate and multivariate relationships. In addition, the test highlighted the actual models relating the response to the independent variables. Correlation analysis sought to highlight the linear relationships between each pair of variables. Results Tests of Regression Assumptions The tests carried out were mainly correlation and regression analysis. While correlation analysis is not exclusively parametric, regression analysis is parametric and its successful implementation depends on the data’s fulfilment of specific assumptions. Accordingly, the data was subjected to various tests to establish whether it was suitable for regression analysis as obtained, or further transformation of the selected variables was necessary before carrying out the analysis. The first of the four assumptions investigated was the test of normality. The procedure involved reading both the Q-Q plots generated via SPSS and the Shapiro-Wilk (S-K) test. This test has high precision and handles both small and relatively large sample/ population sizes, spanning to 2000 (Razali & Wah, 2011). The response across all four tests was the number of drug-related arrests made. The overall null hypothesis was that each of the variables is that the data is normally distributed. The median income per household was normally distributed (S-W = 0.172, df = 9, p = 0.200). The normal Q-Q plot was almost perfectly linear. The average ages for respondents from the various areas within Louisville and Lexington were also normally distributed (S-W = 0.823, df = 4, p = 0.150). The normal Q-Q plot for this variable was also fairly perfect. The variable ‘number of relatives in neighborhood’ was normally distributed (S-W = 0.921, df = 4, p = 0.542). The normal Q-Q plot was fairly linear with slight deviations from the normal line. The final independent variable, ‘proportion of population unemployed’ was normally distributed (S-W = 0.831, df = 8, p = 0.061). The normal Q-Q plot was fairly linear. Therefore, all the independent variables fulfilled the test of normality. Next, tests of linearity between the dependent variable and each of the independent variables was undertaken. Evaluating linearity using scatter plot is the best approach since it involves an easy observance process that is easy to understand and interpret (Statistics Solutions, 2014). The tests yielded fairly similar results for the four independent variables, with their relationships with the number of drug-related arrests producing fairly linear plots. The assumption was consequently considered to be fulfilled for all variables. In addition, a test for the existence of extreme outliers was conducted using the box plot. Values occurring within certain extreme ends are automatically considered to be outliers. This test is especially important for the response, whose distribution is similar to that of the error terms of the independent variables. The test did not detect any extreme outliers, which further reinforced the notion that the above test of linearity was not flawed. The third assumption tested was the lack of existence of extremely weak multicollinearity. The approach adopted involved investigating the level of bivariate correlation between each pair of independent variables. The largest and smallest absolute correlation coefficients were 0.545 and 0.197. Despite these values being relatively smaller (suggesting no problems with collinearity), a further test (test of tolerance) was undertaken. The test involves using all independent variables to create a regression model against which tolerance (1 – R2) can be evaluated. The R2 obtained was 0.400. With the tolerance (1 – 0.400 = 0.600) which is greater than 0.1 there is no risk of multicollinearity. Further, a variance inflated factor did not surpass the rule of thumb benchmark of 10, suggesting conformity to the earlier finding. This assumption was fulfilled as well. The fourth assumption against which the data was tested is that of autocorrelation. Autocorrelation shows that the residues (which, by extension, affects the response variable) are not independent of each other. The problem is most common in time series data. So the default assumption is that the data is not auto-correlated. The Durbin-Watson test was preferred for testing this assumption. It was carried out within the initial regression test involving all the independent variables selected for this analysis. The Durbin-Watson statistic obtained was 1.828, which, by virtue of being very close to 2, indicated absence of autocorrelation (Statistics Solutions, 2014). The final assumption tested was that of homoscedasticity. It is a test of whether the error term is same across all the values of the independent variable (Statistics Solutions, 2014). The test was done using both scatter plots and the chi-square test. The Cook-Weisberg/ Breusch-Pagan test yielded a chi-square value of 2.35 (1 degree of freedom). The p-value associated with the statistic was 0.146, which implies that we cannot allude that the data is heteroscedastic. The plots obtained suggested the data was homoscedastic, thereby fulfilling the final assumption for undertaking regression analysis. Based on the above five tests, the data fulfills all assumptions investigated and, therefore, it qualifies for regression analysis. Descriptive Analysis Descriptive analysis was undertaken on select variables of interest to shed light on who made part of the sample. As noted in the introduction, the population was drawn from the Kentucky cities of Lexington and Louisville. Overall, 34 (51.5%) of the areas that contributed to the present data were based in Lexington while 32 (48.5%) were from Louisville. The neighborhood with the highest total population had 5,364 while the one with the lowest had 195 people. The average population across the neighborhoods was 895.08 (std. dev. = 797.87). The proportion of males in the neighborhoods was highest at 67.58% and lowest at 30.50%, with the average proportion running at 47.06% (std. dev. = 7.20). Some neighborhoods had 100% White populations while some had 0.0%. The average proportion of Whites was 50.20% (std. dev. = 34.48). Similarly, some neighborhoods had 0% Blacks and 100% of persons of the African-American racial group. On average, Blacks comprised 48.92% (std. dev. = 34.84) of all neighborhoods. Hispanics had the highest representation at 4.1% and the lowest at 0.0%. They comprised 0.54% (std. dev. = 0.96) of the inter-neighborhoods’ populations. The neighborhood with the lowest proportion of people employed on a full-time basis had 20% while the highest had 74.29%. On average, the proportion of people employed on full-time basis was 20% (std. dev. = 13.3%). Other variables explored included those used to model the situation in the following step. These include the average number of drug-related arrests made per neighborhood, average age of persons in every neighborhood, the median income, and the proportion of the population that is unemployed. The highest number of drug-related arrests made in any neighborhood was 132 while the lowest was 0. The average number of such arrests was 24.70 (std. dev. = 27.62). The highest median income per neighborhood was $52,324 while the lowest was $4,999. The average median income was computed at $16,982.23 (std. dev. = $10,052.99). The highest proportion of unemployed persons living in any particular neighborhood was 57.14% while the lowest proportion was 0.0%. On average, the unemployment ratio was 19.70% (std. dev. = 13.29). The table below (table 1) shows the descriptive characteristics of the variables used in the regression and correlation analysis. Table 1. Univariate statistics.               Mean Std. Dev. Min Max Dependent Variable Number of drug-related arrests 24.70 27.62 0 132 Independent Variables Proportion unemployed 19.70% 13.29% 0.0% 57.14% Median income 16982.23 11052.99 4999 52324 Age 46.04 5.37 30.42 58.57 Bivariate Analysis Two tests were done at this level. The tests were basically conducted to establish whether there existed any significant relationships between the selected variables. The first test investigated the question: does having more relatives from the neighborhood (who could actively engage in influencing the process of social control) lead to higher lower number of arrests? Effectively, it responds to the hypothesis: H0: Having more relatives (informal social control mechanism) is significantly, inversely correlated with the number of drug-related arrests made (unsuccessful social control), against the alternative: H1: Having more relatives in the neighborhood is not significantly, inversely correlated with the number of drug-related arrests. The number of drug-related arrests were significantly positively correlated with the number of relatives in the neighborhoods (ρ = 0.332, p = 0.006). This implies that as the number of relatives increases within a neighborhood, the number of drug-related arrests are likely to rise. This is in violation of the pre-meditated hypothesis. This paints a picture suggesting that more relatives in any one neighborhood could be linked to collusions to trade in drugs. The other bivariate test was a correlation and regression analysis of the relationship between the number of drug-related arrests and the median family income per neighborhood. Basically, the expectation would be that the higher the income, the lower the number of such arrests made, as highlighted in the introduction. The results suggested that the number of arrests are significantly, inversely correlated with the median income per household (ρ = -0.492, p < 0.001). This finding is in line with the pre-stated hypothesis that established grounds for such findings. The correlation coefficient was also non-zero (F = 20.40, p < 0.001). This implies that there is a true, linear relationship between the two variables. The table of coefficients suggests the following linear model to describe the relationship: Number of Drug-Related Arrests = 45.56 – 0.001*(Median Income). In essence, the higher the number of drug-related arrests, the lower the average income for a neighborhood, and vice versa. The independent variable (median income) accounted for 24.2% of the change observed in the response (number of drug-related arrests made). Suggestively, this figure is low, implying that there could be more factors that account for an even larger portion of the variation in the response. To investigate this notion, the multivariate analysis of the variables was undertaken. Multivariate Analysis Two more independent variables were introduced into the model to determine whether they are as relevant as hypothesized in explaining the variation within the dependent variable. The new variables are the proportion of unemployed people within the neighborhood and age. The test responds to the question: does the addition of proportion of unemployed persons and age variables make the model fit better? The results show that the model is statistically significant (F = 13.78, p < 0.001). The relationship is captured in the model: Number of Drug-Related Arrests = 10.99 – 0.001*(Median Income) + 0.062*(Age) + 101.87*(Proportion of Unemployed Persons in a Neighborhood) + 12.66*(Proportion Attending Neighbourhood Meetings). The three independent variables account for 37.1% (adjusted R2 = 0.371) of the total variation observed within the response. In total, age, unemployment rate, and median income accounted for only 37.1% of the change in the number of drug-related arrests. Inevitably, there are additional variables which have not yet been explored and that partly account for the variation. Nevertheless, the model improved with the addition of the two additional independent variables. Consequently, the addition made it a better fit than when only one independent variable was employed. The multivariate results are shown in table 2 below. The coefficients of the variables in the model had contrasting significances. This raises a question on whether the non-significant ones have an impact on the overall response. Based on the results, only the ‘proportion of the unemployed’ appeared to be statistically significant (β = 101.87, p = 0.001). The implication of this finding is that the other variables are relatively unimportant in the model, and cannot be relied upon to interpret changes in the response. For this reason, they will not be given further consideration. For the significant ‘proportion of the unemployed’, we find that the estimated rate of change in the conditional mean of ‘number of drug-related arrests’ with respect to the stated proportion, holding all other variables constant, ranges between 46.09 and 157.64. This is in line with the confidence interval of the independent variable. Table 2. Unstandardized regression coefficients (and standard errors) for number of drug-related Arrests Constant Constant 10.99 (31.47) Social control variable Proportion attending neighbourhood meetings 12.66 (21.60) Independent variables Age 0.062 (0.604) Proportion unemployed 101.87* (27.90) Median income -0.001 (0.000) R2 0.400 Adjusted R2 0.371       (*) Shows the statistic is significant at the 0.05 level of significance. Discussion/ Conclusion The previous section presented both confirmatory and negative results on the pre-stated hypotheses. There was no full concurrence with the hypotheses. For instance, the results presented in the previous section suggest that indeed, the perceived inverse relationship between the number of drug-related arrests and the number of relatives in a neighborhood is misconceived. This is suggested by the positive (in place of an inverse) correlation between the number of arrests and the number of relatives in the neighborhood. By extension, the results suggest that the more the number of relatives within the neighborhood, the more the expected number of drug-related arrests. There could be high levels of collusion between members of same families to engage in drug-consumption and trade. The literature consulted confirms the opposite: family members help to mold an informal environment of social control, inspired by a sense of responsibility and accountability for ones actions (Warner, Leukefeld & Kraman, 2003). Hence, being close to family member can be viewed as a negative influence in the sense of how it contributes to one being more responsible and accountable; instead the feeling that takes over the relationship between family members is to help each other perfect their drug dealing activities. Income variation is one of the largest contributors of drug-related arrests. Inequality at this level suggests that those people with lower incomes are more likely to engage in drugs, which will ultimately lead to their arrest. On the other hand, employment (or lack of it) is associated with income. This suggests an indirect relationship between the three variables. In its current state, the role of employment could be masked within the role of income, implying difficulties in trying to establish the net contribution of either independent variable towards the change in numbers of drug-related arrests. It is, therefore, wise to introduce income as a mediating factor of the relationship between arrests and employment instead of introducing the two as independent variables. Indeed, the level of correlation between employment (or unemployment) and the income status should ideally be much higher than was observed in this study. Nonetheless, the contribution of age towards the determination of number of drug-related arrests in the neighborhoods was noticeable. The significant, inverse relationship suggests that the younger population is more likely to be arrested for drug-related crimes, while the older people are less likely to experience the same. Delving further into the previous discussion, this suggests that unemployment could be higher among the younger generations, a finding supported by Green et al. (2010). This hypothetical relationship should form a basis of further investigation into this topic. References Fulkerson, J. A., Pasch, K. E., Perry, C. L. & Komro, K. (2008). Relationships between alcohol-related informal social control, parental monitoring and adolescent problem behaviors among racially diverse urban youth. Journal of Community Health. 33(6): 425-433. Green, K. M., Doherty, E. E., Reisinger, H. S., Chilcoat, H. D. & Ensminger, M. (2010). Social integration in young adulthood and the subsequent onset of substance use and disorders among a community population of urban African Americans. Addiction. 105(3): 484-493. Jones-Johnson, G., DeLisi, M. & Hochstetler, A. (2012). Gender differences in social relationships, social integration and substance use. Sociology Mind. 3(1): 106-113. Nilsson, A., Estrada, F. & Backman, O. (2014). Offending, drug abuse and life chances – A longitudinal study of a Stockholm birth cohort. Journal of Scandinavian Studies in Criminology and Crime. 15(2): 128-142. Pratt, T. C., Gau, J. M. & Franklin, T. W. (2011). Key ideas in criminology and criminal justice. New York: SAGE Publications. Razali, N. M. & Wah, Y. B. (2011). Power comparisons of Shapiro-Wilk, Kolmogorov-Smirnov, Lilliefors and Anderson-Darling tests. Journal of Statistical Modeling and Analytics. 2(1): 21-33. Schroeder, R. D., Giordano, P. C. & Cernkovich, S. A. (2007). Drug use and desistance processes. Criminology. 45(1): 191-222. Statistics Solutions (2014). Assumptions of linear regression. Retrieved from http://www.statisticssolutions.com/assumptions-of-linear-regression/. Warner, B. D. (2002). Informal social control of crime in high drug use neighborhoods in Louisville and Lexington, Kentucky, 2000. ICPSR Version. Lexington, Kentucky: University of Kentucky Research Foundation, Inc. Retrieved from http://www.icpsr.umich.edu/icpsrweb/NACJD/studies/3412/version/1. (Accessed 10th April, 2015). Warner, B. D., Leukefeld, C. G. & Kraman, P. (2003). Informal social control of crime in high drug use neighborhood: Final project report. Washington, DC: National Criminal Justice Reference Service. Read More