Impact of Poverty on Violent Crime - Coursework Example

?Investigating the impact of poverty on violent crime Research question The research question this work intends to explore is whether poverty causes violent crime and if so, then to what extent. The relationship between general crime and poverty is a contended one. Citing instances of nations where crime rates are substantially lower in spite of high poverty, some authors argue that there is in reality no causality between crime and poverty. Others however strongly object citing instances such as majority of the population convicted serving out their jail terms come from poorer families or have relatively lower annual incomes. However, even if it is admitted that poverty does engender crimes such as petty thievery or other non-violent crimes, what the impact is of poverty on violent crimes does constitute an open question and this is exactly what the present endeavour aims to investigate. Theory and hypothesis The standard socio-economic argument for criminal behaviour posits that crimes are committed only if the potential benefit outweighs the potential cost. While the potential benefit includes successful evasion of authorities so that there is an increase in the individual’s lifetime income, the potential cost includes getting caught and convicted. If an individual is caught and convicted and sentenced to a jail term, he loses his potential earnings for the duration of the jail term that he would have otherwise earned. The jail term however ensures three square meals a day and a shelter, however small. Therefore, it is easy to see that poorer an individual, the lesser is the potential cost of the crime if he is caught and convicted. Therefore the poorer among the population have a stronger economic incentive to commit a crime. Additionally, if poverty is also coupled with a lack of education which in turn lowers the individual’s potential legal lifetime income, it is likely that he will have relatively stronger incentives to commit to criminal behaviour. The particular concern of this paper is the impact poverty has on violent crimes. We intend to look at whether incentives to commit crimes that are generated because of hailing from backgrounds of poverty can explain the incidences of violent crimes as well. The central idea is that poverty reduces the potential cost of being convicted and sentenced relative to the potential benefit of the crime to the individual and therefore increases the incentive for all types of crime. If this is true, poverty should increase the incidence of violent crime as well. Our hypothesis therefore is that increase in poverty would lead to increase in rate of violent crimes as well. Data In this section we shall briefly comment on the source of the data and present a table to convey the nature of the variables in the dataset chosen. A state-wise cross section dataset from the US for the year 2005 is used for the purpose of the study. The source of the data is the statistical abstract of the united states, 2005 and is available from the webpage for the textbook “Statistical Methods for the Social Sciences,4th Edition” by Agresti and Finlay (2008)1. Table 1 presents and index including details about the variables used. Table 1: Index of variables used and codes Variable Code Details Violent Crime Rate VI Outcome Variable, measures the number of violent crimes per 100,000 population Metropolitan population ME Control Variable, measures the percentage of Population in Metropolitan areas White population density WH Control Variable, measures the percentage of whites in population Education HS Control Variable, measures the percentage of population that are high school graduates Poverty PO Explanatory Variable, measures the percentage of population with incomes below the poverty level Method This section will present the methodological design that is going to utilise the aforementioned data set to explore our research question and test the hypothesis that poverty positively affects violent crime. The basic methodology that we intend to use is that of a linear least squares regression analysis. State wise violent crime rate is our outcome variable and the population in each state with incomes below the poverty level is our main explanatory variable. We include education, number of whites in population and population in metropolitan areas for each state as control variables to account for the variations in violent crimes that changes in these factors could potentially cause. We estimate the following linear relationship: (1) Where is the outcome variable representing violent crime rate is the constant or the autonomous part of crime, the parameters through are the coefficients that measure the degree of variability of violent crimes explained by the variability of the corresponding variable. Thus, the coefficient of PO is our parameter of primary interest. is the random disturbance term or the error term. Our null hypothesis is that PO or the incidence of poverty has no impact on violent crime which translates to the coefficient on PO being zero and we shall test this against the alternative hypothesis that the coefficient is non-zero. Symbolically therefore, we wish to test: Against the alternative: Results In this section we shall present the results of our statistical analysis of the data. We start of by presenting the summary statistics. In table 2 below we present the summary statistics of our data. Table 2: Summary statistics Next, to gain an understanding of the state wide variations in violent crime and the potential correlations, we look at the following matrix of scatter plots. It presents the scatters of violent crimes with each of our independent variables. Figure 1: Scatter-plot matrix of variables of interest In figure 1 above, the left most columns is the column of interest. It shows the scatters diagrams of violent crime with the other variables of interest in each row. A high positive trend is notable in the 2nd row 1st column scatter which traces how violent crime rates vary with the degree of population in metropolitan areas. The scatter of violent crime with percentage of high school graduates does not exhibit any particular patterns but a positive trend is apparent in the scatter of violent crime with poverty (the last row 1st column). From the 3rd row of the 1st column it seems that percentage of whites is negatively correlated with number of violent crimes caused. Thus, not only do the scatters seemingly support our hypothesis, it also seems to show that some of the variables chosen as control (explained further in the next section) exert some influences as well. Now, we turn to the main results. Table 3 presents the main results of the regression. Table 3: Results of regression (all states included) Recall that we used OLS to estimate equation (1). Thus, our estimated equation is: Before we proceed to interpret this estimated equation it is crucial to note the significances of the coefficients. The 5% critical value for the two way t-test is 1.96. From table 3, we find that the variables HS and the Constant are associated with t-values that have smaller absolute values. Thus, the estimates of the coefficients and cannot be treated as being different from zero. From the table it is also evident that all other estimated coefficients are statistically significantly different from zero. Turning to equation (2), the first point to note is that the coefficient on the variable PO is positive and since the corresponding t-value is 3.52, the coefficient is significant. Additionally it is the coefficient with the highest magnitude among our estimates. Therefore, we reject our null hypothesis in favour of the alternative. This implies that we have statistical evidence for the fact that variability in poverty does cause variability in violent crime. The positive sign of the coefficient tells us that if poverty rises then the expected impact on violent crime is that of an increase on average violent crimes. Apart from the parameter of interest, the regression results also provide valuable information regarding the impact of the control variables on the incidence of violent crimes. The estimated parameter that measures the effect of changes in degree of population living in metropolitan areas is positive and significant as well. This implies greater metropolitan areas cause more violent crimes. Finally, the coefficient of the variable WH is negative but significant. This implies that increases in the number of whites in the population reduce violent crime. The alternative implication is that increase in the number of non-whites causes increases in violent crimes. Table 4 presents the Variance Inflating Factors for each of the variables included in the regression to show that multicollinearity does not smudge the efficacy of the estimation procedure. Multicollinearity problems are evident in VIF values being greater than 10. From table 4 we find none of the computed VIFs are even close to this number. Table 4: Testing for multicollinearity using VIF Thus, we find that the basic OLS regression does provide statistical support to the notion that poverty causes violent crime. However, looking at the violent crime rates in the data, it is quite evident that the state DC is an obvious outlier. While the mean violent crime rate for the sample excluding DC is 402.32, the violent crime rate of DC itself is 1608. Thus, inclusion of this outlier could have led to erroneous estimates. Rather than deleting the observation however, we resort to using robust regression analysis. The results of running a robust regression are presented below in table 5. Table 5: Results of robust regression From table 5 what we find is that only the variables ME and WH are significant. The estimated coefficient of PO is no longer statistically significant from zero. This is evident from the t-values presented in the table. Recall that these t-values are those computed to test whether the coefficients equal to zero or not. The 5% critical value is 1.96. The only variables that correspond to t-values that are greater in absolute terms are ME and WH. The coefficient of ME is positive and takes the value 4.17 while that of WH is negative and takes the value -9.76. From tables 3 and 5 we find that the estimated coefficient capturing the impact of education on violent crime is negative in both regressions. This reflects that education has a beneficial social impact in that it reduces violent crime. However, the coefficient fails to be significant in both regressions. Thus, in case of the robust regression, we fail to reject our null hypothesis of poverty having no impact on violent crime. The conclusion is that we fail to find evidence of poverty causing violent crime. However, we did obtain evidence greater the metropolitan area of a state, greater are the incidences of violent crimes. Also, we found evidence that showed an increase in the number of whites does reduce the violent crime rates. However, it is very important to note that this should in no way be interpreted as non-white population are causes of crime. Since the our measure simply tracks the number of violent crimes per a given number of population, there is no way of knowing whether the victim was white or non-white. Thus, it is quite possible that this increase reflects a rise in hate crimes caused by whites on non-whites whenever the non-white population gets large enough to become a threat. Alternatively, it is also possible that non-whites participate in more violent criminal activities than whites in the states under consideration. However in all probability it is likely to be a combination of both. Thus, we cannot infer exactly what this implies and it begs greater investigation. Conclusions The objective of this research was to investigate how poverty impacts incidences of violent crime. We used 2005 state-wide data for crime across the United States. The central idea was that if increases in poverty led to greater incentives to commit violent crime, then increases in number of people living on incomes below the poverty level would lead to increases in the violent crime rate. To account for changes in violent crime caused by other factors such as differences in education, degree of metropolitan areas and racial dispersion we included these variables as controls in a regression set up that had violent crime rate as the dependent variable and poverty as the main independent variable. Our null hypothesis was that we should find that the coefficient on poverty would be zero implying no impact of poverty on violent crimes. Our preliminary OLS results showed that poverty did have a significant positive impact on violent crime. Thus the null hypothesis was rejected. We also found that differences in metropolitan area as well as racial differences (White vs Non-white population) also had significant influences. Education, as measured by the percentage of high school graduates in the population was found to have no significant influences on violent crime incidence. However, we also noted that the state DC was an obvious outlier and that led to the possibility of erroneous estimates because of the presence of outliers. Thus we turned to a robust regression approach to mitigate the effect of the outlier. The results of the robust regression however established that poverty did not have any significant impact on violent crime. Thus the robust regression did not provide evidence that would enable us to reject the null hypothesis. The analysis however provided further support to the impacts of variations in metropolitan population as well as number of Whites in the population. We failed to find any evidence for school education having any impacts on violent crime as well. Thus to conclude, our research fails to find any evidence that increased poverty causes violent crime rate to increase. We do find evidence of rises in metropolitan population increasing the crime rate and increase in the number of Whites is associated with lower violent crimes. However, one should be very careful in interpreting this last result. In all probability it simply reflects that uniformity of population results in fewer hate crimes and racial attacks either way. This is an interesting aspect that emerges and can be a vital topic for further research. APPENDIX A. The annotated STATA do file: **Inital Setting up** clear set memory 10m insheet using "D:\work\crimedata\crime2005.csv" **descriptive stats** summarize vi me wh hs po ** matrix of scatter plots** graph matrix vi me wh hs po **OLS regression** reg vi me wh hs po **Robust regression** rreg vi me wh hs po exit, clear B. The DATA STATE VIC VI2 MU ME WH HS PO AK 593 59 6 65.6 70.8 90.2 8 AL 430 43 7 55.4 71.4 82.4 13.7 AR 456 46 6 52.5 81.3 79.2 12.1 AZ 513 51 8 88.2 87.6 84.4 11.9 CA 579 58 7 94.4 77.2 81.3 10.5 CO 345 34 4 84.5 90.3 88.3 7.3 CT 308 31 3 87.7 85.1 88.8 6.4 DE 658 66 3 80.1 75.3 86.5 5.8 FL 730 73 5 89.3 80.6 85.9 9.7 GA 454 45 8 71.6 66.4 85.2 10.8 HI 270 27 2 91.5 26.5 88 7.4 ID 243 24 2 66.4 95.5 87.9 9.8 IL 557 56 7 87.8 79.4 86.8 8.5 IN 353 35 6 70.8 88.7 87.2 7.5 IO 272 27 2 61.1 95 89.8 6.9 KS 396 40 5 71.4 89.4 89.6 7.1 KY 262 26 5 55.8 90.4 81.8 14.2 LA 646 65 13 72.6 64.1 78.7 16.6 MA 469 47 2 91.4 87 86.9 7.5 MD 704 70 10 86.1 64.5 87.4 6.1 ME 109 11 1 40.2 97 87.1 7.6 MI 511 51 6 74.7 81.4 87.9 8.6 MN 263 26 3 70.9 89.8 92.3 5.6 MO 473 47 5 69.4 85.4 87.9 8.6 MS 326 33 9 48.8 61.3 83 16.4 MT 365 36 3 54.1 91 91.9 9.9 NC 455 46 6 60.2 74.1 80.9 10.7 ND 78 8 2 55.9 92.4 89.5 8.4 NE 289 29 3 69.8 92.1 91.3 8.2 NH 149 15 1 59.3 96.2 90.8 5.1 NJ 366 37 5 94.4 76.9 87.6 6.6 NM 665 66 6 75 84.7 82.9 14.8 NV 614 61 9 91.5 82.5 86.3 8.7 NY 465 46 5 87.5 73.9 85.4 10.7 OH 333 33 5 77.4 85.2 88.1 9.4 OK 506 51 6 65.3 78.6 85.2 12.4 OR 296 30 2 78.7 90.9 87.4 9.7 PA 398 40 5 77.1 86.2 86.5 8.2 RI 286 29 2 90.9 89 81.1 8.2 SC 794 79 7 60.5 68.3 83.6 11.3 SD 173 17 1 51.9 88.7 87.5 7.2 TN 688 69 7 63.6 80.7 82.9 10.6 TX 553 55 6 82.5 83.3 78.3 13.1 UT 249 25 3 88.2 93.8 91 7.6 VA 276 28 6 73 73.8 88.4 6.6 VT 110 11 2 38.2 96.9 90.8 6.4 WA 347 35 3 82 85.3 89.7 7.9 WI 221 22 3 68.3 90.2 88.8 7.2 WV 258 26 4 46.1 95.2 80.9 15.5 WY 262 26 3 65.1 94.7 91.9 7.3 DC 1608 161 44 100 37.4 86.4 18.5 Read More