B.2. Relación de lo interpretado con los propósitos de la investigación
2. En el contexto ‘Cu-2’: veía, y me preguntaba qué era lo que hacía que se aprendiera más en un centro que en otro (caso de CEIP B.)
The only significant change made by applying the longer data set is the effect of real earnings on fraud: while the effect is positive earlier reflecting the opportunity effect, it becomes negative and significant when using the longer data set picking up its motivation effect. As we have observed the same change in the analyses of burglary, we will explain such phenomenon together in later discussion.
The GMM estimation shows that the once-lagged crime rate is correlated with contemporary crime rate at a rate of 0.30, which is coherent with the previous finding.
Furthermore, the over-identification test has also confirmed the validation of employed instruments.
4.5.4 Robbery
According to the Home Office, robbery has been categorized as violent crime because the occurring of such crimes is usually accompanied with physical harm implied onto
the victims. However, it cannot neglect that the motivation of committing such crimes is usually of financial interest. Therefore, robbery is expected to respond in similar way to the independent variables as do with property crimes. The estimation results based on the primary data set 1992-2005 are reported as following in table 4-13.
Table 4-13
*** significant at 1% level; ** significant at 5% level; * significant at 10% level.
The J-statistic is computed for the Sargan/Hansen over-identifying restrictions. Under the null hypothesis that the over-identifying restrictions are valid, the J-statistic follows a Chi-Squared distribution with the degree of freedom being the difference between the instrument rank and the number of coefficients estimated. The reported over-identification test is the corresponding p-value.
The first thing to notice in table 4-13 is that both detection rate and prison population have displayed negative and significant impacts on robbery as law enforcement instruments. Whilst the coefficient of detection rate is constantly negative and significant in all three estimations, the effect of prison population is insignificant in both the OLS and fixed-effect estimations and becomes highly significant (at 1% level) once the relevant estimation issues are controlled by the GMM technique.
Meanwhile, we also find that both Gini coefficient and the proportion of young people present positive and significant correlations with robbery, thus supporting our expectations. As we have predicted that an increase in the income inequality will increase crimes by reducing the opportunity cost of committing crimes for the people at the lower end of wealth distribution, our results have indeed reinforced such predictions. Furthermore, the proportion of young people could positively affect robbery through two channels: on the one hand, higher proportion of young people increases the number of motivated robbers given the lower opportunity cost of young population; on the other hand, higher proportion of young people also means more potential targets for robbery. Due to its nature, robbery often occurs on the streets of less affluent areas. In addition, the targets of robbery are usually the valuable belongings carried by passengers and pedestrians, such as mobile phones, ipods, laptops, wallets and purses and so on. Consequently, for robbers, young people are usually seen as attractive targets since they are relatively less cautious and more likely to possess trendy electronic gargets. As a result, higher proportion of young people increased the number of both potential robbers and potential victims, hence, significantly increases the crime rate of robbery.
Whilst unemployment rate shows negative and significant correlation with robbery, real earnings exhibits positive and significant effect on the same crime. Both of them have picked up their opportunity effects. In particular, as long as the potential estimation issues are controlled, higher unemployment rate will reduce the crime rate of robbery, reflecting the fact that robbery primarily occurs on the streets and, the more people being unemployed, the fewer commutes would be required. Thus, it
provides fewer opportunities for potential robbers. Whereas, with higher income, people tend to make more purchases, thus provide more opportunities for robbery.
The coefficient of once-lagged robbery rate is positive and significant at 1% level.
The magnitude of 0.44 suggests quite strong persistence in robbery which is again consistent with expectation. Meanwhile, the over-identification test has confirmed that the instruments employed for the endogenous variables in the GMM estimation are indeed valid and independent from the error term.
Table 4-14 below shows the counterpart results generated by the data set 1987-2005.
Table 4-14
*** significant at 1% level; ** significant at 5% level; * significant at 10% level.
The J-statistic is computed for the Sargan/Hansen over-identifying restrictions. Under the null hypothesis that the over-identifying restrictions are valid, the J-statistic follows a Chi-Squared distribution with the degree of freedom being the difference between the instrument rank and the number of coefficients estimated. The reported over-identification test is the corresponding p-value.
As shown in table 4-14, the estimation results based on the data set 1987-2005 are quite consistent with previously findings, confirming the robustness of our results.
Especially, both detection rate and prison population have shown negative correlations with robbery indicating their deterrent (and probably also incapacitation) effects on crime as law enforcements. Gini coefficient displays positive and significant correlation with robbery implying that higher income inequality will increase robbery through lowering the opportunity cost to the least affluent.
Additionally, both unemployment rate and real earnings have picked up their opportunities effects on robbery: higher unemployment rate will reduce the opportunities and hence the crime rate for robbery; while higher income will increase the opportunities for potential robbers and hence increase the crime rate of robbery.
The only exception is the proportion of young people which has obtained unstable coefficients across estimations. The correlation between the proportion of young people and robbery is positive in both the OLS and fixed-effect estimations. Once the GMM estimation is applied, the coefficient of young people turns into negative and significant at 1% level.38 Therefore, the coefficient changing is probably as a result of the inclusion of extra 5 years in the data set 1987-2005. In this case, the coefficient of young people is rather sensitive to the period under examination.
The GMM estimation based on the data set 1987-2005 has confirmed the significant self-correlation in robbery. In addition, the over-identification test has proved the validity of using lagged values as instruments for the endogenous variables.
38 By restricting the data set 1987-2005 to only include the period 1992-2005, the coefficient of young