Appendix B: Detailed Statistical Results

This appendix provides greater detail about the equitable sharing results. The discussion assumes a working knowledge of quantitative research methods.

Table B1 reports estimates for equitable sharing payments received per person from the DOJ.  The coefficient on the share of forfeiture proceeds variable in Table B1 indicates that law enforcement agencies residing in generous forfeiture states receive significantly lower equitable sharing payments from the DOJ.  Although this standard interpretation is informative, the decomposition provides additional substantive information regarding the effects of the percent of forfeiture proceeds returned to law enforcement on proceeds received through equitable sharing payments (per person).  These results were discussed in the text above; the precise findings are presented in Table B1.

With respect to the legal hurdles faced by agencies in forfeiting drug-related assets, the coefficient on the state standard of proof variable is in the expected positive direction.  It should be pointed out, however, that the coefficient is not statistically significant at conventional significance levels.  On the other hand, the coefficient for the innocent owner variable is statistically significant in the expected positive direction.

Next, we explored the possibility of two-way interactions between the three forfeiture law variables.  We believe there is a strong theoretical basis to expect that the effects of any one forfeiture law variable on equitable sharing payments may be moderated by values for one of the other forfeiture law variables.  For example, agencies located in states where the standard of proof required by authorities to forfeit assets is greater (e.g., beyond a reasonable doubt) than the preponderance of evidence requirement at the federal level may be more inclined to turn forfeiture cases over to federal authorities even if state law permits agencies to keep a generous portion of the forfeited proceeds.  Conversely, agencies located in more generous states with similar standard of proof requirements for seizing assets may be less inclined to turn cases over to federal authorities.

The results of the interaction analysis are reported in Table B2.  These interactions must be interpreted with care.  Most importantly, the statistical significance of each law variable in isolation cannot be determined by looking at their t-statistics (coefficient divided by standard error) separately.  In other words, the fact that the coefficient for the innocent owner variable is not significant does not mean this aspect of forfeiture restrictiveness does not have a significant impact on equitable sharing payments.  Rather, the statistical significance of each law variable can only be determined when testing its importance in conjunction (referred to as a joint hypothesis test) with the interaction terms of which it is a part.  In this case, the F test of the joint hypothesis for both interactions involving the law variable measuring innocent owner burden were statistically significant.  To determine exactly which parts of the interactions were significant, the F tests were followed by more focused tests (often referred to as “simple slopes” tests).  The results of the tests indicated that the interaction between the state profit motive variable was significant regardless of whether state innocent owner statutes place the burden of proof on the property owner or the government.  On the other hand, the only part of the interaction that was significant between the state standard of proof and the innocent owner burden occurred when the burden of proof was placed on the government, i.e., the property owner is presumed innocent.  In addition, readers should note that each law variable is centered to facilitate interpretation of its marginal effect on equitable sharing payments when evaluated at different values for the other law variables.  Lastly, readers should be aware that the interpretations provided in the text above only apply to agencies reporting equitable sharing payments (i.e., the agency received at least one equitable sharing payment between fiscal years 2000 and 2004).  Readers interested in changes in the probability of receiving equitable sharing payments for those agencies not receiving payments should focus their attention on the coefficients presented in column (3).

Table B3 examines whether the interactions observed in Table B2 persist after controlling for potential confounding factors.  With respect to the control variables, the most notable finding is the negative, albeit nonsignificant association between the drug arrest rate and equitable sharing payments.  The most likely explanation is that agencies in high drug activity areas are also located in more generous forfeiture states.  On the other hand, the coefficient for the violent crime rate variable indicates agencies embedded in high crime areas receive significantly greater equitable sharing payments from DOJ.  Lastly, the dummy variable for agency type shows that municipal agencies receive larger equitable sharing payments than sheriff’s offices.

Of course, the most important results in Table B3 pertain to the importance of the interactions observed in Table B2 when addressing potential omitted variable bias.  The results for the two-way interaction terms between the innocent owner defense variable and the remaining two law variables remain largely unaffected when controlling for potential confounding factors.