A critical question regarding the expected utility of a national registry is the number of interstate perpetrators that might be expected each year. As the Interim Report to Congress pointed out, no such estimates had been established, and what information did exist was anecdotal. To fill this gap, national estimates of interstate perpetrators were developed. This examination of prevalence also examined the seriousness of maltreatment by instate and interstate perpetrators. Estimates of the proportion of interstate perpetrators identified in geographically adjacent States were also produced.
Twenty-two States contributed data for the prevalence study by contributing encoded names, first initial, and date of birth information for all substantiated adult perpetrators from 2005-2009. (See table 1.) These data were appended to individual records from the National Child Abuse and Neglect Data System (NCANDS), which includes additional perpetrator characteristics, victim information, and information on type of maltreatment. (See appendices B and C for more details on the prevalence study.)
Using data for those 22 States, data were compared from 2009 records from each State to records for all other participating States from 2005-2009. The goal was to estimate the number of interstate perpetrators that would be identified by States in a given year through a national registry. The 5-year time period was assumed to be adequate for identifying most interstate perpetrators that would be found through a national registry, though many State registries maintain records for longer than 5 years. Several matching algorithms were explored, and an optimal algorithm was identified (see more detailed discussion below).
Estimates of the number of matches for each participating State were adjusted upward to account for matches that would have come from nonparticipating States, using interstate migration data from the 2000 Census. To create estimates for nonparticipating States, Census-based interstate migration rates were applied to aggregate data on the total number of substantiated child maltreatment perpetrators in 2009.
The Matching Process
The Adam Walsh Act limits the case-specific content of a national registry to the name of the perpetrator and the type of substantiated maltreatment. In order to explore the capability of such a registry to identify interstate perpetrators, matches were carried out using an encoded last name and first initial only. Encoded names were used in this exercise for two reasons. First, it allowed participating States to supply name information without breaching confidentiality restrictions. Limiting first name information to first initial was done for the same reason. Second, the encoding allowed for matching names with different though similar spellings. It is a common practice in record matching, and it is likely that a national registry would use a similar technique in order to capture minor variations in spelling that inevitably occur in such records.
The results of the matching process show that 88.6 percent of all records in 2009 found a match in one or more States for the previous 5 years. Of those that did match, 85.1 percent found matches in three or more States. This is an overestimate of matches due to the use of only the encoded last name and first initial since many names are similar or the same. Therefore, using the encoded last name and first initial only is inadequate as a matching algorithm for purposes of developing national estimates. If complete names are used on a national registry it would substantially reduce the percentage of matches; however, it is very likely that even full names would result in too many matches to be of practical use to States. (See tables 23 and 24.)
By adding sex and date of birth to the matching algorithm, the number of matches was reduced dramatically to 2,022 across the 22 participating States, representing 0.7 percent of all substantiated perpetrators in 2009. Of those 2,022 matches, only 44 included matches in more than one State, and only 345 matched more than two other records. (See table 25.) Across the 22 States, matching rates were fairly consistent, ranging from 0.6 percent in California to 2.0 percent in Nevada and Wyoming. These rates are highly likely to be lower than would be found if data for all States were available. Adjustments to account for those missing States would roughly double the matches from 0.7 percent to 1.5 percent.
While there is no way to confirm the accuracy of this estimate, it is reasonable given what is known about revictimization and reperpetration rates over similar time periods. If one assumes that about 16.7 percent of all substantiated perpetrators will reoffend within 5 years, and that they move between States at a rate similar to the general population in 2000 (8.9 percent), then one might expect an interstate match rate of about 1.5 percent.,,
National Estimates of Interstate Child Maltreatment Perpetrators
As noted in the preceding section, a reasonable formula or algorithm for matching perpetrator records across States was developed using encoded last name, first initial, sex, and date of birth of the perpetrator. Using this approach to matching records, State and national estimates for interstate child maltreatment perpetrators were produced based on a model that combines this prevalence study data with Census interstate migration data.
The modeling exercise is complex, involving four steps. First, the number of matches for each State participating in the prevalence study was estimated relative to other participating States. Second, these estimates were adjusted upward to account for perpetrators from nonparticipating States using 2000 Census data, based on the percentage of all interstate in-migrants in each State who were from States that did not participate in the prevalence study.
The results of this second step represent complete estimates of the number of interstate perpetrators in each of the 22 States that participated in the prevalence study. With this adjustment, the total number increased from 2,022 to 4,216. The number of interstate perpetrators varied considerably by State, from highs of 785 (New York), 712 (Texas), and 572 (California), to as few as 15 (Wyoming) and 16 (New Hampshire). (See table 26.)
The third step provided estimates of interstate perpetrators for the 29 States that did not provide data for the prevalence study. The estimates used the aggregate number of substantiated perpetrators for 2009, and multiplied that by an estimated interstate perpetration rate that had been constructed for each State based on an average rate derived from the participating States, and adjusted using Census interstate migration rates particular to that State.
Fourth, adding all of the State-specific estimates together yielded a national estimate of 7,852 interstate perpetrators for 2009. Based on a total of 512,790 unique perpetrators in the U.S. in 2009, interstate perpetrators represented 1.5 percent of all substantiated perpetrators.
An important caveat to these estimates is the fact that there are undoubtedly false positives resulting from the matching algorithm, which, all else being equal will result in an over-estimate of the national incidence of interstate perpetrators. While there is no direct way to test this, it is possible to examine what happened to the number of interstate matches when the requirement that the birth date of at least one child in the case records match was added. When this was done, only 27 percent of the matches used to create the national estimates matched. There are many reasons why a true match might be lost with this additional requirement including inaccurate date of birth information, and the fact that fewer than one-half of all States include all children living in the household in their records as a regular practice. Still, this finding suggests that the actual number of interstate perpetrators at the national level may be less than the 7,852 estimated here, though there is no way to know how much less.
Most States already have working relationships with the child welfare agencies in their neighboring States. If a large proportion of matches come from only a few neighboring States the burden on States to make such inquiries may not be very great. To explore this issue, the small number of participating States in the prevalence study that include adjacent States also in the study and contributing at least 75 percent of their overall interstate migration flow were examined. These included California, Texas, Louisiana, Maine, and Arizona.
For these States, the percentage of all interstate perpetrators from neighboring States was estimated as follows: Louisiana (40%), Arizona (35.8%), Texas (15.4%), California (12.4%), and Maine (9.5%). To the extent that there are false positive matches in the estimates, these percentages may be underestimated, since false matches are likely to be spread more evenly across the country. Even taking this into account, however, it seems likely that, for most States, a majority of interstate perpetrators do not come from adjacent States, though clearly there are important differences across States.
Instate and Interstate Perpetrator Comparisons
If interstate perpetrators engage in more serious forms of abuse and neglect, this would provide additional evidence of the utility of a national registry, since it would be identifying more serious cases. To examine this issue, the study compared instate and interstate perpetrators on four outcomes—type of child maltreatment, whether one or more children was removed from the home, whether there were any court petitions, and any child fatalities—all for 2009. Analyses used the same data file that was used to create the national estimates, the prevalence data file.
For type of maltreatment, the typology used for establishing seriousness was least serious to most serious as follows: neglect, medical neglect, emotional maltreatment, physical abuse, and sexual abuse. Each perpetrator was assigned a single value representing the most serious form of abuse or neglect for which he or she was substantiated. The results show very similar patterns across the two groups, with nearly two-thirds (64.7 percent) in each group substantiated for neglect. Chi square analyses indicate no significant difference in type of maltreatment between instate and interstate perpetrators. (See table 27.)
Examining whether a child was removed from the home for 24 hours or more, however, resulted in different results. Thirty percent of interstate perpetrators included a child removed from the home in their current State of residence compared to just more than 20 percent among instate perpetrators. This difference is statistically significant at the .001 level. (See table 28.) Similar results were observed for court involvement, with rates of 28 percent for interstate perpetrators and about 19 percent for intrastate perpetrators. (See table 29.) Finally, interstate perpetrators do not appear any more likely to be associated with a child death than intrastate perpetrators, though the small number of child deaths associated with interstate perpetrators, 4, was too few to support any tests of statistical significance. (See table 30.)
Clearly, interstate perpetrator cases are more serious in that children are more likely to be removed from the home. To some extent this is expected, since interstate perpetrators have by definition been substantiated as perpetrators at least two times, while instate perpetrators may or may not have had prior substantiations. While they may represent more serious cases, they do not appear to be more likely to be sexual predators or to exhibit different overall patterns of maltreatment.