Research of Note of July 2005

Print More

Firearm Violence Exposure and Serious Violent Behavior

University of Michigan/Science magazine
Abstract available free at, or contact American Association for the Advancement of Science at (202) 326-6440 for a copy of the article.

A University of Michigan researcher’s novel use of a complex statistical method called “propensity stratification” has provided evidence of a cause-and-effect relationship between adolescents’ exposure to firearm violence and their subsequent perpetration of violence.

Youth exposed to gun violence are twice as likely over the subsequent two years to perpetrate serious violence themselves, doctoral candidate Jeffrey Bingenheimer found in a study that is summarized in the May 27 issue of Science.

Bingenheimer used a subset of data from the Project on Human Development in Chicago Neighborhoods, led by Harvard professors Felton Earls and Robert Brennan, co-authors of his study. The Chicago study is following the lives of approximately 7,000 randomly selected urban youth, ages 18 and under, to uncover the roots of crime. Bingenheimer analyzed interview and demographic data collected in three waves over five years on more than 1,500 of those youth, ages 12 to 15.

Unable to control his subjects’ lives using a randomized model – the “gold standard” of scientific experimentation, in which subjects are randomly assigned to treatment or non-treatment groups – Bingenheimer instead opted to use propensity stratification. That is a less-used method that simulates randomization to strengthen estimates of causal relationships based on observed data.

Bingenheimer assigned each subject a propensity score. The scores ranged from 0 (no predicted probability of exposure to gun violence) to 1 (the greatest predicted probability of exposure), and were based on more than 150 personal and environmental variables, such as gender, impulsivity, substance use, family’s criminal history, reading proficiency and neighborhood disorder. The information on those variables was collected in the initial interviews with the youth and their families.

The subjects were then randomly divided (by lottery) into two groups of exposure propensity – nonexposure (in which subjects’ scores were subtracted from a score of 1) or exposure (in which subjects’ scores were unaltered) – and reassigned to one of 12 strata with scores of 0 to 1. In theory, this randomization controls for the effects of personal and environmental variables on later violent behaviors, allowing Bingenheimer to study any subsequent firearm exposure as an isolated variable.

Next, Bingenheimer looked at the second wave of the youths’ reports of exposure to firearm violence in the previous year. Exposure was defined as being shot or shot at, or seeing someone shot or shot at. Of those reporting, 23 percent had actually been exposed to gun violence, while 77 percent had not.

The researcher’s final assessment examined reports in the third wave about the youths’ perpetration of violent behavior in the previous year. Violence was defined as carrying a hidden weapon, attacking someone with a weapon, shooting someone, shooting at someone, or participating in a gang fight in which someone was hurt or threatened with harm. Of those reporting, 12 percent were classified as perpetrators, while 87 percent reported no perpetration.

Bingenheimer showed a persistent association between actual exposure to firearm violence and the perpetration of violence among adolescents who had similar predicted probabilities of exposure, independent of family and environmental variables.
“These findings suggest there is a substantial cause-and-effect relationship between exposure and perpetration,” Bingenheimer said, in a statement accompanying the study’s release. “Violence can be transmitted from person to person by means of exposure in the community.”

Understanding Foster Parenting: Using Administrative Data to Explore Retention

Office of the Assistant Secretary for Planning and Evaluation, U.S. Department of Health and Human Services
Available free at 68 pages.

This innovative federal study makes good use of often-overlooked administrative data from foster parent licensure, demographic and placement records to add to the little that is known about foster parent utilization and retention.

The study, based on data provided by child welfare agencies in New Mexico, Oklahoma and Oregon for the years from 1998 through 2003, reveals several significant insights:

• About 20 percent of foster families leave service each year.
• Infant foster children are typically cared for by young, urban and two-parent families, while adolescents are more often cared for by older, rural and single-parent families.
• “Occupancy rates” (the average number of foster children in a home on a given day) and “new placement” rates (the extent to which foster parents deal with new children over time) are highest among foster families that have two parents, are nonwhite, and are in rural and nonmetropolitan communities.
• The median length of foster parent service is eight to 14 months – a stark contrast to the five to eight years reported in previous retention studies, and shorter than the average child’s stay in foster care. In each state studied, nearly half to more than 60 percent of foster parents left service within a year of receiving their first placement.
• Short lengths of service cannot be attributed to a “burn-out” factor often associated with providing higher levels of care to infants, adolescents and children with special needs, or with having high numbers of foster children in the home. Those factors are consistently associated with longer lengths of service.
• Foster parents with longer service tend to be older and live in metropolitan areas.
• Twenty percent of available foster parents provided between 60 percent and 72 percent of all foster care days across the three states. This core group of foster parents may be overutilized by placement workers, who value their experience.

The researchers warn readers about three limitations to their study: that the experiences of the three states studied cannot be generalized to other groups of foster parents; that the analysis provides little information on why foster parents stay or leave service; and that the influences of child welfare workers’ decisions and the preferences of foster parents are too subtle to be revealed in administrative data.

Previous studies, including a 2002 assessment by the Office of the Inspector General at the U.S. Department of Health and Human Services, have blamed poor foster parent retention on unsatisfactory interaction with child welfare agencies, difficult behavior in foster children, and poor experiences with licensure training. (See, “Is This Any Way to Treat Foster Parents?” August 2002.)