To measure the impact of a local homicide, Sharkey compared the test scores of children who were assessed directly after a homicide in their neighborhood with other children in the same neighborhood who were assessed at different times. He took into account three geographic areas of increasing size: "block groups," which are small sets of city blocks that have about 1,500 residents; "census tracts," which are slightly larger and have about 4,000 residents; and "neighborhood clusters," which have about 8,000 residents. Because all comparisons were made among children living within the same neighborhoods, the analysis can be thought of like an experiment in which some children are randomly picked to be assessed in the days following a local homicide while other children are picked to be assessed at a different point, further removed from the date of the homicide.
Sharkey examined separately the potential impact of a local homicide on African Americans and Hispanics. Whites and other ethnic groups were excluded because they were almost never exposed to local homicides in the samples Sharkey used for his study. ...I have no problem believing that living in a bad neighborhood is a risk factor for being a bad reader, but this study mainly tells me how resilient kids are. Somebody was murdered in their neighborhood a couple of days ago, and we give them a test on reading and verbal skills, and in a week or so, they're back to normal.
Overall, the results showed that African-American children who were assessed directly after a local homicide occurred scored substantially lower than their peers who live in the same neighborhood, but were assessed at different times. As the duration of time between the homicide and the assessment increased beyond a week, the estimated effects of homicides faded away. Further, as the distance between the child's home and the location of the homicide widened, the impact of the homicide became weaker. While the results were extremely strong for African Americans, there was no effect of local homicides for Hispanics—a finding that Sharkey plans to explore in ongoing research. Very similar patterns were found when the analysis was replicated in the second dataset, the Three City Study. Again, the effects of local homicides were extremely strong for African Americans and non-existent for Hispanics. (My bold.)
It also means that, if schools want to find out a kid's true reading ability (especially after No Child Left Untested), they should be flexible in the testing schedule, and not test kids when someone has been killed in their neighborhood in the last week or so.