final defense Flashcards
explain the really large increases at some sites
combination of real increases, some wonkiness in models, large negative weights, estimates of decline, small numbers. relative increases can look large when numbers are small, newer sites less implementation time so more sensitive to upticks in violence
how do you understand null or negative program effects
implementation fidelity, understaffing, hiring the wrong people, including people who are still “in the game”. more gang focused interventions have suggested that focusing on gangs could increase gang affiliation, though gangs less relevant in baltimore. impact of structural factors outside of safe streets control, violence that is out of their purview. shouldnt be held accountable for a beef that erupts between two people at a bus stop over nothing, esp. if not from their community
impact of program on youth vs. overall
more effective overall than among youth, similar that more effective homicides, older sites. more variation among youth why average effect not stat sig. half of sites effective/effective, half sites mismatched, no clear pattern emerged. focus on youth diminished, dont have formal caseload, accountable violence in all areas. old eval 90% mediation involved youth.
issues of aging staff, lower credibility, not on social media
difference between impact on NFS and H
general: homicide victims more likely to be older, shot once, shot in incident related to retaliation or drug trade. some randomness- issue of cm, time to hospital, quality of surgeon
homicide may be more likely to be premeditated and more experienced shooter, safe streets may be more likely to intervene
limitations to ascm models
small number of eligible donor pool neighborhoods, result in more negative weights and larger negative weights than “ideal” though no threshold for acceptable, though models not sensitive to SCM v ASCM
natural outcome volatility, small numbers- 3 mo moving averages, pretreatment fit old data for old sites! may not be representative, fiddled with things, used theory and rmspe to guide decision making, dont account any spillover effects
unlimited money! what research
survey of community members- opinions of the program, normative beliefs (how widespread, what are they), importance of events vs community investment.
include prolonged observation of SS work/embeddedness in a site
cross cities- a lot to learn from other cities, different implementation structure, different relationship/openness to police
top things for CBO/MONSE to take away?
key issues from staff: staffing, media, safety (vests)
