Before you demand a “data-driven” justice system, make sure you know who is actually behind the wheel.
Research evidence is not simply discovered; it is purchased by the government agencies and philanthropies that pay for evaluation research.
Evidence does not emerge from a pristine and impartial search for truth. It is subject to the political economy of research. Evaluations of justice interventions can be very expensive. Some statistical analyses may be conducted by academics operating on their own, but good evaluation research often requires new data collection and this can be expensive, usually requiring outside funding.
This means our evaluation evidence comes largely from investments made by policymakers, government agencies, and foundations. These investments are not free of bias. They reflect the goals, beliefs, and values of funding bodies as shaped by cultural, class-based, and racial biases. When we review the findings of evaluation research, we only see answers to the questions funding sources paid to ask. What sort of questions do you think they avoid, and how does the lack of such information affect justice policy and community well-being?
One of the most influential sources of bias relates to the level of measurement used in evaluation research. Most justice research tests interventions on individual behavior, although behavior is obviously shaped by social and community context. Are wealthy neighborhoods relatively free of violence because non-violent people decide to live there, or do structural and economic advantages lead to lower rates of violence? Reasonable answers to this question would suggest a range of public safety policies, but research funding tends to focus on one approach– individual interventions. In part, this may be due to the greater complexity of community-level evaluation designs. It is also clear, however, that foundations and elected officials tend to favor explanations of crime that point to individual pathology rather than inequality and injustice. It makes their jobs easier.