SAGE

So how can analysts who are not computer scientists or PhD-level data researchers incorporate this type of "evidence-based policy" into their work? Authors Eugene Bardach and Eric M. Patashnik offer these tips: Understand where to begin: Remember the opportunity cost of obtaining rigorous evidence. The value of the additional evidence equals the value of the outcome you would get from a decision with better evidence minus the value of the outcome you'd get from a decision without it. If your time is constrained, you can still consult registries of RCTs and other rigorous studies as part of the "Assemble Some Evidence" step to see if a policy similar to your proposal has been evaluated in the past. Know your data: It is always important for policy analysts to understand the data they are working with, but this is especially so when you are working with an administrative dataset that includes hundreds of thousands or even millions of observations. It is easy to mistake data quantity for data quality. Make sure to clean your data regularly and identify the characteristics of "good" and "bad" administrative datasets. Use administrative data and experiments to inform problem defi nition: Big data can inform problem defi nition by allowing the analyst to get a quick sense of what is happening on the ground. Experiments can help you evaluate the causal chains that go from situations to bad effects. Expand your option set and see constraints as learning opportunities: Machine learning and other methods can pinpoint factors associated with policy outcomes of interest. This can help you expand your menu of policy alternatives by identifying potentially low-hanging fruit–low-cost changes that might (although this would have to be confi rmed through more rigorous studies) make a big difference. Use data visualization to tell your story: Data visualization allows you to present key lessons to your client or audience in digestible parts. Rather than trying to present as much information as possible from large datasets, use visualization to show key relationships and dynamics and make your narrative come alive. Two of the most signifi cant trends in the fi eld of policy analysis are: (1) the growing use of randomized controlled trials (RCTs) to generate rigorous scientifi c evidence on the impact of policy interventions and (2) the increasing use of large, digitized administrative datasets (a type of "Big Data"). These trends have been encouraged by the creation of "policy labs" such as the Lab@DC, the Adul Latif Jameel Poverty Action Lab (J-PAL), and the California Policy Lab, which work closely with governments to evaluate public programs and inform future decisions. A Practical Guide for Policy Analysis: The Eightfold Path to More Effective Problem Solving, Sixth Edition Eugene Bardach and Eric M. Patashnik See page 12 for details. It is easy to mistake data quantity for data quality. Make sure to clean your data regularly and identify the characteristics of 'good' and 'bad' administrative datasets. Big data and policy analysis Award-winning authors Eugene Bardach and Eric M. Patashnik offer tips for analysts who wish to incorporate "evidence-based policy" into their work. Policy Analysis 3

• Cover