NIJ's 2021 Recidivism Forecasting Challenge aimed to improve the ability to forecast recidivism using person- and place-based variables with the goal of improving outcomes for those serving a community supervision sentence.
As a condition of receiving their prize, each of the 25 winners[1] were asked to submit a research paper that describes which variables did and did not matter to the final forecasting model, and when applicable, what type of models outperformed other models.
The dashboard below allows you to filter down the winning submissions to ones that may be of particular interest to you. When using the dashboard:
- Selecting filters that they work with an “or” within a filter and an “and” between filters (e.g., if you select multiple types of models it will select papers that describe using either model, but if you also pick a variable from any of the variable dropdown it will look for papers that use any selected model AND any of the selected variables).
- Selecting the row of a submission, you will see a link to the “Winning Paper” in the popup.
Static and accessible lists of all winners and all winning papers also are available as is a downloadable spreadsheet data presented below.
NIJ has published a meta-analysis of the challenge results, see "A Synthesis of the 2021 NIJ Forecasting Challenge Winning Reports."
- Year Prize Won – Filter by year prize was won. Winners that won in multiple years submitted only one overall analysis.
- Data, sub-data, feature extraction – Filter by additional data, forms of feature engineering, or their algorithm processed by gender or race.
- Model – Filter by the specific model types used or tried. Most winners used (or tried) multiple models.
- Any Mention of a Variable – Filter if a variable was specifically identified as a top-5 variable in a specific model or in general.
- Variable (Year 1 Variables) – Filter if a variable was specifically identified as a top-5 variable in their year-1 forecast.
- Variable (Year 2 Variables) – Filter if a variable was specifically identified as a top-5 variable in their year-2 forecast.
- Variable (Year 3 Variables) – Filter if a variable was specifically identified as a top-5 variable in their year-3 forecast.
You may access form controls using your keyboard.
- Tab: Move you through the page from filter to filter.
- Space: Open a filter; select and unselect options
- Arrow keys: Move up and down within the filter.
- Escape (Esc): Close filter