Data-Driven Decisionmaking & Decision Support Services │ Data Science

Rapid response evaluation and support for decisionmaking

“Big data” offer more options for the continuous measurement and assessment of data needed for effective rapid-cycle innovation and evaluation of policies and programs. Our data scientists and content experts help clients determine how to use that data through iterative research design and development. Through this process, we can devise interventions that are shown to work.

Data-driven decision making requires access to large amounts of heterogeneous data. A set of diverse methods—ranging from traditional statistical inference to machine learning, artificial intelligence, and text mining—ensures that these data can be processed and analyzed appropriately. At AIR, we work closely with our clients to build custom reports and visualizations to effectively communicate results and inform decision making.

Contact

Featured AIR Projects

Population Explorer (2009–Present)

USAID responds to humanitarian needs caused by natural and human-made disasters. While responses to emergency situations need to happen immediately, estimating the populations that are affected by such disasters can take months. To address the need for rapid and reliable small-area population estimates, AIR developed Population Explorer (www.populationexplorer.com). This web-based service allows anyone with access to the internet to estimate human populations, including age and gender breakdowns, anywhere in the world. Today, Population Explorer is used by USAID’s Famine Early Warning Systems Network as a front-line tool in disaster management and response. The tool is also used by the international private sector for market analysis.

Launch My Career

The American Institutes for Research designed Launch My Career to answer key questions about the return on investment for specific post-secondary education paths to help students identify programs that are likely to lead to success in the job market. Since its launch, Launch My Career has partnered with eight states to match state-specific education data from certificate and degree completers with information on wages maintained by state workforce officials. Using these data, we built consumer-facing web tools that allow students and other interested parties to explore and compare a variety of potential educational and career opportunities. AIR offers these web tools as a “software as service”, where AIR and are publicly available from any internet-enabled device. For the latest iteration of the tool, see Launch My Career Utah.

Microsimulation Model for Predicting Paid Leave

AIR developed an open-source microsimulation model to predict the use of paid leave and other labor market behavior under various policy scenarios. The model leverages the most up-to-date individual-level public data from the 2018 U.S. Department of Labor Family and Medical Leave Act employee survey and the American Community Survey. The model features a simulation engine driven by modern machine learning methods. It offers a user-friendly tool for states, federal agencies, and researchers to test different scenarios of paid leave programs, estimate the implications on costs in benefits paid out, and assess the costs of administering a paid family leave program.

In addition to the model itself, the DSAA team developed code to create, operate, and administer an open-sourced modeling platform with full capacity for modeling complex labor market behavior. Our model currently features a user-selectable full suite of microsimulation methods for simulating 13 different types of behaviors under different policy rules. Simulation methods offered include traditional and regularized logistic regressions, nearest neighbor-based classifiers, Naïve Bayes, support vector machine, decision tree, random forest, and XGBoost.