How this translates
Research topics map to evidence. Systems topics map to workflows that ingest data, model changes, and generate outputs.
See systemsResearch
19Evidence and analysis of how policy, economics, and systems affect outcomes
AI Tutoring Outcomes
Do AI tutoring systems improve student learning compared with business-as-usual instruction and high-dosage human tutoring?
Example findings
-AI tutoring systems improve learning compared with business-as-usual instruction when they provide structured, adaptive practice in well-bounded domains.
-The key boundary between helpful AI tutoring and harmful AI assistance is whether the tool preserves productive student effort while providing timely support.
Sports Arena Impact Boundary
Do arenas reduce more impact through facility upgrades or event-day transportation and operations?
Example findings
-When sports-event accounting includes full Scope 3 activity, the largest climate gains often come from transportation and event-system design outside the facility envelope rather than from building systems alone.
-On-site facility improvements remain meaningful because they reduce the operator's direct and purchased-energy footprint, lower utility and water impacts, improve waste outcomes, and create auditable annual performance data.
Homeownership vs Renting: Is Owning Financially Advantaged?
When buying beats renting-and-investing the difference, and when it does not, under realistic US tax, cost, and regional assumptions
Example findings
-Whether homeownership is financially advantaged over renting in the US is a conditional question.
-The user cost of housing framework is the accepted analytical frame in tenure-choice economics.
Is Homeownership Tax-Advantaged?
What the research actually says about the MID, imputed rent, and who captures the housing tax subsidy
Example findings
-The Tax Cuts and Jobs Act of 2017 reduced the number of itemizing tax filers from approximately 46 million (TY2017) to approximately 16 million (TY2018), a drop of roughly 30 million filers driven by the near-doubling of the standard deduction and the $10,000 SALT cap.
-The share of US tax filers claiming the mortgage interest deduction fell from approximately 21% in TY2017 to approximately 8% in TY2018, a 13-percentage-point drop in a single year driven by the TCJA standard deduction change, not by any change in mortgage holding or mortgage rates.
The Science of Reading: What Works in Early Literacy Instruction
Evidence on phonics, structured literacy, and the instructional strands that support early reading for children ages 0 through K-2
Example findings
-Systematic phonics instruction in kindergarten and first grade has been associated with small-to-moderate gains in decoding and word reading compared with less-systematic or non-phonics instruction.
-Phonemic awareness has been consistently associated with later word reading, and short explicit phonemic awareness instruction linked to letters has been associated with gains in decoding for beginning readers.
Oral Language, Vocabulary and Comprehension in Early Literacy
Evidence on the non-decoding strands of early literacy, including caregiver talk, vocabulary development, and the word-gap debate
Example findings
Early Literacy Assessment: Screening, Benchmarks and Dyslexia Detection
Evidence on DIBELS, universal screening, dyslexia identification, progress monitoring, and the validity of early literacy measures
Example findings
-Oral reading fluency measures have been associated with moderate to strong prediction of early grade reading comprehension, with reported correlations often in the 0.
-Multi-stage or gated screening models, in which children flagged by an initial brief screener are reassessed through dynamic assessment or short-term progress monitoring, have been associated with lower false positive rates than single-gate screening.
Play-Based Learning vs Direct Instruction in Early Childhood
Evidence on the relative effectiveness of guided play, free play, and direct instruction for young children
Example findings
-Academic gains from preschool programs have frequently faded by first or third grade, even when other effects persist.
-US kindergarten classrooms have become substantially more academic and teacher-directed since the late 1990s, with less time for play, art, and child-initiated activity.
The Developmental Science of Play
Cognitive, social, and regulatory functions of play in young children
Example findings
-Juvenile rats deprived of play with peers show altered prefrontal cortex development and impaired adult social behavior, providing the strongest causal evidence for play's developmental functions in any species.
-Preschoolers' exploratory play is sensitive to evidence quality and can support causal inference, with children targeting ambiguous or confounded contexts for fuller exploration.
Children's TV, Film and Early Literacy
Evidence on how children's television and film affect early literacy, vocabulary, and learning outcomes
Example findings
-Regular preschool exposure to curriculum-based educational programming such as Sesame Street has been associated with small-to-moderate gains in vocabulary, letter recognition, and other school-readiness skills.
-Children under roughly two and a half years generally learn less vocabulary and less transferable content from pre-recorded video than from equivalent live interaction.
Digital Apps, E-Books and Touchscreen Learning in Early Childhood
Evidence on interactive digital media, e-books, and adaptive apps for early literacy
Example findings
-Educational apps for young children vary widely in quality and most top-downloaded apps score low on learning-science criteria.
-Manipulative design features are common in popular children's apps and may shift time away from learning content.
In-Person Children's Programming: Libraries, Preschool and Community Programs
Evidence on library storytimes, preschool programs, home visiting, and other in-person literacy interventions
Example findings
-High-quality center-based preschool programs have been associated with short-term gains on language and early literacy measures in the range of 0.
-Early Head Start has been associated with small positive effects on toddler language and cognitive outcomes at age 3 (around d = 0.
Home Literacy Environment and Parent-Child Interactions
Evidence on shared reading, caregiver talk, book access, and the home as a literacy-relevant environment
Example findings
Emerging Interventions Beyond Traditional Phonics
Evidence on high-dosage tutoring, state structured literacy reform, and dyslexia-specific interventions
Example findings
-High-dosage tutoring programs with small groups, multiple weekly sessions, and in-school delivery have been associated with moderate average effect sizes on reading outcomes in randomized trials.
-Phonics-based structured literacy interventions have been associated with statistically meaningful gains in reading and spelling for students with dyslexia or severe reading difficulties.
Benefit Cliff Insights
How benefit cliffs affect labor force participation in rural vs urban environments
Example findings
-Some studies observe workers approaching benefit eligibility thresholds constraining their hours or declining overtime to keep earnings below cliff cutoffs.
-When a worker simultaneously risks losing two or more benefits (e.
Administrative Burden & Workforce Participation
How paperwork, recertification, and procedural complexity in benefit programs affect labor force participation
Example findings
-During the 2023-2024 Medicaid unwinding, approximately 69% of the 25+ million disenrollments were procedural rather than eligibility-based, suggesting that the majority of coverage losses were driven by administrative failure to complete renewal processes rather than genuine changes in eligibility.
-Experimental evidence from LA County found that one-third of SNAP applications were denied due to missed interviews, five times the number denied for ineligibility.
How Income Thresholds Affect Work Decisions
Evidence on how eligibility cutoffs, phase-outs, and benefit cliffs in public programs influence labor supply, household decisions, and economic mobility
Example findings
-Research consistently finds that income thresholds in benefit programs have larger effects on the extensive margin (whether to work) than the intensive margin (how much to work).
-Observable income bunching at benefit thresholds is limited and concentrated among specific populations.
Wage Growth vs Total Compensation
Why a raise may not increase take-home income, and how benefit phase-outs, health insurance costs, and effective marginal tax rates shape real compensation
Example findings
-Effective marginal tax rates for low- and moderate-income workers range from near-zero to over 80%, with some household configurations exceeding 100%.
-When multiple benefit programs phase out across the same income range, the combined effective marginal tax rate exceeds what any individual program designer intended.
Work Incentives in Low- and Middle-Income Households
How the tax-benefit system shapes labor supply decisions for low- and middle-income households
Example findings
-Research consistently finds that the extensive margin of labor supply (whether to work at all) is more responsive to tax-benefit policy than the intensive margin (how many hours to work).
-Optimization frictions, including fixed work schedules, adjustment costs, and imperfect information, help explain why micro-econometric studies find small labor supply elasticities while macro-level evidence suggests larger responses.
Systems
5How workflows are executed and how they can be automated
Manual Workflows at Scale
Evidence, failure modes, and system outcomes for manual coordination, data entry, and approval processes
Key claims
-Industry and analyst research consistently shows that inefficient manual processes are not marginal.
-Empirical case studies show dramatic performance differences between manual and automated execution in structured, repetitive tasks.
AI Impact on Reporting Workflows
Evidence on how AI tools are changing data aggregation, report generation, and compliance reporting across organizations
Key claims
-Industry surveys consistently show high AI adoption rates across organizations, but deep workflow integration remains uncommon.
-A Federal Reserve working paper tracking U.
Document Processing & Data Extraction Automation
Evidence on AI-driven extraction from PDFs, invoices, forms, and unstructured documents compared to manual data entry
Key claims
-Academic benchmarks and case studies show substantial variance in extraction accuracy depending on the approach and document complexity.
-Multiple case studies from named organizations report order-of-magnitude reductions in document processing time after implementing AI-driven extraction.
Email and Task Automation in Operations
Evidence on how organizations convert high-volume email into structured tasks, and the productivity costs of email-driven workflows
Key claims
-Multiple large-scale studies converge on a consistent finding: email occupies a disproportionate share of work time.
-Research documents that email has evolved beyond communication into an informal task management platform.
Data Fragmentation & Operational Inefficiency
Evidence on how data silos, disconnected systems, and fragmented data sources create operational costs and productivity losses
Key claims
-Multiple studies converge on a consistent finding: workers spend a substantial portion of their day locating information rather than using it.
-The economic costs of data fragmentation are documented at multiple scales.