Emerging Technology
day one project

Incorporate open science standards into the identification of evidence-based social programs

02.09.24 | 5 min read | Text by Sean Grant

Evidence-based policy uses peer-reviewed research to identify programs that effectively address important societal issues. For example, several agencies in the federal government run clearinghouses that review and assess the quality of peer-reviewed research to identify programs with evidence of effectiveness. However, the replication crisis in the social and behavioral sciences raises concerns that research publications may contain an alarming rate of false positives (rather than true effects), in part due to selective reporting of positive results. The use of open and rigorous practices — like study registration and availability of replication code and data — can ensure that studies provide valid information to decision-makers, but these characteristics are not currently collected or incorporated into assessments of research evidence. 

To rectify this issue, federal clearinghouses should incorporate open science practices into their standards and procedures used to identify evidence-based social programs eligible for federal funding.

Details

The federal government is increasingly prioritizing the curation and use of research evidence in making policy and supporting social programs. In this effort, federal evidence clearinghouses—influential repositories of evidence on the effectiveness of programs—are widely relied upon to assess whether policies and programs across various policy sectors are truly “evidence-based.” As one example, the Every Student Succeeds Act (ESSA) directs states, districts, and schools to implement programs with research evidence of effectiveness when using federal funds for K-12 public education; the What Works Clearinghouse—an initiative of the U.S. Department of Education—identifies programs that meet the evidence-based funding requirements of the ESSA. Similar mechanisms exist in the Departments of Health and Human Services (the Prevention Services Clearinghouse and the Pathways to Work Evidence Clearinghouse), Justice (CrimeSolutions), and Labor (the Clearinghouse for Labor and Evaluation Research). Consequently, clearinghouse ratings have the potential to influence the allocation of billions of dollars appropriated by the federal government for social programs. 

Clearinghouses generally follow explicit standards and procedures to assess whether published studies used rigorous methods and reported positive results on outcomes of interest. Yet this approach rests on assumptions that peer-reviewed research is credible enough to inform important decisions about resource allocation and is reported accurately enough for clearinghouses to distinguish which reported results represent true effects likely to replicate at scale. Unfortunately, published research often contains results that are wrong, exaggerated, or not replicable. The social and behavioral sciences are experiencing a replication crisis as a result of numerous large-scale collaborative efforts that had difficulty replicating novel findings in published peer-reviewed research. This issue is partly attributed to closed scientific workflows, which hinder reviewers’ and evaluators’ attempts to detect issues that negatively impact the validity of reported research findings—such as undisclosed multiple hypothesis testing and the selective reporting of results.

Research transparency and openness can mitigate the risk of informing policy decisions on false positives. Open science practices like prospectively sharing protocols and analysis plans, or releasing code and data required to replicate key results, would allow independent third parties such as journals and clearinghouses to fully assess the credibility and replicability of research evidence. Such openness in the design, execution, and analysis of studies on program effectiveness is paramount to increasing public trust in the translation of peer-reviewed research into evidence-based policy.

Currently, standards and procedures to measure and encourage open workflows—and facilitate detection of detrimental practices in the research evidence—are not implemented by either clearinghouses or the peer-reviewed journals publishing the research on program effectiveness that clearinghouses review. When these practices are left unchecked, incomplete, misleading, or invalid research evidence may threaten the ability of evidence-based policy to live up to its promise of producing population-level impacts on important societal issues.

Recommendations

Policymakers should enable clearinghouses to incorporate open science into their standards and procedures used to identify evidence-based social programs eligible for federal funding, and increase the funds appropriated to clearinghouse budgets to allow them to take on this extra work. There are several barriers to clearinghouses incorporating open science into their standards and procedures. To address these barriers and facilitate implementation, we recommend that:

  1. Dedicated funding should be appropriated by Congress and allocated by federal agencies to clearinghouse budgets so they can better incorporate the assessment of open science practices into research evaluation.
    • Funding should facilitate the hiring of additional personnel dedicated to collecting data on whether open science practices were used—and if so, whether they were used well enough to assess the comprehensive of reporting (e.g., checking publications on results with prospective protocols) and reproducibility of results (e.g., rerunning analyses using study data and code).
  2. The Office of Management and Budget should establish a formal mechanism for federal agencies that run clearinghouses to collaborate on shared standards and procedures for reviewing open science practices in program evaluations. For example, an interagency working group can develop and implement updated standards of evidence that include assessment of open science practices, in alignment with the Transparency and Openness Promotion (TOP) Guidelines for Clearinghouses.
  3. Once funding, standards, and procedures are in place, federal agencies sponsoring clearinghouses should create a roadmap for eventual requirements on open science practices in studies on program effectiveness.
    • Other open science initiatives targeting researchers, research funders, and journals are increasing the prevalence of open science practices in newly published research. As open science practices become more common, agencies can introduce requirements on open science practices for evidence-based social programs, similar to research transparency requirements implemented by the Department of Health and Human Services for the marketing and reimbursement of medical interventions. 
    • For example, evidence-based funding mechanisms often have several tiers of evidence to distinguish the level of certainty that a study produced true results. Agencies with tiered-evidence funding mechanisms can begin by requiring open science practices in the highest tier, with the long-term goal of requiring a program meeting any tier to be based on open evidence.

Conclusion

The momentum from the White House’s 2022 Year of Evidence for Action and 2023 Year of Open Science provides an unmatched opportunity for connecting federal efforts to bolster the infrastructure for evidence-based decision-making with federal efforts to advance open research. Evidence of program effectiveness would be even more trustworthy if favorable results were found in multiple studies that were registered prospectively, reported comprehensively, and computationally reproducible using open data and code. With policymaker support, incorporating these open science practices in clearinghouse standards for identifying evidence-based social programs is an impactful way to connect these federal initiatives that can increase the trustworthiness of evidence used for policymaking.

To learn more about the importance of opening science and to read the rest of the published memos, visit the Open Science Policy sprint landing page.